Wednesday, 2015-07-29

*** vilobhmm has quit IRC00:00
thingeecool I don't feel as dumb00:00
*** merooney has quit IRC00:02
*** hodos has quit IRC00:03
*** merooney has joined #openstack-cinder00:11
*** jasondotstar has quit IRC00:11
*** merooney has quit IRC00:11
*** merooney has joined #openstack-cinder00:12
*** _cjones_ has quit IRC00:14
*** merooeny_ has joined #openstack-cinder00:14
thingeewinston-d_zZZ: thinking of something like
*** merooney has quit IRC00:16
*** jasondotstar has joined #openstack-cinder00:17
*** merooeny_ has quit IRC00:18
*** xek_ has joined #openstack-cinder00:19
*** xek has quit IRC00:20
*** vokt has quit IRC00:20
*** IlyaG has joined #openstack-cinder00:23
*** esker has quit IRC00:23
*** david-lyle has quit IRC00:24
thingeescottda: you still around?00:28
*** tsekiyam_ has joined #openstack-cinder00:32
*** xyang1 has quit IRC00:33
*** changbl has joined #openstack-cinder00:34
*** mtanino has quit IRC00:34
*** Yogi1 has quit IRC00:34
*** tsekiyama has quit IRC00:35
*** garthb has quit IRC00:35
thingeescottda: how hard would it be to have one of the people that brings in coffee to instead bring an empty thermos dispensers? If people are interested, I could make coffee :)00:35
*** tsekiyam_ has quit IRC00:36
scottdaI'm not sure...I'll ask tomorrow.00:37
scottdaIf they cannot, I've one, and could see about scrounging up some others.00:37
scottdaIt is packaged with pastries/fruit, so I'm not sure if we can just cancel the coffee.00:38
jungleboyjOoooh, real thingee coffee!  :-)00:38
thingeewhoa I think my messages didn't make it over cause I switching to vpn...00:38
scottdaI'll ask catering, and we can see if people care about the food part at tomorrow's meeting, if that sounds good?00:38
thingeescottda: I'm talking about that hipster coffee stuff. pourover, fresh beans and just grinded.00:39
thingeescottda: all I need is a single outlet for the kettle, dispenser and water.00:39
*** albertom has quit IRC00:39
thingeejust so people are happy and productive :)00:39
scottdawell, there will certainly be an outlet.00:39
thingeet have answer for tea people though00:40
scottdaso, you still want one or more thermos to hold the coffee?00:40
thingeescottda: heh just an idea, if it's too much trouble then we can just use the nearby cafe or whatever00:41
*** leeantho has quit IRC00:41
*** albertom has joined #openstack-cinder00:42
thingeescottda: so I usually use a 3liter canister...I can use whatever they have.00:42
scottdaWell, I've wrangled kmartin and hemna to arrange for coffee and pastries brought in. We could cancel or change, and/or I could bring a thermos or see if they have an extra for "custom coffee".00:42
thingeescottda: ah that's fine.. I don't want to be a trouble on canceling things00:43
scottdaI think they generally use 3L canisters.00:43
scottdaI don't think we have arrangements for the afternoon either, so one way or another some good coffee will be welcome.00:43
*** Yogi1 has joined #openstack-cinder00:43
scottdaI'll work on the logistics tomorrow, when catering is around to answer questions.00:43
thingeeI have a bunch of canisters I use for burning man, but fitting on those in my bag would take up a bit of room, especially since this won't be my first stop.00:44
scottdanp, we'll figure it out.00:44
openstackgerritwanghao proposed openstack/cinder: Fixing notify message of manage_existing flow
*** xek_ has quit IRC00:46
*** xek_ has joined #openstack-cinder00:47
*** zhenguo has joined #openstack-cinder00:57
*** IlyaG has quit IRC00:58
*** Yogi1 has quit IRC00:59
*** zhonghua-lee has joined #openstack-cinder00:59
*** garthb has joined #openstack-cinder01:00
*** garthb_ has joined #openstack-cinder01:00
*** garthb_ has quit IRC01:00
*** laughterwym has joined #openstack-cinder01:07
*** lixiaoy1 has joined #openstack-cinder01:12
*** jasondotstar has quit IRC01:12
*** davechen has joined #openstack-cinder01:15
*** zhonghua-li has joined #openstack-cinder01:16
*** Liu has joined #openstack-cinder01:16
lixiaoy1jdurgin: Hi Josh, could like to talk with you ceph backup.OK?01:17
*** jasondotstar has joined #openstack-cinder01:17
*** zhonghua-lee has quit IRC01:19
*** LiuXinguo has quit IRC01:19
lixiaoy1Or anyone know that why we need to do full restore in ceph if source volume is same as destination volume:
*** r-daneel has quit IRC01:19
*** salv-orlando has joined #openstack-cinder01:20
*** heyun has joined #openstack-cinder01:21
*** jasondotstar has quit IRC01:22
*** breitz has quit IRC01:22
*** lixiaoy1 has quit IRC01:23
*** salv-orlando has quit IRC01:25
*** patrickeast has quit IRC01:25
*** lixiaoy1 has joined #openstack-cinder01:26
davechenlixiaoy1: seems like this is not a good time to talk :)01:26
*** annashen has joined #openstack-cinder01:26
*** mriedem has quit IRC01:26
*** Lee1092 has joined #openstack-cinder01:28
*** annashen has quit IRC01:31
*** lixiaoy11 has joined #openstack-cinder01:33
openstackgerritVincent Hou proposed openstack/cinder: Adds the migration progress support for migration
*** lixiaoy1 has quit IRC01:35
*** breitz has joined #openstack-cinder01:37
openstackgerritLi Yingjun proposed openstack/cinder: Validate 'is_public' when creating volume type
*** vincent_hou has joined #openstack-cinder01:46
vincent_houtbarron: Hi Tom.01:47
*** annashen has joined #openstack-cinder01:55
*** daneyon has quit IRC01:57
*** breitz has quit IRC01:57
tbarronhi vincent_hou01:58
*** Apoorva has quit IRC01:58
*** annashen has quit IRC02:00
openstackgerritLi Yingjun proposed openstack/cinder: Add the ability to update volume type public status
*** garthb has quit IRC02:10
*** annashen has joined #openstack-cinder02:10
*** annashen has quit IRC02:12
*** Tross has joined #openstack-cinder02:16
*** vincent_hou has quit IRC02:17
*** david-lyle has joined #openstack-cinder02:19
*** haomaiwang has quit IRC02:21
*** haomaiwa_ has joined #openstack-cinder02:22
*** david-lyle has quit IRC02:23
*** annashen has joined #openstack-cinder02:27
*** jwcroppe has quit IRC02:27
*** jwcroppe has joined #openstack-cinder02:28
*** annashen has quit IRC02:32
*** jwcroppe has quit IRC02:32
*** haomaiwa_ has quit IRC02:36
*** heyun has quit IRC02:36
*** haomaiwang has joined #openstack-cinder02:36
openstackgerritVincent Hou proposed openstack/python-cinderclient: Change cinderclient according to volume migration improvement
*** boris-42 has quit IRC02:40
*** zhonghua-li has quit IRC02:41
*** zhonghua-lee has joined #openstack-cinder02:41
*** angela-s has quit IRC02:43
*** IlyaG has joined #openstack-cinder02:50
*** hakimo has joined #openstack-cinder02:51
*** hakimo_ has quit IRC02:53
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
*** annashen has joined #openstack-cinder03:10
*** mragupat has joined #openstack-cinder03:12
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
*** mragupat_ has joined #openstack-cinder03:13
*** haomaiwang has quit IRC03:14
openstackgerritliuxinguo proposed openstack/cinder: Add volume migration support for Huawei driver
*** srkr has joined #openstack-cinder03:15
*** mragupat has quit IRC03:16
*** coolsvap|away is now known as coolsvap03:26
*** annashen_ has joined #openstack-cinder03:32
*** david-lyle has joined #openstack-cinder03:34
*** annashen has quit IRC03:35
openstackgerritBharat Kumar Kobagana (BharatK) proposed openstack/cinder: GlusterFS: Using 'fallocate' instead of 'dd'
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** ishant has joined #openstack-cinder03:48
openstackgerritOpenStack Proposal Bot proposed openstack/os-brick: Updated from global requirements
*** BharatK has joined #openstack-cinder03:49
*** davechen is now known as davechen_afk03:58
*** annashen has joined #openstack-cinder03:58
*** zhonghua-lee has quit IRC03:59
*** zhonghua-lee has joined #openstack-cinder04:01
*** Apoorva has joined #openstack-cinder04:03
*** annashen has quit IRC04:03
*** annashen_ has quit IRC04:07
*** mragupat_ has quit IRC04:09
*** jwcroppe has joined #openstack-cinder04:14
*** IlyaG has quit IRC04:17
*** harlowja_at_home has joined #openstack-cinder04:17
*** zhonghua-lee has quit IRC04:27
*** avishay has joined #openstack-cinder04:32
*** zhonghua-lee has joined #openstack-cinder04:36
*** vilobhmm has joined #openstack-cinder04:49
*** annashen has joined #openstack-cinder04:49
*** rushiagr_away is now known as rushiagr04:50
srkrhi, is there any limitation on the size of a cinder volume that can be used for "boot from volume" provided tenant quota and cinder-backend allows it04:51
*** IlyaG has joined #openstack-cinder04:51
*** ub has joined #openstack-cinder04:53
*** LiuXinguo has joined #openstack-cinder04:53
*** simondodsley_ has joined #openstack-cinder04:53
*** Liu has quit IRC04:54
*** ub2 has quit IRC04:54
*** simondodsley has quit IRC04:54
*** annashen_ has joined #openstack-cinder04:59
*** harlowja_still_a has joined #openstack-cinder05:02
*** annashen_ has quit IRC05:03
*** shyama has joined #openstack-cinder05:05
*** harlowja_at_home has quit IRC05:05
*** daneyon has joined #openstack-cinder05:05
*** srkr has quit IRC05:06
*** daneyon_ has joined #openstack-cinder05:07
*** ub has quit IRC05:08
*** BharatK has quit IRC05:09
*** daneyon has quit IRC05:10
*** annashen has quit IRC05:10
*** annashen has joined #openstack-cinder05:16
*** harlowja_still_a has quit IRC05:16
*** Apoorva has quit IRC05:16
*** harlowja_at_home has joined #openstack-cinder05:16
*** Apoorva has joined #openstack-cinder05:22
*** IlyaG has quit IRC05:24
*** jwcroppe has quit IRC05:25
*** jwcroppe has joined #openstack-cinder05:25
*** BharatK has joined #openstack-cinder05:25
*** harlowja_still_a has joined #openstack-cinder05:29
*** jwcroppe has quit IRC05:30
*** deepakcs has joined #openstack-cinder05:31
*** lpetrut has joined #openstack-cinder05:31
*** harlowja_at_home has quit IRC05:32
*** harlowja_still_a has quit IRC05:38
openstackgerritVilobh Meshram proposed openstack/cinder: Cinder Nested Quota Driver
*** jwcroppe has joined #openstack-cinder05:45
*** shyama has quit IRC05:45
*** heyun has joined #openstack-cinder05:49
*** agarciam has joined #openstack-cinder05:55
*** annashen_ has joined #openstack-cinder06:00
*** annashen has quit IRC06:00
*** lpetrut has quit IRC06:00
*** Apoorva has quit IRC06:01
*** vlaza has joined #openstack-cinder06:04
*** lpetrut has joined #openstack-cinder06:04
*** annashen_ has quit IRC06:04
*** rushiagr is now known as rushiagr_away06:07
*** afazekas has joined #openstack-cinder06:09
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** rushiagr_away is now known as rushiagr06:16
*** nkrinner has joined #openstack-cinder06:18
*** wanghao has joined #openstack-cinder06:30
*** ub has joined #openstack-cinder06:32
*** zhenguo has quit IRC06:32
*** rushiagr is now known as rushiagr_away06:34
*** abhishekk has joined #openstack-cinder06:34
deepakcsanyone with DB expertise here ? Whats the advantage of doing index=True in Column('project_id', String(length=255), index=True)06:35
*** alonmarx has quit IRC06:35
*** zhenguo has joined #openstack-cinder06:40
dulekdeepakcs: It gives better performance when filtering by project_id, which is done quite often I think.06:41
deepakcsdulek: so does it help in sorting or just plain retrievign ?06:42
dulekdeepakcs: It definitely helps with "WHERE project_id='foo'" clauses. Let me think about sorting...06:42
deepakcsdulek: thanks, I am not a DB guy, and i recd some comments from e0ne @
deepakcsdulek: could you help me understand what he means by "I propose to use UUID as primary key to make it more scalable"06:43
dulekdeepakcs: - it seems that index can help with sorting performance.06:44
deepakcsdulek: i did a grep on primary_key in cinder/db/sqlalchemy/migrate_repo/versions/* and none of them use UUID for primary_keys, so not sure how to address that comment of his. Any help appreciated :)06:44
*** abehl has joined #openstack-cinder06:44
deepakcsdulek: good to know.06:45
*** lpetrut has quit IRC06:45
dulekdeepakcs: You should have did a grep on cinder/db/sqlalchemy/models06:46
dulekdeepakcs: Even volume has it's id (which is primary_key) as UUID.06:47
deepakcsdulek: ah, so he meant to use a String type instead of integer type, so that id can hold UUID ?06:48
dulekdeepakcs: This is how it's done in current code.06:49
deepakcsdulek: I am adding snapshot_admin_metadata on similar lines of volume_admin_metadata and i just followed
deepakcsdulek: as u can see there ^^ it uses Integer for 'id' field, which i just followed for my case :)06:49
dulekdeepakcs: Then I think you should respond with that and ping e0ne.06:51
deepakcsdulek: ok, but i think agreeing to what he says also makes sense... so in a dilema right now06:52
deepakcsdulek: i just wanted to check with other DB experts around here, if there is any specific reason why 'id'is a Integer for volume_admin_md case ? can u think of any ?06:52
dulekdeepakcs: I don't have enough expertise to make decision here, so let's pass it to Ivan. ;)06:53
deepakcsdulek: and i guess Ivan is available in US times ?06:53
dulekdeepakcs: No, he is living in eastern Europe, I guess he's eating his breakfast now. ;)06:54
* deepakcs really wishes if people can provide more constructive comments for DB files, since not many have DB expertise06:54
lifelessints are faster to compare than uuids06:54
deepakcslifeless: e0ne says UUID makes it more scalable, so feel free to debate (see the link i gave above with e0ne's comments)06:55
dulekdeepakcs: Well, I find knowledge on indexes rather basic stuff on databases. Anyway the good side of reviews is that you can always learn.06:55
deepakcsdulek: given the TZ difference, its better for people to provide hints that can expedite the work. I am old enuf to justify saying that i don't remember much of DB specifics06:56
deepakcsdulek: esp when adding a new DB table is not the primary focus of my patch series06:57
dulekdeepakcs: More descriptive comments are always welcomed, not only in DB-related patches.06:57
*** kjelly has joined #openstack-cinder06:57
deepakcsdulek: any idea how far from now e0ne would be online ?06:57
dulekdeepakcs: It's just sometimes hard to realize what level of description is enough.06:58
*** davechen_afk is now known as davechen06:58
*** markus_z has joined #openstack-cinder06:58
dulekI would expect him to be here in an hour or so. If he haven't started his vacation or something.06:59
*** vilobhmm has quit IRC06:59
*** zhonghua-lee has quit IRC07:00
lifelessconsistencygroup has a string pk07:01
lifelessas does cgsnapshot and volune07:01
*** annashen has joined #openstack-cinder07:01
*** abehl has quit IRC07:03
deepakcsthanks dulek lifeless07:03
*** annashen has quit IRC07:05
openstackgerritwanghao proposed openstack/cinder: Implement function to manage/unmanage snapshots
*** rushiagr_away is now known as rushiagr07:10
*** salv-orlando has joined #openstack-cinder07:12
*** abehl has joined #openstack-cinder07:16
*** ronis has joined #openstack-cinder07:19
*** rushiagr is now known as rushiagr_away07:20
openstackgerritSergey Vilgelm proposed openstack/cinder: Switch to the oslo_utils.fileutils
*** jwcroppe has quit IRC07:22
openstackgerritVincent Hou proposed openstack/cinder: Volume status management during migration
openstackgerritliuxinguo proposed openstack/cinder: Add hypermetro support for Huawei driver
openstackgerritliuxinguo proposed openstack/cinder: Fix use of urllib for py3 compatibility in Huawei driver
*** markus_z has quit IRC07:34
*** lpetrut has joined #openstack-cinder07:46
*** abehl has quit IRC07:48
*** zhonghua-lee has joined #openstack-cinder07:48
*** jwcroppe has joined #openstack-cinder07:51
*** markus_z has joined #openstack-cinder07:51
*** bluex-pl has joined #openstack-cinder07:51
*** bluex-pl has quit IRC07:51
openstackgerritXinXiaohui proposed openstack/cinder: capacity-headroom
openstackgerritXi Yang proposed openstack/cinder: Snap copy feature for EMC VNX Cinder driver
*** annashen has joined #openstack-cinder08:01
*** Liu has joined #openstack-cinder08:01
*** abehl has joined #openstack-cinder08:01
*** zhonghua-li has joined #openstack-cinder08:01
*** annashen_ has joined #openstack-cinder08:02
openstackgerritLi Yingjun proposed openstack/cinder: Add the ability to update type public status
*** zhonghua-lee has quit IRC08:04
*** LiuXinguo has quit IRC08:04
*** annashen has quit IRC08:05
*** annashen_ has quit IRC08:06
*** jordanP has joined #openstack-cinder08:06
*** daneyon_ has quit IRC08:07
*** jistr has joined #openstack-cinder08:08
*** ndipanov has quit IRC08:14
*** ndipanov has joined #openstack-cinder08:14
*** jasondotstar has joined #openstack-cinder08:25
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
*** zhonghua-li has quit IRC08:28
*** lixiaoy11 has quit IRC08:28
*** jwcroppe has quit IRC08:35
*** davechen is now known as davechen_afk08:36
openstackgerritwanghao proposed openstack/cinder: Improve performance listing detail for volumes
*** shyama has joined #openstack-cinder08:37
*** e0ne has joined #openstack-cinder08:40
*** markus_z has quit IRC08:43
*** e0ne has quit IRC08:43
openstackgerritJordan Pittier proposed openstack/os-brick: Remotefs: add ScalityFS support
openstackgerritMichal Dulko proposed openstack/cinder-specs: Add Cinder Volume A/A HA lock spec
*** markus_z has joined #openstack-cinder08:43
dulekgeguileo: ^08:44
dulekgeguileo: I've just updated some nits... I also don't get double spaces after dots, but let's not dig into it.08:45
*** zhonghua-li has joined #openstack-cinder08:48
*** davechen_afk is now known as davechen08:50
*** zhonghua-li has quit IRC08:56
*** zhonghua-li has joined #openstack-cinder08:57
*** zhonghua has joined #openstack-cinder08:58
geguileodulek: In a meeting, give me 10 minutes08:58
*** zhonghua-li has quit IRC09:02
*** jwcroppe has joined #openstack-cinder09:02
*** markus_z has quit IRC09:02
*** annashen has joined #openstack-cinder09:02
*** coolsvap is now known as coolsvap|away09:04
*** BharatK has quit IRC09:05
geguileodulek: Ok, I'm back09:05
openstackgerritliuxinguo proposed openstack/cinder: Add hypermetro support for Huawei driver
geguileodulek: I'll read it now.  We can remove double spaces, no problem09:06
*** annashen has quit IRC09:07
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
*** coolsvap|away is now known as coolsvap09:12
geguileodulek: Look good to me09:12
openstackgerritliuxinguo proposed openstack/cinder: Add volume migration support for Huawei driver
geguileodulek: I must have been hungry yesterday, because I ate a bunch of "ed"s  ;-)09:12
*** davechen has left #openstack-cinder09:13
*** markus_z has joined #openstack-cinder09:19
*** alexpilotti has joined #openstack-cinder09:21
*** ankit_ag has joined #openstack-cinder09:26
-openstackstatus- NOTICE: Currently our CI system is broken, jobs are not getting processed at all.09:27
*** ChanServ changes topic to "Currently our CI system is broken, jobs are not getting processed at all."09:27
*** gaurangt has joined #openstack-cinder09:32
openstackgerritwanghao proposed openstack/cinder: Add pagination to backups
*** bluex-pl has joined #openstack-cinder09:37
*** bluex-pl has quit IRC09:37
*** bluex-pl has joined #openstack-cinder09:37
*** heyun has quit IRC09:41
*** nikeshm has joined #openstack-cinder09:45
*** markus_z has quit IRC09:46
*** mdbooth has quit IRC09:46
*** mdbooth has joined #openstack-cinder09:47
nikeshmliuxg: hi09:49
nikeshmLiu: are you Liu Xinguo?09:50
*** coolsvap is now known as coolsvap|away09:50
*** e0ne has joined #openstack-cinder09:57
*** jasondotstar has quit IRC09:57
*** shyama has quit IRC09:58
*** coolsvap|away is now known as coolsvap09:58
*** jwcroppe has quit IRC10:00
*** ankit_ag has quit IRC10:00
*** abehl has quit IRC10:00
*** ub has quit IRC10:00
*** rhe00 has quit IRC10:00
*** jwang has quit IRC10:00
*** Swanson has quit IRC10:00
*** mitz has quit IRC10:00
*** mitz has joined #openstack-cinder10:00
*** ankit_ag has joined #openstack-cinder10:01
*** Swanson has joined #openstack-cinder10:01
*** jwang has joined #openstack-cinder10:01
*** abhishekkk has joined #openstack-cinder10:01
*** ub has joined #openstack-cinder10:01
*** rhe00 has joined #openstack-cinder10:01
*** annashen has joined #openstack-cinder10:02
deepakcse0ne: need some help wrt your comment @
deepakcse0ne: i changed the field 'id' from Integer to String(36)10:02
*** e0ne_ has joined #openstack-cinder10:02
*** abehl has joined #openstack-cinder10:03
deepakcse0ne: I am hitting this error when i try to update an entry in snapshot_admin_metadata :
*** annashen_ has joined #openstack-cinder10:04
deepakcse0ne: I even tried to adding (if not 'id', id = uuid) my snapshot_admin_metadata_update function. See this:
deepakcse0ne_: ^^10:05
*** rushiagr_away has quit IRC10:05
*** Lee1092 has quit IRC10:05
*** e0ne has quit IRC10:05
*** abhishekk has quit IRC10:05
*** kaisers has quit IRC10:05
*** timcl1 has quit IRC10:05
*** DTadrzak has quit IRC10:05
deepakcse0ne: e0ne_ still hitting the same error (field id doesn't have default value)10:05
e0ne_deepakcs: is it on aster or some patch?10:06
*** e0ne_ is now known as e0ne10:06
*** DTadrzak has joined #openstack-cinder10:06
*** timcl has joined #openstack-cinder10:06
deepakcse0ne: looks like u missed my earlier msg, let me repaste it10:06
deepakcse0ne: need some help wrt your comment @
deepakcse0ne: i changed the field 'id' from Integer to String(36)10:06
deepakcse0ne: aster meaning ?10:06
*** annashen has quit IRC10:06
*** kaisers has joined #openstack-cinder10:06
e0nedeepakcs: yes, i missed this message. give me 5 minutes, please10:07
deepakcse0ne: np10:07
*** Lee1092 has joined #openstack-cinder10:07
*** BharatK has joined #openstack-cinder10:08
*** annashen_ has quit IRC10:08
*** jasondotstar has joined #openstack-cinder10:09
*** zhenguo has quit IRC10:13
deepakcse0ne: there ?10:29
e0nedeepakcs: looking10:30
*** shyama has joined #openstack-cinder10:31
e0nedeepakcs: sqlalchemy can't generate uuid automacitly for pk10:32
deepakcse0ne: ^^ i did that, still hitting the error as seen here  :
deepakcse0ne: lines 5 & 6 in add the uuid if not present.10:33
e0nedeepakcs: could you please share full traceback?10:35
Liunikeshm: yeah, I am Liu Xinguo10:36
deepakcse0ne: is what i see in my c-api log10:36
openstackgerritVipin Balachandran proposed openstack/cinder: VMware: Set virtual disk UUID to volume ID
Liunikeshm: or liuxinguo  or liuxg :)10:36
e0nedeepakcs: i don't see any cinder code there. only oslo.db traceback10:37
deepakcse0ne: thats all i am getting. see line 1.. thats the cinder api contrib_actions invoking a snapshot.update followed by save in the code10:38
deepakcse0ne: after which i see the traceback10:38
*** aix has quit IRC10:38
openstackgerritGorka Eguileor proposed openstack/cinder: Move import and export backup metadata to object
openstackgerritGorka Eguileor proposed openstack/cinder: Fix backup metadata import missing fields
Liunikeshm: I will leave for dinner time, and will be back about two hours latter10:39
nikeshmLiu: ok, thanks, are you thinking to optimize the code for smartx feature,10:39
deepakcse0ne: can u review my snapshot_admin_metadata _update function in
deepakcse0ne: and what needs to be done there if i change 'id' field of my table from Integer to String(36)10:40
Liunikeshm: yes, I have thought about it10:40
deepakcse0ne: bcos thats what i did, and landed in this issue, so something in my _update function isn't correct10:40
*** vlaza has left #openstack-cinder10:40
Liunikeshm: but have not yet a good idea about it10:41
e0nedeepakcs: the code looks fine10:41
nikeshmLiu: i have commented some suggestions on the patch, that might help10:42
deepakcse0ne: to begin with it doesn't initialize the 'id' field (when it was Integer, it was autoincrement, but now its string and i don't set it anywhere)10:42
deepakcse0ne: there is no create() for metadata tables, they get rows added via snapshot['admin_metadata']10:42
Liunikeshm: that's greate!10:43
e0nedeepakcs: it could be an issue10:43
e0nedeepakcs: it affects a lot of code10:43
Liunikeshm: I will see it and do the optimization if if's OK10:43
deepakcse0ne: so i did this : (see line 5 & 6)10:44
nikeshmLiu: donot mind, what should i use he/she for you?10:44
e0nedeepakcs: maybe making pk as uuid was a bad idea10:44
Liuhe, thanks10:44
deepakcse0ne: even after that, sql error happens saying during update it is not finding 'id' field10:44
nikeshmLiu: ok, thanks10:44
deepakcse0ne: shall i revert back to Integer then ?10:44
e0nedeepakcs: let's do it for this time10:44
e0nedeepakcs: i won't -1 for it10:45
deepakcse0ne: can u pls put that in
nikeshmLiu: which city are you working?10:45
e0nedeepakcs: i need more time to investigate this issue10:45
deepakcse0ne: ok, but i can proceed by usign Integer, thats fine right ?10:45
e0nedeepakcs: yes10:45
deepakcse0ne: pls update ur comment in
e0nedeepakcs: done:)10:46
Liunikeshm: chengdu, a beaufiful chinese city10:46
deepakcse0ne: thanks, i was debuggign this for most of my today :(10:46
e0nedeepakcs: i review your patch again after you'll fix other comments10:46
deepakcse0ne: wasted a lot of time10:46
nikeshmLiu: nice10:46
deepakcse0ne: yes, willl do10:46
e0nedeepakcs: it was my fail too:(10:47
Liunikeshm: and what about you10:47
deepakcse0ne: its ok, it happens :)10:47
nikeshmLiu: i work in hyderabad, INDIA10:48
e0nedeepakcs: in cinder, we need some notes or even guidelines what do we want for PKs: int or uuid10:48
Liunikeshm: Nice place10:48
e0nedeepakcs: now, it's very confusing to have both of them w/o any algorithm to choose one10:48
deepakcse0ne: yes, and if its uuid, and if its a lazy loaded (like metadata or admin_metadata) how to handle it, bcos it doesn't work with what i just did10:48
nikeshmLiu: my hometown is near to Varanasi,Uttar Pradesh,India10:49
deepakcse0ne: it should logically have worke with what i did tho :( so i would be interested to know if u root cause it as to why its not working10:49
*** lennyb_ has quit IRC10:49
e0nedeepakcs: yep. it's very interesting for me to find what gone wrong. i'll try to find time for it10:49
deepakcse0ne: sure10:49
Liunikeshm: hah, very nick10:49
Liunikeshm: I must leave for dinner10:50
deepakcse0ne: just as FYI, i also get this error: "2015-07-29 10:50:48.381 ERROR cinder.api.middleware.fault [req-0e17993b-8e5a-4581-a425-58e82ef65f0d 4702a0f683f14183a55cd0e1d6a6c68f 3d53b60a4b154988a851e8b887231946] Caught error: SQL expression, column, or mapped entity expected - got 'True'10:51
deepakcse0ne: when i try to delete the cinder volume10:51
deepakcse0ne: not sure if it helps in root causing10:51
Liunikeshm: it's almost evening now in my time zone10:51
*** markus_z has joined #openstack-cinder10:53
*** coolsvap is now known as coolsvap|away10:55
ubhey guys! One week ago i fixed a little bug and it's still not merged. I think one core reviewer (+2) is missing. Did i forgot something? Is it normal?
openstackgerritMichal Dulko proposed openstack/cinder-specs: Add Cinder Volume A/A HA lock spec
dulekub: It's completely normal, core reviewers time is a limited resource. Unfortunately you need to be patient on that.11:04
*** annashen has joined #openstack-cinder11:04
*** annashen has quit IRC11:09
openstackgerritGorka Eguileor proposed openstack/cinder: Restore backup to Vol ID when it doesn't exist
openstackgerritPranali Deore proposed openstack/cinder: Validate string, integer limit for input parameter
*** jasondot_ has joined #openstack-cinder11:23
sgotliv_tbarron, good morning! yt?11:24
*** e0ne has quit IRC11:25
tbarronsgotliv_: yes, good afternoon?11:25
sgotliv_tbarron, yes :-) afternoon. I have a quick question about the NetApp driver11:25
*** jasondotstar has quit IRC11:25
tbarronsgotliv_: sure11:25
*** zhonghua has quit IRC11:26
sgotliv_tbarron, One of my customers gets this error NaApiError: NetApp api failed. Reason - 17163:Clone operation failed to start: Max Dense limit is reached..\n']11:26
sgotliv_tbarron, I only can guess what does it mean11:27
tbarronsgotliv_: I can't defend that message :-)11:28
tbarronsgotliv_: it means that the filers hit a limit on the number of clones they could do in that burst or clone deletes11:28
tbarronI've seen it on clone deletes.11:29
tbarronI think some deferred COW work has to be done and there is a rate limit.11:29
* sgotliv_ processing :-)11:29
tbarronsgotliv_: if you want I can look at logs and see if I can tell you something less speculative.11:30
sgotliv_tbarron, thanks, let me check if I can send you logs. All these security measures are very restrictive11:31
tbarronsgotliv_: sure, understood.11:31
*** DTadrzak has quit IRC11:31
*** belmoreira has joined #openstack-cinder11:32
tbarronsgotliv_: if they are also netapp customer ...11:32
sgotliv_tbarron, I already thought about it a sec ago :-)11:32
sgotliv_i want to find a way to add you to the loop11:32
sgotliv_but meantime, you are right! They are trying to create a lot of snapshots11:33
*** cdelatte has joined #openstack-cinder11:33
*** delattec has joined #openstack-cinder11:33
tbarronsgotliv_: cinder snapshots we implement as file (or LUN) clones11:33
openstackgerritSzymon Wróblewski proposed openstack/cinder: Tooz locks
*** e0ne has joined #openstack-cinder11:35
*** rushiagr_away has joined #openstack-cinder11:37
*** delatte has joined #openstack-cinder11:40
deepakcse0ne: ping11:40
*** delattec has quit IRC11:41
*** cdelatte has quit IRC11:42
*** shyama has quit IRC11:46
*** bswartz has quit IRC11:46
*** jwcroppe has joined #openstack-cinder11:47
*** delattec has joined #openstack-cinder11:48
*** jwcroppe has quit IRC11:48
*** svilgelm has joined #openstack-cinder11:49
*** delatte has quit IRC11:51
*** coolsvap|away is now known as coolsvap11:51
*** kodoku_ has joined #openstack-cinder11:52
*** svilgelm has left #openstack-cinder11:52
*** svilgelm has joined #openstack-cinder11:52
kodoku_Hi, I have issue when I attach volume ==> Unable to fetch connection information from backend: 'NoneType' object has no attribute 'service_content' to caller11:58
kodoku_I use KILO version. python-cinderclient-1.1.1-1.el7.noarchpython-cinder-2015.1.0-3.el7.noarchopenstack-cinder-2015.1.0-3.el7.noarch11:59
*** e0ne has quit IRC11:59
kodoku_Somebody have idea ?12:00
ccesarioHello, does someone have idea about this SSL error in cinder-volume service ?
*** annashen has joined #openstack-cinder12:02
*** annashen_ has joined #openstack-cinder12:05
*** aix has joined #openstack-cinder12:07
*** annashen has quit IRC12:07
*** Lee1092 has quit IRC12:07
*** annashen_ has quit IRC12:10
*** svilgelm has quit IRC12:10
*** gaurangt has quit IRC12:14
*** erlon has joined #openstack-cinder12:16
*** ByPasS has joined #openstack-cinder12:16
*** marcusvrn has joined #openstack-cinder12:16
*** coolsvap is now known as coolsvap|away12:17
*** kodoku_ has quit IRC12:19
*** kodoku has joined #openstack-cinder12:19
*** kaisers_ has joined #openstack-cinder12:19
kodokuccesario hI12:19
kodokuccesario cinder connect to vcenter with https ?12:20
openstackgerritErickson Filipe Guedes dos Santos proposed openstack/cinder: Cinder Nested Quota Driver
ccesariokodoku, hi,.. yes its try.... with nova service I solve it with insecure = True .... but with cinder I dot not know how solve it :/12:21
*** salv-orl_ has joined #openstack-cinder12:22
kodokuccesario You can add your CA in ca-trust on your control node12:22
*** xek_ is now known as xek12:23
*** akerr has joined #openstack-cinder12:24
*** avishay has quit IRC12:24
*** salv-orlando has quit IRC12:25
*** e0ne has joined #openstack-cinder12:25
*** enikher has joined #openstack-cinder12:25
ccesariokodoku, insecure mode is not possible with cinder ?12:26
*** salv-orlando has joined #openstack-cinder12:26
e0nedeepakcs: ping12:26
*** salv-orl_ has quit IRC12:26
kodokuccesario I use cinder with https vcenter, I don't see insecure option :/ But I don't add CA in ca-trust12:26
*** edmondsw has joined #openstack-cinder12:27
*** salv-orl_ has joined #openstack-cinder12:27
*** kaisers_ has quit IRC12:29
*** vlaza has joined #openstack-cinder12:29
*** salv-orlando has quit IRC12:30
openstackgerritAnton Arefiev proposed openstack/cinder: Add Cinder API WSGI app to run under Apache/Nginx
openstackgerritAnton Arefiev proposed openstack/cinder: Add instructions on how deploy API under Apache
*** bluex-pl has quit IRC12:30
openstackgerritAnton Arefiev proposed openstack/cinder: Add instructions on how deploy API under Apache
openstackgerritAnton Arefiev proposed openstack/cinder: Add Cinder API WSGI app to run under Apache/Nginx
openstackgerritAnton Arefiev proposed openstack/cinder: Add instructions on how deploy API under Apache
openstackgerritAnton Arefiev proposed openstack/cinder: Add Cinder API WSGI app to run under Apache/Nginx
*** deepakcs has quit IRC12:33
openstackgerritDeepak C Shetty (deepakcs) proposed openstack/cinder: Cinder part of volume snapshot improvements
openstackgerritDeepak C Shetty (deepakcs) proposed openstack/cinder: db: Add support for snapshot_admin_metadata
*** salv-orlando has joined #openstack-cinder12:35
*** ganso_ has joined #openstack-cinder12:35
*** ishant has quit IRC12:36
*** salv-orl_ has quit IRC12:38
*** Tross has quit IRC12:39
*** Tross1 has joined #openstack-cinder12:39
*** salv-orl_ has joined #openstack-cinder12:44
ccesariokodoku, is it possible you show me your cinder.conf ?12:44
kodokuccesario Yes, one second12:46
*** salv-orlando has quit IRC12:47
*** salv-orl_ has quit IRC12:47
*** coolsvap|away is now known as coolsvap12:48
*** salv-orlando has joined #openstack-cinder12:48
*** ChanServ changes topic to "The OpenStack Block Storage Project Cinder | The New Kids On the Block |"12:48
-openstackstatus- NOTICE: zuul's disks were at capacity. Space has been freed up and jobs are being re-queued.12:48
*** bswartz has joined #openstack-cinder12:49
kodokuccesario Send by pm12:49
*** shyama has joined #openstack-cinder12:53
ccesariokodoku, thank you.... let me check12:55
*** dims has quit IRC12:57
*** salv-orl_ has joined #openstack-cinder12:57
*** annegentle has joined #openstack-cinder12:57
*** dims has joined #openstack-cinder12:57
*** mriedem has joined #openstack-cinder12:57
*** salv-orlando has quit IRC13:00
*** salv-or__ has joined #openstack-cinder13:00
*** zhipeng has joined #openstack-cinder13:00
*** liuxg has joined #openstack-cinder13:03
*** salv-orl_ has quit IRC13:04
ccesariokodoku, it is similar.... with mime ....13:04
*** Yogi1 has joined #openstack-cinder13:05
kodokuccesario Hum I don't because I use auto sign for my vcenter13:05
*** annashen has joined #openstack-cinder13:06
ccesarioI will try search any wayt to use insecure mode to it13:06
ccesarioauto sign!?13:07
kodokuccesario If I curl on https://myvcenter I have "Use -k because this cert is not sign"13:08
kodokuBut cinder works :/13:08
kodokuWhat is your version of cinder ?13:09
*** salv-or__ has quit IRC13:09
*** annashen has quit IRC13:10
*** salv-orlando has joined #openstack-cinder13:11
*** coolsvap is now known as coolsvap|away13:14
*** salv-orl_ has joined #openstack-cinder13:14
*** salv-orlando has quit IRC13:17
*** kaisers_ has joined #openstack-cinder13:17
*** markstur_ has joined #openstack-cinder13:18
*** salv-orlando has joined #openstack-cinder13:18
ccesariokodoku, I 'm testing with devstack13:19
ccesariocinder-volume --version13:19
*** markstur_ has left #openstack-cinder13:19
ccesariocinder --version13:20
*** salv-orl_ has quit IRC13:21
openstackgerritJoel Coffman proposed openstack/cinder-specs: Add volume encryption metadata spec
*** eharney has joined #openstack-cinder13:23
-openstackstatus- NOTICE: zuul jobs after about 07:00 UTC may need a 'recheck' to enter the queue. Look if your change is in and recheck if not.13:26
*** akshai has joined #openstack-cinder13:26
*** BharatK has quit IRC13:29
*** kodoku has quit IRC13:30
*** kaisers_ has quit IRC13:34
*** kaisers_ has joined #openstack-cinder13:36
*** annegentle has quit IRC13:36
*** abhishekkk has quit IRC13:39
*** DTadrzak has joined #openstack-cinder13:39
*** kaisers_ has quit IRC13:41
*** jwcroppe has joined #openstack-cinder13:43
*** BharatK has joined #openstack-cinder13:43
*** markvoelker has joined #openstack-cinder13:44
*** jasondotstar has joined #openstack-cinder13:48
*** jungleboyj has quit IRC13:52
*** salv-orlando has quit IRC13:52
*** jasondot_ has quit IRC13:52
*** timcl has quit IRC13:54
*** mragupat has joined #openstack-cinder13:59
*** mragupat has quit IRC13:59
*** breitz has joined #openstack-cinder13:59
*** thangp has joined #openstack-cinder13:59
*** timcl has joined #openstack-cinder14:00
*** mragupat has joined #openstack-cinder14:00
*** mylu has joined #openstack-cinder14:01
*** annashen has joined #openstack-cinder14:03
*** markvoelker has quit IRC14:04
*** lpetrut has quit IRC14:04
*** EmilienM has quit IRC14:05
*** lpetrut has joined #openstack-cinder14:06
*** annashen_ has joined #openstack-cinder14:07
*** annashen has quit IRC14:08
*** zhipeng has quit IRC14:08
*** EmilienM has joined #openstack-cinder14:09
*** mylu has quit IRC14:11
*** annashen_ has quit IRC14:11
*** mylu has joined #openstack-cinder14:12
*** ankit_ag has quit IRC14:13
*** annegentle has joined #openstack-cinder14:14
*** Miouge has joined #openstack-cinder14:18
*** markvoelker_ has joined #openstack-cinder14:21
*** vlaza has left #openstack-cinder14:24
*** mtanino has joined #openstack-cinder14:25
*** r-daneel has joined #openstack-cinder14:26
*** kodoku has joined #openstack-cinder14:31
*** lpetrut has quit IRC14:32
*** lpetrut has joined #openstack-cinder14:36
*** coolsvap|away is now known as coolsvap14:37
*** markvoelker_ has quit IRC14:38
*** markvoelker has joined #openstack-cinder14:39
openstackgerritErlon R. Cruz proposed openstack/cinder: Fix HNAS iSCSI 32 targets limitation error
*** kjelly_ has joined #openstack-cinder14:45
*** markvoelker has quit IRC14:47
*** markvoelker has joined #openstack-cinder14:47
*** crose has joined #openstack-cinder14:48
*** Guest100_ has joined #openstack-cinder14:51
*** esker has joined #openstack-cinder14:54
*** mdenny has joined #openstack-cinder14:56
kodokuHi, I have issue with cinder (KILO), I can create volume, attache fresh volume, but after detach, I can't re-attach this volume ===> VolumeBackendAPIException: Bad or unexpected response from the storage volume backend API: Unable to fetch connection information from backend: 'NoneType' object has no attribute 'service_content'14:56
thingeekodoku: I'm assuming this is vmware?14:57
*** kodoku_ has joined #openstack-cinder14:58
kodoku_thingee yes14:58
*** jwcroppe has quit IRC14:58
*** belmoreira has quit IRC14:59
*** jasondotstar has quit IRC14:59
*** tsekiyama has joined #openstack-cinder14:59
kodoku_thingee Do you know this issue ?14:59
thangpdidnt kilo reached EOL?14:59
thingeekodoku_: I'm not associate with vmware, I only assumed it's vmware because they're the only one with service_content in their code.14:59
kodoku_ok ^^15:00
thingeekodoku_: I don't know if their driver maintainers hang out here though15:00
kodoku_I guess it's vipin15:00
thingeevipin balachandran is the maintainer15:00
*** jistr has quit IRC15:00
*** kodoku has quit IRC15:01
*** ByPasS has quit IRC15:01
*** haomaiwang has joined #openstack-cinder15:01
*** mragupat has quit IRC15:01
*** e0ne has quit IRC15:02
*** rex_lee has quit IRC15:02
*** sweston has quit IRC15:02
*** Adriano__ has quit IRC15:02
thingeekodoku_: I looked through their attach related bugs here, and didn't find anything. You mind want to consider reporting something15:02
*** jistr has joined #openstack-cinder15:02
*** ByPasS has joined #openstack-cinder15:02
*** haomaiwang has quit IRC15:03
* thingee disappears a bit for meeting15:03
*** edtubill has joined #openstack-cinder15:03
*** haomaiwang has joined #openstack-cinder15:03
kodoku_thingee ok I open bug, but on vmware team or cinder ?15:04
*** e0ne has joined #openstack-cinder15:04
thingeekodoku_: you can report the bug and tag it vmware to get their attention15:04
thingeekodoku_: if they have their notifications setup correctly, they'll know15:04
thingeeif you want to use other means of letting them know about the bug, that'll be fine too15:05
*** nkrinner has quit IRC15:05
*** IlyaG has joined #openstack-cinder15:06
*** annashen has joined #openstack-cinder15:07
*** Lee1092 has joined #openstack-cinder15:08
openstackLaunchpad bug 1479402 in oslo.vmware "Can't re-attach volume : 'NoneType' object has no attribute 'service_content'" [Undecided,New]15:09
breitzkoduku_:  from the bug……."But I can Re-Attach this volume :"  might want to change that "can to a "can't"15:12
*** annashen has quit IRC15:13
*** svilgelm has joined #openstack-cinder15:14
kodoku_can't ><15:15
kodoku_edit :)15:15
*** mylu has quit IRC15:15
breitzit is easy to figure out from the title and the error msg  - but if it is changed (as you have done) it makes life a little easier.    Thanks for doing that.15:16
*** mylu has joined #openstack-cinder15:17
kodoku_breitz yes of course :)15:18
kjelly_Hi guys, If there are many ways to solve a bug/issue, where is the best place to discuss the solution?15:19
openstackgerritDeepak C Shetty (deepakcs) proposed openstack/cinder: Cinder part of volume snapshot improvements
openstackgerritDeepak C Shetty (deepakcs) proposed openstack/cinder: db: Add support for snapshot_admin_metadata
openstackgerritRyan Rossiter proposed openstack/os-brick: Changed connector protocols to use constants
*** ronis has quit IRC15:20
scottdakjelly_: you could put the alternatives in the launchpad bug report. You could also post a patch with your preferred solution and make note in a comment on the review that alternatives exist (in the bug report).15:21
scottdaIf someone comments on either, you could ping them here. Or you could just post the info here and see who responds.15:21
kjelly_scottda: Thank you for your suggestion.15:22
*** tsekiyama has quit IRC15:23
scottdasure. It kind of depends on the level of detail. Just ask here if it's simple.15:23
SwansonAre we getting any successful CI runs?15:24
SwansonWrong window.15:24
*** Miouge has quit IRC15:25
*** tsekiyama has joined #openstack-cinder15:25
smcginnisSwanson: Haha, yes, some. :)15:27
smcginnisSwanson: Seems to be something going on that started yesterday morning. Haven't had time to dig in to it.15:27
openstackgerritKendall Nelson proposed openstack/cinder: Enable cinder-manage to remove services
thingeeSwanson: lol15:28
kjelly_scottda: Ok. I am working on the bug. bug/1477625 . There are many ways to solve the problem. The simple way is . Another way is change the interface . And don't use context.elevated()  .   Any suggestion for the bug ?15:28
smcginnisWith split screen irissi I'm just waiting for the day when I accidentally put something private in channel.15:28
smcginnisI've certainly done the reverse plenty of times.15:29
scottdakjelly_: I'll look in a minute, I've a meeting starting ATM....15:29
thingeesmcginnis: can't be as bad when I accidentally announced I was leaving my previous company.15:29
kjelly_scottda: Ok. Thank you.15:29
smcginnisthingee: Doh! That could be a little awkward.15:30
thingeeirssi, pasting a message with a return in it.15:30
*** jdurgin1 has joined #openstack-cinder15:30
*** alexpilotti has quit IRC15:34
tbarronthingee: that's one way to give notice :)15:35
guitarzankjelly_: it needs to be changed to not use context elevated for the volume_get()15:35
guitarzankjelly_: and then later use context elevated to fetch the viewable admin metadata in a second db call15:35
Swansonthingee: nice!15:35
*** esker has quit IRC15:39
*** diogogmt has joined #openstack-cinder15:40
openstackgerritGorka Eguileor proposed openstack/cinder-specs: Remove Cinder API races
*** dannywilson has joined #openstack-cinder15:42
*** lpetrut has quit IRC15:43
*** lpetrut has joined #openstack-cinder15:43
kjelly_guitarzan: Thank you for your suggestion. This seems to be a good solution. What do you think about letting add another argument to self.db.volume_get for limiting get volume in the same project ?15:43
*** harlowja_at_home has joined #openstack-cinder15:44
guitarzankjelly_: I don't quite understand15:45
guitarzanwhat would the flag do?15:45
guitarzandb.volume_get should always limit getting the volume to the same project or an admin context15:45
*** jungleboyj has joined #openstack-cinder15:46
openstackgerritabhiram moturi proposed openstack/cinder: Fix Oracle ZFSSA iSCSI driver to return free space
kjelly_guitarzan: but becasue the code "context.elevated()", all user is admin context.15:47
guitarzankjelly_: it's entirely possible I don't understand what you're saying though :)15:47
guitarzankjelly_: that is the bug :)15:47
guitarzanit should not be using context.elevated for volume.get()15:47
openstackgerritJon Bernard proposed openstack/cinder: RBD: use user-configured value for chunk size
kjelly_guitarzan: yes. I agree your solution. I will use your suggestion to solve the bugs. Thank you.15:48
*** rhagarty_ has quit IRC15:48
*** rhagarty has quit IRC15:48
guitarzankjelly_: looking forward to seeing it15:49
*** esker has joined #openstack-cinder15:50
*** esker has quit IRC15:50
*** jseiler has joined #openstack-cinder15:53
*** patrickeast has joined #openstack-cinder15:53
*** esker has joined #openstack-cinder15:54
*** esker has quit IRC15:54
*** vincent_hou has joined #openstack-cinder15:56
*** adurbin_ has joined #openstack-cinder15:56
openstackgerritTomoki Sekiyama proposed openstack/cinder: Efficient image transfer for Glance cinder store
openstackgerritTomoki Sekiyama proposed openstack/cinder: Generic image-volume cache
*** rhagarty has joined #openstack-cinder15:58
*** rhagarty_ has joined #openstack-cinder15:58
*** ub has quit IRC15:58
*** haomaiwang has quit IRC16:01
*** lpetrut has quit IRC16:01
kjelly_scottda: It's time for me to sleep. And thank you for your help. If you have any suggesion, plaease comment in or . Tomorrow, I will try the solution suggested by guitarzan.  "it needs to be changed to not use context elevated for the volume_get() and then later use context elevated to fetch the viewable admin metadata in a second db call"16:01
openstackLaunchpad bug 1477625 in Cinder "DB access to show volumes may not be properly controlled" [Undecided,In progress] - Assigned to jelly (coding1314)16:01
*** lpetrut has joined #openstack-cinder16:01
*** esker has joined #openstack-cinder16:01
*** svilgelm has quit IRC16:02
scottdakjelly_: I think guitarzan has you pointed in the right direction.16:02
*** haomaiwang has joined #openstack-cinder16:02
jungleboyjthingee: Are we meeting?16:02
kjelly_scottda: Ok, I will follow your suggesion. Thanks.16:03
*** kjelly_ has quit IRC16:04
*** unibeard has joined #openstack-cinder16:04
*** annashen has joined #openstack-cinder16:04
*** Yogi1 has quit IRC16:04
openstackgerritGorka Eguileor proposed openstack/cinder-specs: Add Cinder Volume A/A HA lock spec
*** vilobhmm has joined #openstack-cinder16:05
*** angela-s has joined #openstack-cinder16:06
*** garthb has joined #openstack-cinder16:06
*** garthb_ has joined #openstack-cinder16:06
*** garthb_ has quit IRC16:07
openstackgerritPatrick East proposed openstack/cinder: Generic image-volume cache
*** jistr has quit IRC16:08
*** annashen_ has joined #openstack-cinder16:08
*** annashen has quit IRC16:09
e0negeguileo: hi16:09
geguileoe0ne: Hi16:09
e0negeguileo: did you see my comments for
openstackLaunchpad bug 1476416 in Cinder "backup.api:import_record sets volume_id to '0000-0000-0000-0000'" [High,In progress] - Assigned to Ivan Kolodyazhny (e0ne)16:10
geguileoe0ne: No I haven't16:10
geguileoe0ne: Today's been crazy for me16:10
geguileoe0ne: Let me have a look now16:10
*** unibeard has left #openstack-cinder16:10
*** thangp has quit IRC16:11
*** daneyon has joined #openstack-cinder16:11
geguileoe0ne: With my patch it gets fixed16:11
*** Apoorva has joined #openstack-cinder16:11
geguileoe0ne: For real16:11
e0negeguileo: no16:11
geguileoe0ne: Because I restore the original Backup ID16:11
e0negeguileo: it's true only if we've got it in the db16:12
*** jaypipes has joined #openstack-cinder16:12
geguileoe0ne: What do you mean if we've got it in the DB?16:12
*** coolsvap is now known as coolsvap|away16:12
geguileoI haven't tested the code in a while, so I may be missing something16:13
e0negeguileo: and after your patch i can't import backup before original is  deleted16:13
*** scottda has quit IRC16:13
*** ctracey has quit IRC16:13
*** briancurtin has quit IRC16:13
*** zhiyan has quit IRC16:13
*** ameade has quit IRC16:13
*** xyang has quit IRC16:13
*** wanghao has quit IRC16:13
*** thingee has quit IRC16:13
*** rhefner has quit IRC16:13
*** ganso_ has quit IRC16:13
*** serverascode has quit IRC16:13
*** annashen_ has quit IRC16:13
geguileoe0ne: Of course you can't import something that's already there16:13
geguileoe0ne: And is still alive and kicking16:13
e0negeguileo: operators could purge raws from db which are marked as deleted16:13
geguileoe0ne: Ok, and when the entry is not there you're saying it doesn't work?16:14
e0negeguileo: yes16:14
geguileoe0ne: Because it should, unless I've made a mistake in one of the latest patches16:14
mriedemearlephilhower: i forgot to tell you that i got that series rebased last night16:14
geguileoe0ne: Ok, the patch should fix, and before I had to completely redo the patch 3 or 4 times because people didn't like it16:14
geguileoe0ne: It used to work16:15
mriedemi see that xing-yang already rebased the scaleio shim on top of it16:15
geguileoe0ne: I'll fix it today16:15
geguileoe0ne: Because it must be something silly with current import mechanism I have16:15
*** harlowja_at_home has quit IRC16:15
e0negeguileo: why do you think that restoring backup is a good idea?16:15
*** thangp has joined #openstack-cinder16:16
geguileoe0ne: What do you mean restoring backup?16:16
e0negeguileo: i mean restore backup io16:16
geguileoe0ne: You mean the ID16:16
geguileoe0ne: Because if you don't not only do you break Ceph backend16:16
geguileoe0ne: But you also break Incremental backup I believe16:16
e0negeguileo: you're right16:16
geguileoe0ne: Because parent-id will be referring to a non-existen backup id16:16
geguileoe0ne: Otherwise I would have just changed Ceph16:17
geguileoe0ne: But since this affects incremental16:17
e0negeguileo: but we can remove backup id from the image name16:17
geguileoe0ne: I'd rather leave Ceph as it is and fix the root16:17
geguileoe0ne: Then you break backwards compatibility :(16:17
*** leeantho has joined #openstack-cinder16:17
e0negeguileo: yes:(16:17
*** lpetrut has quit IRC16:17
geguileoe0ne: And that's a tough sell   XD16:17
*** xyang1 has joined #openstack-cinder16:18
geguileoe0ne: I'll work on fixing it after the meeting is over16:18
geguileoe0ne: And ping you once it's resolved16:18
*** david-lyle has quit IRC16:18
*** kodoku_ has quit IRC16:19
e0negeguileo: thanks16:20
e0negeguileo: i've just tested you patch again16:20
geguileoe0ne: And it still fails?16:20
e0negeguileo: restore doesn't work16:20
geguileoe0ne: Ok, I'll fix it16:20
geguileoe0ne: And thank you for testing it and pointing out the problem16:21
*** thingee has joined #openstack-cinder16:21
e0negeguileo: create volume, backup, export backup, delete original backup, import backup, try to restore imported backup16:21
geguileoe0ne: I should have done it myself, but I got tired after the 3rd time I completely redid the fix16:21
*** vokt has joined #openstack-cinder16:21
geguileoe0ne: Yep, that's the process16:21
geguileoe0ne: Thanks16:22
e0negeguileo: thanks for fixing it. i'm wating for a new patch16:22
*** sree has joined #openstack-cinder16:22
*** belmoreira has joined #openstack-cinder16:23
markus_zCould a second core have a look at please?16:24
openstackgerritPatrick East proposed openstack/cinder: Generic image-volume cache
*** afazekas has quit IRC16:27
*** sree has quit IRC16:27
*** agarciam has quit IRC16:28
*** markus_z has quit IRC16:29
*** sweston has joined #openstack-cinder16:33
e0negeguileo: i'm going like to ask question about restoring backup is in openstack-dev ml. do you have any concerns?16:34
geguileoe0ne: What is there to ask? r:-??16:35
e0negeguileo: i think that id should be one time thing16:35
geguileoe0ne: Sure, everyone says that16:36
geguileoe0ne: But then you need an additional field to store an ID for backups16:36
e0negeguileo: and you re-use it16:36
geguileoe0ne: And that ID must be unique16:36
geguileoe0ne: And it cannot be repeated when you reimport the backup16:36
e0negeguileo: two IDs is also bad16:37
geguileoe0ne: That new field really, really sound like thebackup ID16:37
*** ctracey has joined #openstack-cinder16:37
e0negeguileo: and i'm not fan of my idea to break backward compatibility16:37
geguileoe0ne: Well, you would basically break incremental backups as well...16:37
*** jdurgin1 has quit IRC16:38
e0negeguileo: i need to test it to be sure what things will be broken or not16:38
geguileoe0ne: If you want to...16:38
geguileoe0ne: But parent-id points to backup-id (if I'm not mistaken)16:38
geguileoe0ne: So it's pretty much a done deal16:38
*** coolsvap|away is now known as coolsvap16:39
*** svilgelm has joined #openstack-cinder16:39
*** jordanP has quit IRC16:39
e0negeguileo: sounds you're right. i don't remember how it is implemented16:40
*** serverascode has joined #openstack-cinder16:41
*** david-lyle has joined #openstack-cinder16:42
geguileoe0ne: I agree that nobody likes to reuse IDs16:42
geguileoe0ne: But if we don't in this case you basically cannot export/import backups16:42
*** enikher has quit IRC16:43
*** Apoorva has quit IRC16:43
*** ameade has joined #openstack-cinder16:43
*** Apoorva has joined #openstack-cinder16:43
*** ub has joined #openstack-cinder16:43
geguileoe0ne: Anyway, a message to dev list is probably the right idea16:45
*** alexpilotti has joined #openstack-cinder16:45
*** zhiyan has joined #openstack-cinder16:46
*** ganso_ has joined #openstack-cinder16:48
*** svilgelm has quit IRC16:48
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
*** crose has quit IRC16:50
*** ronis has joined #openstack-cinder16:51
*** Adriano__ has joined #openstack-cinder16:52
*** laughterwym has quit IRC16:53
*** _cjones_ has joined #openstack-cinder16:53
*** _cjones_ has quit IRC16:53
*** _cjones_ has joined #openstack-cinder16:53
*** briancurtin has joined #openstack-cinder16:57
*** lpetrut has joined #openstack-cinder16:58
*** coolsvap is now known as coolsvap|away16:58
*** coolsvap|away is now known as coolsvap16:58
*** scottda__ has joined #openstack-cinder16:59
openstackgerritMatthew Edmonds proposed openstack/cinder: move oslo.vmware to test-requirements
*** haomaiwang has quit IRC17:01
dannywilsonthingee: bp here for trying to enable discard/unmap/trim for Cinder attached volumes
dannywilsonpretty simple implementation and spec too17:01
dannywilsonthingee: just wanted to get it on your radar, thank you17:01
*** haomaiwang has joined #openstack-cinder17:02
*** e0ne has quit IRC17:03
jgriffithgeguileo: thingee got a minute ?17:03
geguileojgriffith: Yep17:03
thingeejgriffith: yes17:03
thingeedannywilson: cool, it's on my list to look at17:03
jgriffithgeguileo: thingee so I'm trying to better understand the A/A issues that folks are trying to solve17:04
thingeedannywilson: if it's already got DuncanT's approval, that's a good sign17:04
hemnadannywilson, can you explain what you mean by "discard"?17:04
geguileojgriffith: Just for the record, I have an easier solution to the most complex part of my proposal17:04
*** avishay has joined #openstack-cinder17:04
*** thingee_ has joined #openstack-cinder17:04
jgriffithgeguileo: thingee in particular IMO the biggest problem seemed to be the DB no?17:04
jgriffithgeguileo: LOL... I think you know where I'm going with this then :)17:05
geguileojgriffith: Not really17:05
geguileojgriffith: When you say the DB you mean the atomic changes?17:05
geguileojgriffith: Because that's easy17:05
*** rhefner has joined #openstack-cinder17:05
*** rex_lee has joined #openstack-cinder17:05
geguileojgriffith: If we don't change Cinder-Nova interactions17:05
jgriffithgeguileo: thingee so I don't have any issues or objections to anything... I just wanted to sort of step back and level set on this a bit17:05
patrickeasthemna: its for scsi trim/unmap/discard17:06
patrickeasthemna: lots of names depending on who is talking about it17:06
geguileojgriffith: Oh, if you have issues they are welcome!! :-)17:06
patrickeasthemna: in the nova side it got referred to as discard i think because of how libvirt enables it17:06
thingeepatrickeast: what does it do exactly?17:06
geguileoBut I have to write another posibility that it's easier for the job distribution/unstucking of resources17:06
dannywilsonhemna: yes, Nova side uses trim,
*** scottda has joined #openstack-cinder17:06
jgriffithgeguileo: thingee I've heard a couple of people going on and on about the HA problem for c-vol over the years, but no good explanation of what they want exactly17:06
patrickeastthingee: if i understand correctly it allows for flash disks to do better GC17:07
geguileojgriffith: I can't tell you what they want, but I can tell you some of the issues ;-)17:07
jgriffithgeguileo: well... that's a great first step :)17:07
dannywilsonhemna: sorry Nova uses discard for the terminology, so many similar words17:07
geguileojgriffith: First of all we have to decide how jobs get distributed to the different cluster Nodes17:08
*** angela-s has quit IRC17:08
jgriffithgeguileo: how come?17:08
thingeejgriffith: so some deployers like ebay use different cinder backends to the same storage. One c-vol does everything except delete, the other c-vol does just deletes.... and that's because their choice of storage solution is slow on deletes.17:08
thingeejgriffith: but they do out of tree hacks to accomplish this17:08
jgriffithgeguileo: what I mean to say is "is that true, do we have to care"17:08
thingeejgriffith: I know you would agree those solution just don't belong17:08
jgriffithgeguileo: as long as the db is replicated it shouldn't matter should it?17:09
geguileojgriffith: Well, it does matter17:09
thingeepatrickeast: oh right, thanks17:09
*** annashen has joined #openstack-cinder17:09
geguileojgriffith: Do we want to send jobs based on a topic queue that uses the host as reference17:09
thingeedannywilson, patrickeast yea that seems pretty straight forward17:09
jgriffiththingee: I'm sadly not really familiar with that... but that's interesting and I don't really understand it17:09
geguileoAnd then all nodes need to have the same host configured17:09
jgriffithgeguileo: I don't know :)17:10
*** wanghao has joined #openstack-cinder17:10
thingeejgriffith: they end up with a bunch of delete requests, and things slow down for other operations to come in17:10
geguileoOr do we want to configure the nodes with different host configurations17:10
thingeejgriffith: there's not really much more to explain17:10
geguileoand use the backend name as unique identifier17:10
*** angela-s has joined #openstack-cinder17:10
thingeejgriffith: it's a bandaid fix17:10
geguileoto use for the topic queue where we would replace the host with * instead of the host name17:10
*** svilgelm has joined #openstack-cinder17:10
jgriffiththingee: maybe we can circle back and you can help me understand that config better?17:11
geguileoAnyway, that's one of the decissions we need to make17:11
guitarzanthingee: they hacked up the rpc layer to do that? that's pretty wild17:11
thingeejgriffith: I don't understand it myself. I might be able to dig up what DreamHost did...they're doing something similar17:11
thingeealso using the same solution, if you can figure that out ;)17:11
jgriffithgeguileo: yeah... so what I was wondering was why you couldn't just say "host = [foo, biz, baz....]17:11
jgriffithgeguileo: and just issue the roc call to any available from the list17:11
geguileojgriffith: In the DB you mean?17:12
guitarzanI think ideally all sharing the same host would be preferred17:12
earlephilhowermriedem:  Thanks for the update on the rebase.  Been in meetings since 7:30AM, will check out at lunch.17:12
jgriffithgeguileo: no...we were talking c-vol17:12
jgriffithdb is another piece.. that I think is already solved17:12
jgriffithor "solvable"17:12
geguileojgriffith: And how does the API and the scheduler know the node to send it to?17:12
jgriffithbut I could be completely wrong too :)17:12
geguileojgriffith: Or the list of nodes?17:12
jgriffithgeguileo: what I'm saying is why should it care?  Why shouldn't17:12
jgriffithgeguileo: yes... it has a list, any node will do17:13
geguileojgriffith: Ok, then how does he maps the list of nodes that he has to the resources?17:13
jgriffithgeguileo: this assumes the c-vol node is NOT the storage of course17:13
geguileojgriffith: He receives a delete, looks in the DB the host field17:13
*** annashen has quit IRC17:13
jgriffithgeguileo: same way we do today... we just modify the definition of the "host" column17:13
geguileojgriffith: I think that doesn't really matter, because if the node is the storage you cannot configure it as A-A17:13
geguileojgriffith: So the host is a list?17:14
jgriffithgeguileo: my point is do we care what path it takes to delete the volume?  So long as the command makes it to the backend device?17:14
*** salv-orlando has joined #openstack-cinder17:14
geguileojgriffith: No, we don't17:14
nikeshmsmcginnis: hi17:14
geguileojgriffith: That's why I propose to use the same host for all nodes in the same cluster17:14
*** alexpilotti has quit IRC17:14
geguileojgriffith: And any of those nodes can pick the job17:15
geguileojgriffith: When we queue it we don't care17:15
jgriffithgeguileo: ok, so we seem to be on the same page there then17:15
geguileojgriffith: Good  :-)17:15
geguileojgriffith: But then we have the problem with the init_host of volume manager17:15
geguileojgriffith: That on start it assumes he's the only one with access to the backend17:15
jgriffithgeguileo: although I don't understand your comment "same host for all nodes in same cluster"  You mean, they all report same host name and we just treat them as a queue?17:16
geguileojgriffith: And will break any ongoing operations17:16
geguileojgriffith: Yes, they report the same host17:16
*** avishay has quit IRC17:16
*** avishay_ has joined #openstack-cinder17:16
geguileojgriffith: So they will access the same topic queue17:16
jgriffithgeguileo: ok17:16
geguileojgriffith: And we don't care how many there are getting jobs from the queue17:16
guitarzangeguileo: the HA discussions have all hinged on having some sort of distributed locking mechanism17:16
guitarzanwhich should also be the answer to init_host17:17
jgriffithguitarzan: yeah17:17
geguileoguitarzan: The DLM is for resource locking at cinder volume17:17
jgriffithguitarzan: although I'm not sure we need that TBH17:17
jgriffithwhich is what I'm getting at here17:17
geguileoguitarzan: But now we are talking about API to Volume job distribution17:17
geguileoguitarzan: Which is a step earlier17:17
guitarzanjgriffith: maybe, but init_host does do some goofy things like deleting volumes stuck in 'deleting'17:17
jgriffithguitarzan: hmm...17:18
geguileoguitarzan: Read a few lines before when I begin talking about it17:18
jgriffithguitarzan: I'll have to look at that17:18
guitarzanso they still may need locks17:18
geguileojgriffith: It's also mentioned in the post17:18
jgriffithguitarzan: honestly we shouldn't do that17:18
*** mylu has quit IRC17:18
geguileojgriffith: Since now we share the same host we need to fix the init_host17:18
guitarzanjgriffith: I don't think I disagree :)17:18
*** ub has quit IRC17:18
thingeejgriffith: why?17:19
*** annashen has joined #openstack-cinder17:19
*** xyang has joined #openstack-cinder17:19
*** mylu has joined #openstack-cinder17:19
jgriffiththingee: guitarzan ahh.. that :)  Well, that's not a terribly big deal I don't think... but anyway17:19
geguileothingee: That as well17:19
guitarzanno it isn't, it's just a possible point of contention where multiple drivers do the same thing17:20
*** ub has joined #openstack-cinder17:20
*** salv-orlando has quit IRC17:20
jgriffiththingee: I might have misunderstood; so re-issuing a delete on a volume that's "deleting" shouldn't be terrible depending on the backend17:20
guitarzanalso, I just saw the snapshot stuck in creating thing... I don't like that one at all :)17:20
geguileojgriffith: The problem is not so much deleting as the other link I've posted17:20
jgriffithguitarzan: yeah... 'creating' tasks are trickier17:21
geguileojgriffith: Where any creating operation is set to error17:21
jgriffithgeguileo: looking17:21
jgriffithgeguileo: yeah17:21
jgriffithgeguileo: so TBH though... I think this is sort of wonky17:21
geguileoSo we need a way to differentiate between stuck jobs and ongoing jobs17:21
geguileojgriffith: It made sense when only 1 node could access the backend17:22
jgriffithgeguileo: so use a db tag that says "this node owns the task"17:22
geguileojgriffith: For Active-Passive is ok17:22
thingeeI've said it in past discussions and the last summit, we should not attempt to resume creating tasks. Just give up and have the user try again.17:22
thingeeit's not worth it17:22
jgriffithgeguileo: so when you restart or bring a new c-vol online it knows "what it's allowed to do and what it's not"17:22
geguileothingee: I agree17:22
*** salv-orlando has joined #openstack-cinder17:22
jgriffiththingee: I fully agree with that17:22
geguileothingee: But we still need to know when an operation is stuck17:22
geguileothingee: To set it to error17:23
*** patrickeast has quit IRC17:23
jgriffiththingee: but I see geguileo 's point about a new service coming up currently "doesn't know"17:23
geguileothingee: Otherwise the user can't tell the difference between a slow operation and a stuck one17:23
thingeejgriffith: oh I see, yeah good point17:23
geguileoAnd maybe the sevice doesn't come up17:23
geguileoBut we still have other 2 nodes working with that backend17:23
jgriffithgeguileo: thingee so I'd propose we change our model slightly; and for the volume when it goes into creating, deleting or whatever active op, we mark what node is "doing" it17:23
*** alexpilotti has joined #openstack-cinder17:23
geguileoThat should know that resource is in error17:23
jgriffithas part of the status update17:24
*** lpetrut has quit IRC17:24
*** mylu has quit IRC17:24
thingeecan't we just the c-vol coming up ask all other c-vols what uuid's they're working with?17:24
jgriffiththen when a new service is started (or another one restarted) it knows not to mess with things that aren't under it's current task state17:24
thingeeotherwise set to error17:24
eharneyjgriffith: but what if that node never comes back up?17:24
thingeeworking with at this point in time17:24
*** ub has quit IRC17:25
geguileothingee: I don't think that's reasonable17:25
jgriffitheharney: the it's errored out...17:25
jgriffitheharney: oh... wait17:25
jgriffitheharney: that's a different case... just a minute17:25
*** mylu has joined #openstack-cinder17:25
jgriffitheharney: so I'm separating this into two cases:17:25
geguileoI think the solution would be to use the backend as the message queue topic identifier17:25
jgriffith1. Everything's "ok" I just started a new service instance17:26
jgriffith2. The actual "oh, one of my services failed"17:26
thingeegeguileo: why is it not reasonable?17:26
jgriffithSo that's where the point thingee just brought up comes in to play (situation 2)17:26
geguileojgriffith: How do you know it's yours if we are using the same host in them all?17:26
jgriffithgeguileo: a new column "assigned-node" or whatever17:26
geguileothingee: Because then you need to keep track in Volume of all ongoing operations17:26
jgriffitheach node under the host gets a UUID17:27
geguileojgriffith: Oh, that, ok17:27
jgriffiththat UUID is known by the manager; and when it sets a state on a volume it also sets that it "owns" that task17:27
thingeeSure, c-vol has a list of active uuids. If it gets a request from another c-vol of what it's working with, it sends it17:27
guitarzanthis is just a homegrown locking implementation17:28
*** lpetrut has joined #openstack-cinder17:28
guitarzanand I don't think it'd be any cleaner17:28
jgriffithguitarzan: kinda... you might be right17:28
geguileothingee: Then each node must be aware of other nodes17:28
jgriffithguitarzan: I'm kinda of the opinion that "we" in the OpenStack world like to make our solutions as complex as possible and it's not always needed17:29
geguileothingee: Or they could use the broadcast queue17:29
jgriffithgeguileo: nahh17:29
thingeegeguileo: yeah just have c-vol subscribe. they don't need to know about each other17:29
jgriffiththingee: +117:29
thingeejust that someone is doing something with that uuid17:29
jgriffiththingee: geguileo but guitarzan pointed out that this might not be any simpler/better... so maybe I'm just wasting your time :)17:29
thingeejgriffith: never give up! :P17:30
geguileothingee: But then you need cross-reference those jobs with the DB and exclude newer jobs17:30
jgriffiththingee: geguileo guitarzan I think it was worth talking through a bit tough17:30
thingeegeguileo: have c-api subscribe17:30
*** aix has quit IRC17:30
jgriffithgeguileo: so there's only a single case that this matters it seems17:30
thingee"can't do anything with this uuid, some c-vol is active with it."17:30
jgriffithgeguileo: and I honestly see a lot of this as being very similar to a problem that's already solved via scheduler17:31
thingeewe've already agreed in previous discussions we're not interested in doing parallel operations with resources17:31
jgriffithor I should say "in the" scheduler17:31
*** lpetrut has quit IRC17:31
geguileojgriffith: I don't think it's all that similar17:31
jgriffithgeguileo: thingee anyway... just some additional things to think about17:32
eharneythingee: can you clarify that a little?17:32
geguileojgriffith: I actually considered scheduler to do the job distribution and I didn't like it17:32
*** Yogi1 has joined #openstack-cinder17:32
jgriffithalso, some of this is harder for me to really dig into, because I don't necessarily agree with the philosophy17:32
geguileojgriffith: What philosophy?17:33
jgriffithphilosophy being the whole "A/A HA" thing17:33
geguileojgriffith: You don't like A-A?17:33
jgriffithgeguileo: well, not saying that...17:33
hemnaha ha philosophy17:33
jgriffithgeguileo: just saying that there are a couple of folks that I "think" make a bigger issue of that than it really is (as in the real need for it)17:34
*** lpetrut has joined #openstack-cinder17:34
jgriffithgeguileo: but that's my own personal view... take it for what it's worth17:34
jgriffithor don't :)17:34
thingeeeharney: so each c-vol and c-api are listening for messages of uuids being worked with. each keeping a list in memory and updating as things come in and are done. If c-api gets a request from the user on a particular uuid and sees it has been mentioned by a c-vol as doing something with it, it stops the requests. In c-vol land, they all know which uuids are actively being worked with and will stop operations even on their own layer in addi17:34
thingeeI have not figured out the release part if a c-vol goes down17:35
eharneythingee: ahh gotcha17:35
hemnaI'd just like to see is get rid of the local file locks in the volume manager.17:35
geguileojgriffith: I think it's a little weird to say to people that cinder does not support A-A being the block storage of a cloud infrastructure17:35
jgriffithhemna: LOL.. amen to that17:35
hemnageguileo, +117:35
jgriffithgeguileo: sure, but it's not REALLY going to support it anyway17:36
geguileojgriffith: Why??  :''-(17:36
jgriffithgeguileo: if your c-vol node goes down and you're using LVM you're still toast... if your Ceph cluster barfs, you're toast... etc etc etc17:36
thingeeso I'm probably crazy, but I think the broadcast idea of resources solves it both in c-api layer and c-vol layer as a fallback. Helping with issues with nova bypassing the api in some places17:36
hemnabut folks that typically want A A don't use LVM or ceph, they buy a solidfire/3par/ibm/etc/etc17:37
jgriffiththingee: well.. you are crazy... but not because of that opinion :)17:37
jgriffithhemna: my argument applies to all of them as well... that was the "etc" part :)17:37
jgriffithhemna: it includes every 3'rd party driver17:37
hemnajgriffith, you know you've always wanted to color your hair purple.  you are just jealous17:37
* thingee is behind on the backlog from all the talking, but it probably restating someone else's idea here17:37
geguileojgriffith: Well, if you are using LVM with local storage you CAN'T configure it as A-A17:37
jgriffithhemna: and I don't know if I agree with Ceph being in that first category17:37
jgriffithgeguileo: I know.. that's my point :)17:38
geguileojgriffith: And with Ceph you can have A-A17:38
*** patrickeast has joined #openstack-cinder17:38
hemnajgriffith, but I don't think that precludes us from fixing some "simple" things that would make A-A possible for the rest of us.17:38
jgriffithgeguileo: so that leave the case that all of this effort going towards solving of being "my c-vol service node died and I can't get to the backend"17:38
jgriffithgeguileo: seems goofy17:38
hemnaI don't think it's an either or.17:39
geguileojgriffith: And also increase workload capacity17:39
*** rhagarty_ has quit IRC17:39
thingeegeguileo: thought's?17:39
jgriffithgeguileo: so my other crazy thing is docker containers running the services across different nodes, shared storage and just spawn/fail over17:39
*** rhagarty has quit IRC17:39
jgriffithcuz it solves all these problems higher in the stack17:39
geguileojgriffith: For example if you do backups, they are "expensive" and you may need more nodes17:39
jgriffithand for the most part "just works"17:39
*** jaypipes has quit IRC17:40
jgriffithgeguileo: yeah... backups is something that can always be used as a detractor for us17:40
jgriffithgeguileo: anyway, I'm certainly not arguing against anything.  Just talking through it17:41
hemnaputting in 'ing' checks against volumes in the API, would help us get rid of the local file locks, and allow for a limited version of A-A for some backends to work.17:41
jgriffithgeguileo: educating myself a bit17:41
geguileothingee: I think that's quite a complicated solution17:41
jgriffithhemna: we already have them in most places by the way17:41
thingeegeguileo: why?17:41
hemnayah that's good17:41
geguileohemna: Checking ING is not enough17:41
thingeegeguileo: I think when people talk about DLM's that's complicated17:41
geguileohemna: Because reading operations don't have an ING status17:41
hemnageguileo, no it's not, but it has to be a first step17:41
thingeemostly because don't know how they work and throw the problems to them17:41
ccesariohello... does someone have idea about this error on create volume process!?
*** mylu has quit IRC17:42
*** annashen has quit IRC17:42
jgriffiththingee: that's why I come back to things like etc and letting an orchestration layer above us deal with it :)17:42
geguileothingee: Because I think it's almost as complicated as the one I proposed.  XD17:42
*** rhagarty has joined #openstack-cinder17:42
*** rhagarty_ has joined #openstack-cinder17:42
geguileothingee: And there is an easier way to do it that it's similar to how Neutron is doing it17:42
geguileothingee: I'll try to write something up this weekend17:43
thingeegeguileo: I disagree. c-vol and c-apis subscribe. communicate and keep in memory for check on uuid of resources.17:43
thingeethat's it17:43
openstackgerritVincent Hou proposed openstack/cinder: Adds the migration progress support for migration
openstackgerritVincent Hou proposed openstack/cinder: Volume status management during migration
thingeeI can even describe the solution in one sentence17:43
geguileothingee: Then they need to update before any operation?17:43
thingeethe updates are pushed?17:43
avishay_hemna: gentle reminder :)
*** r-daneel has quit IRC17:44
geguileothingee: Let me sleep on your solution17:44
hemnaavishay_, thanks, I haven't had time to look at reviews just yet this morning...late start17:44
geguileothingee: And I will compare it with the other solution I'm thinking about17:44
avishay_hemna: no worries, thanks!17:45
*** avishay_ is now known as avishay17:45
*** ByPasS has quit IRC17:45
geguileothingee: I'd rather think it well  ;-)17:45
thingeegeguileo: if anything I learned about working in computer science, is if the solution is simple, it probably doesn't work. That's why I was asking people if I'm crazy on suggesting this.17:45
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
geguileothingee: XD XD XD17:45
erlonthingee: Mike, can you explain the deadlines for BP approval? you said in the meeting that liberty-2 is tagged, but feature free is next week, so I'm assuming that will be possible to get changes merged until next friday right?17:46
geguileothingee: I'll think it more carefully17:46
thingeeerlon: code freeze is early september17:46
geguileoerlon: I think it's this Friday17:46
avishayjgriffith: we have a great experience using consul for HA, it's similar to etcd17:46
geguileoerlon: For spec/bp acceptance17:46
*** annashen has joined #openstack-cinder17:46
*** dannywilson has quit IRC17:46
jgriffithavishay: nice, I haven't looked at consul at all17:47
thingeeI will be looking over specs some more today to see where we're at. I appreciate people understanding that we just got done with a new milestone, whether people keep track of that stuff or not17:47
jgriffithavishay: so the paxos based stuff like ZK etc are sort of phasing out IMO17:47
hemnaavishay, ?17:47
angela-sthingee, hemna: just wanted to mention that brocade CI is up/running for zoning driver.  would be great to get final reviews on the two zoning specs, as CI was the blocker17:47
avishayjgriffith: it gives us a super scalable key-value store for metadata, DNS for all services to route requests to them, cluster membership, etc17:47
avishayhemna: yes17:47
avishayjgriffith: agreed17:48
thingeeangela-s: should I start running stats on the ci?17:48
thingeeangela-s: I was under the impression this was slow moving17:48
hemnaangela-s, nice job, congrats on getting that up.17:48
angela-sthingee: what do you mean by slow moving?17:48
*** ByPasS has joined #openstack-cinder17:48
*** edtubill has quit IRC17:48
thingeeangela-s: no one has communicated anything, until now...17:49
thingeeangela-s: so if you don't hear about things, you assume slow moving17:49
thingeeangela-s: communication is good17:49
erlongeguileo: ok, thingee, can you approve this BP, or is there any criteria for this?17:49
vilobhmmjungleboyj : If you can review this patch it would be helpful :
angela-sthingee: i see.  sorry about that.  it's up now, running on all cinder patchsets17:50
thingeeangela-s: I can start running stats on the CI today, but really I wanted to be assured what's in tree today is working17:50
*** ub has joined #openstack-cinder17:51
jungleboyjvilobhmm: Ok, I apologize I haven't been on those.  Been a wild week.17:51
angela-sthingee: ok, thank you17:51
vilobhmmjungleboyj : no worries; whenever you are free17:51
thingeeangela-s: thank you and great work to hear this moving forward!17:52
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder: Add connector object to create_export
thingeeangela-s: happy brocade has you to help with this17:52
thingeeangela-s: hope they're giving you enough recognition for keep things in tree.17:53
erlonthingee: thanks!17:53
angela-sthingee: thanks17:53
angela-sthingee: it's definitely a challenge, that's for sure17:53
thingeeangela-s: but just think how happy your customers will be17:54
thingeehemna: do we have someone from cisco in this room?17:54
thingeeroom channel whatever17:54
hemnathingee, great question.  I've not heard a single thing from cisco since Vancouver17:54
*** jasondotstar has joined #openstack-cinder17:54
thingeehemna: I think you started a ML thread17:55
thingeehemna: that I can reference17:55
patrickeastthingee: hemna: fyi the pure storage ci FC jobs have been/are currently running the cisco zone manager17:56
patrickeastsince all i could get for my system was a cisco swich :(17:56
*** mylu has joined #openstack-cinder17:56
*** e0ne has joined #openstack-cinder17:56
angela-sthingee, hemna: i was contacted by cisco engineer for CI, trying to remember his name17:56
patrickeastso its kind of getting tested17:56
thingeepatrickeast: oh17:57
thingeepatrickeast, hemna: that makes me feel good17:57
hemnathingee, I have an email thread with a guy from Cisco that included you, but I don't think it was on the ML.17:57
thingeepatrickeast: can you speak on the integration work so far?17:57
thingeehemna: alright that works17:57
thingeehemna: thanks17:57
* thingee list of things to do is growing :(17:57
patrickeastthingee: wasn’t really any integration work, just plug and play following the documentation17:58
patrickeastthingee: and it only craps out every once and a while due to parallal request issues or something (its on my list to see if thats our fault or the zone manager)17:58
*** xyang1 has quit IRC17:58
thingeeso dumb fc question from your fellow ptl, what exactly is cisco's integration with cinder then? What's the difference with all this work happening from brocade?17:58
*** mylu has quit IRC17:59
*** mylu has joined #openstack-cinder17:59
angela-sthingee: you don't want me to answer that question, i'm biased17:59
patrickeastthey have an implementation of the zone manager interface that can do cisco-specific remote api calls to the switch to configure things…17:59
patrickeasti haven’t seen anyone actually working on it before17:59
patrickeastso it looks like maybe a little bit neglected17:59
patrickeastbut does seem to mostly work18:00
hemnathingee, I found the thread.  want me to forward it to you?  It was with Al Lau18:00
thingeehemna: sure18:00
thingeeor subject18:00
thingeeI can find it in archives18:00
*** haomaiwang has quit IRC18:01
*** shyama has quit IRC18:02
hemnathingee, sent18:02
thingeehemna: I see... "hello" isn't a good subject to search for18:02
*** haomaiwang has joined #openstack-cinder18:02
hemnayah not the best18:03
thingeehemna: cool, I'll send an email to the list about the state of FC zone management in Cinder.18:04
hemnaok great thanks18:04
* thingee adds another item to the list :(18:04
thingeebusiness is booming18:04
thingeehemna: anything we need to connect on with brick?18:04
hemnaI think we are doing well right now.18:05
*** annashen_ has joined #openstack-cinder18:05
thingeehemna: I haven't looked at the recent updates with nova integration patch18:05
thingeehemna: I assume mriedem has been of help though18:06
hemnathe brick nova patch landed yesterday18:06
hemnaso it's in.18:06
hemnayah mriedem was very helpful :)18:06
*** e0ne has quit IRC18:06
thingeehemna: well good work. I don't know if it was recognized in the meeting18:07
thingeehemna: since I got cut off irc18:07
thingeehemna: but I blame smcginnis then since he was chair18:07
hemnathanks.   it was a long road.  :)18:07
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
openstackgerritVilobh Meshram proposed openstack/cinder: Nested Quota Driver: Get Project Hierarchy
*** e0ne has joined #openstack-cinder18:09
*** annashen_ has quit IRC18:10
openstackgerritabhiram moturi proposed openstack/cinder: Oracle ZFSSA iSCSI driver to return free space for non thin provisioned storage
openstackgerritabhiram moturi proposed openstack/cinder: Oracle ZFSSA iSCSI driver to return free space for non thin provisioned storage
*** jwcroppe has joined #openstack-cinder18:11
*** lpetrut has quit IRC18:12
*** angela-s is now known as angela-s|away18:12
*** ByPasS has quit IRC18:13
*** ub2 has joined #openstack-cinder18:14
mriedemthingee: hemna: regarding nova and brick, the scaleio change is here - it need some work, but some cinder eyes on the technical connect/disconnect details would be helpful18:15
mriedemkeep in mind that nova FF is tomorrow18:15
*** mylu has quit IRC18:15
mriedemand earlephilhower has to get the HGST connector for nova rebased today too
johnthetubaguymriedem: +1 thank you for helping push on those18:15
mriedemjohnthetubaguy: you're supposed to be at dinner18:16
johnthetubaguymriedem: thats what sally was hinting too, just sending the FF emial :)18:16
*** ub has quit IRC18:16
*** dannywilson has joined #openstack-cinder18:18
*** coolsvap is now known as coolsvap|brb18:18
*** lcurtis has joined #openstack-cinder18:19
*** ByPasS has joined #openstack-cinder18:19
*** mylu has joined #openstack-cinder18:20
*** annashen has quit IRC18:21
*** jasondotstar has quit IRC18:23
openstackgerritVincent Hou proposed openstack/python-cinderclient: Change cinderclient according to volume migration improvement
*** annashen has joined #openstack-cinder18:24
*** dims_ has joined #openstack-cinder18:24
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
hemnamriedem, ok I'll look at it, thanks for the heads up.  If you have anything else that needs some Cinder eyes I'm available18:25
hemnaxyang, ping18:25
*** annegentle has quit IRC18:26
*** annegentle has joined #openstack-cinder18:26
*** svilgelm has quit IRC18:27
*** dims has quit IRC18:28
*** amickus has joined #openstack-cinder18:30
*** openstackgerrit has quit IRC18:31
*** openstackgerrit has joined #openstack-cinder18:32
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
*** svilgelm has joined #openstack-cinder18:34
*** jaypipes has joined #openstack-cinder18:34
*** dannywilson has quit IRC18:34
*** dannywilson has joined #openstack-cinder18:34
*** ByPasS has quit IRC18:34
kaisersthingee: ping (regarding the Quobyte CI / Driver)18:35
*** ByPasS has joined #openstack-cinder18:36
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
*** ByPasS has quit IRC18:40
*** salv-orlando has quit IRC18:40
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
openstackgerritRyan Lucio proposed openstack/cinder: VMEM v7000: support storage pools / volume types
openstackgerritliuxinguo proposed openstack/cinder: Add smartX support for Huawei driver
*** annashen has quit IRC18:42
*** ByPasS has joined #openstack-cinder18:43
*** jwcroppe has quit IRC18:43
*** jwcroppe has joined #openstack-cinder18:44
*** jwcroppe has quit IRC18:48
openstackgerritApoorva Deshpande proposed openstack/cinder: Tintri snapshot id
*** edtubill has joined #openstack-cinder18:49
*** svilgelm has quit IRC18:52
openstackgerritThang Pham proposed openstack/cinder: Continue conversion to volume object
openstackgerritMatthew Edmonds proposed openstack/cinder: move oslo.vmware to test-requirements
openstackgerritVincent Hou proposed openstack/cinder: Adds the migration progress support for migration
*** avishay has quit IRC18:56
*** svilgelm has joined #openstack-cinder18:57
*** Adriano__ has quit IRC18:57
*** svilgelm has quit IRC18:59
*** zhonghua has joined #openstack-cinder19:00
*** haomaiwang has quit IRC19:01
*** haomaiwa_ has joined #openstack-cinder19:01
*** salv-orlando has joined #openstack-cinder19:01
xyanghemna: hi19:02
hemnaxyang, looks like you need to get this one updated.  nova FF is tomorrow.
xyanghemna: really?  I am in manila meetup today, it will work on that.  Thanks19:04
hemnaxyang, yup.  mriedem pinged me about it earlier.  FYI.19:04
*** krtaylor has quit IRC19:04
xyanghemna: thanks for the reminder19:05
hemnaxyang, np.19:05
*** zhonghua has quit IRC19:07
*** coolsvap|brb is now known as coolsvap|away19:10
*** IanGovett has joined #openstack-cinder19:13
*** krtaylor has joined #openstack-cinder19:17
*** vincent_hou has quit IRC19:18
*** jasondotstar has joined #openstack-cinder19:18
*** lpetrut has joined #openstack-cinder19:18
*** IanGovett has quit IRC19:19
*** dims_ has quit IRC19:27
*** dims has joined #openstack-cinder19:27
thingeekaisers: just keep the convo on the ML thread. If I haven't gotten to it yet, it's because I have everyone pinging me. :)19:31
thingeekaisers: unfortunately I can't drop everything at the moment.19:31
thingeekaisers: thanks19:31
thingeemriedem: adding this to my list19:31
*** sgotliv_ has quit IRC19:32
*** sgotliv has joined #openstack-cinder19:32
*** salv-orlando has quit IRC19:35
*** ByPasS has quit IRC19:35
*** ByPasS has joined #openstack-cinder19:35
*** annashen has joined #openstack-cinder19:42
openstackgerritErlon R. Cruz proposed openstack/cinder: Fix HNAS iSCSI 32 targets limitation error
openstackgerritErlon R. Cruz proposed openstack/cinder: Fix concurrent attaches on HNAS iSCSI driver
kaisersthingee_: ok19:44
openstackgerritErickson Filipe Guedes dos Santos proposed openstack/cinder: Nested Quota Driver: Get Project Hierarchy
openstackgerritErickson Filipe Guedes dos Santos proposed openstack/cinder: Cinder Nested Quota Driver
*** krtaylor has quit IRC19:46
*** ByPasS has quit IRC19:47
*** belmoreira has quit IRC19:47
*** HenryG has quit IRC19:49
*** e0ne has quit IRC19:51
*** salv-orlando has joined #openstack-cinder19:51
*** timcl has quit IRC19:52
*** HenryG has joined #openstack-cinder19:52
*** ByPasS has joined #openstack-cinder19:52
*** e0ne has joined #openstack-cinder19:55
*** mylu has quit IRC19:56
*** krtaylor has joined #openstack-cinder19:59
*** haomaiwa_ has quit IRC20:01
*** lpetrut has quit IRC20:01
*** edmondsw has quit IRC20:01
*** haomaiwa_ has joined #openstack-cinder20:02
*** thangp has quit IRC20:02
*** mylu has joined #openstack-cinder20:03
*** e0ne has quit IRC20:03
*** hodos has joined #openstack-cinder20:03
*** annashen_ has joined #openstack-cinder20:06
*** bill_az has joined #openstack-cinder20:06
*** Strony has joined #openstack-cinder20:08
openstackgerritliuxinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
*** ByPasS has quit IRC20:09
*** ByPasS has joined #openstack-cinder20:09
*** annashen_ has quit IRC20:11
*** e0ne has joined #openstack-cinder20:11
*** liuxg has quit IRC20:13
*** openstackgerrit has quit IRC20:16
*** openstackgerrit has joined #openstack-cinder20:17
*** annegentle has quit IRC20:17
*** jasondotstar has quit IRC20:19
*** xyang1 has joined #openstack-cinder20:26
xyanghemna: mriedem , hi, I submitted another patch:
xyanghemna: mriedem,  please take a look20:27
*** Lee1092 has quit IRC20:28
*** mylu has quit IRC20:28
*** mylu has joined #openstack-cinder20:28
*** xyang1 has quit IRC20:29
hemnaxyang, ok thanks20:29
hemnaxyang, does that root_wrap filter work?20:30
xyanghemna: thank you20:30
hemnaI'm hacking around trying to add a new /lib/udev/scsi_id CommandFilter and can't get the filter to work at all20:30
hemnaI had to add /lib/udev to the paths in rootwrap.conf to make it work20:31
xyanghemna: seems work for us20:31
hemnaok, I'll try hacking some more20:32
*** jasondotstar has joined #openstack-cinder20:32
hemnaI'd like to not have to add /lib/udev to the rootwrap.conf paths20:32
xyanghemna: ha20:33
xyanghemna: wrong word20:33
xyanghemna: where is it? It is not in there20:34
*** kaisers has quit IRC20:34
hemnathis is on a patch I'm working on locally20:34
xyanghemna: ok20:34
hemnaxyang, so the ScaleIOConnector has the command being executed as drv_cfg, but the rootwrap filter lists the entire path.20:35
hemnaI'm curious how that works20:35
hemnasince the /opt/emc/scaleio/sdc/bin isn't in the env path anywhere20:36
*** salv-orlando has quit IRC20:36
xyangHemna: there is other lib needs to be installed20:37
xyanghemna: the Scaleio sdc20:37
hemnaI just updated my filter to include the full path20:38
hemnabut only call scsi_id from the putils.exec and it seems to work as well20:38
hemnaoh...I think I still have the paths addition in rootwrap.conf though20:38
hemnathat still works20:38
hemna :)20:38
xyanghemna: which one?:)20:39
hemnamy patch20:39
hemnaI was calling /lib/udev/scsi_id in my python code's putils.exec20:39
hemnaand the filter was puking20:39
hemnahad to change it to simply call 'scsi_id'20:40
hemnaand the filter found it and worked20:40
xyanghemna: oh, ok20:40
jgriffiththingee: guitarzan geguileo eharney Ok... I've voiced my opinion here:
jgriffithI'll stay out of the way if the consensus is that I'm "crazy"20:40
jgriffithand also fully support and help no matter what direction we go with things20:41
*** ronis has quit IRC20:41
guitarzanjgriffith: does you being crazy have any other impact here? :)20:41
geguileojgriffith: Thanks for reviewing the spec20:41
jgriffithguitarzan: LOL20:41
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
geguileojgriffith: I'll read and reply tomorrow, as I was just about to go have dinner (it's 22:42 here)20:42
jgriffithgeguileo: Oh, no rush20:42
hemnaso, how does the load balancing work when one node has a local file lock operating on a volume, and another request for some other action happens on the same volume getting balanced to another node ?20:43
geguileojgriffith: Well, Friday is the last day for specs, so there is some hurry  ;-)20:43
hemnaif we are going to use a higher level package to do this?20:43
guitarzanjgriffith: that's a pretty good post, but I think zookeeper is probably a better example than docker20:44
geguileohemna: Local file locks are only the firs step20:44
geguileohemna: They will only be used for A-P configurations20:44
hemnaI think for this spec to have value, we have to nuke the local file locks20:44
geguileohemna: For A-A a DLM will be needed20:44
hemnabut I'm sure I'm missing something20:44
geguileohemna: But then only a configuration change will be needed20:44
hemnaor confused, or stupid....or both :)20:45
geguileohemna: I don't know if I was able to explain it properly20:45
jgriffithguitarzan: blahhh... ZK20:46
*** salv-orlando has joined #openstack-cinder20:46
guitarzanI think whatever route is "blessed" you have to deal with multiple c-vols for the same backend, and that'll require some level of cooperation20:46
*** Yogi1 has quit IRC20:46
hemnaguitarzan, yup, and the local file locks are a big blocker to that working right now.20:47
guitarzanhemna: no argument there at all20:47
geguileoThe idea is to use Tooz with a DLM for Active-Active configurations, but as a first step we just replace current local locks with Tooz local locks, thus leaving everything as it is. And for Active-Passive configuration this will not change, but for Active-Active cinder will be configured to use a DLM (one of those provided by Tooz, like Zookeeper or Redis).20:47
jgriffithguitarzan: yeah, or a backend that's just able to behave in a multi-tenant env.  Which is the second/bigger part of the problem here IMO.20:47
hemnaand I don't think switching those local file locks to tooz locks is really the right way to go.20:47
geguileoguitarzan: That's what Tooz and the DLM are for20:47
jgriffithgeguileo: go eat!!!20:47
geguileoguitarzan: Instead fo local file locks you use a DLM (redis or Zookeeper)20:47
jgriffithgeguileo: I didn't mean to stir all this up again late in the day20:48
geguileojgriffith: I want to, but the conversation here is interesting!!  ;-)20:48
guitarzangeguileo: I think I'm on the same page as you20:48
jgriffithgeguileo: FWIW, I think guitarzan was agreeing with you20:48
geguileoXD XD XD20:48
hemnaok, so if this is the first baby step towards moving to atomic volume operations utilizing the API, then I'm ok with it.20:48
* jgriffith is the only dissenting voice :)20:48
geguileoOk, then I must rest if I'm arguing with those that agree with me  XD20:48
guitarzanI think there is something to be said to just using db state transitions as the only real source of locks20:49
hemnaheh, we churned on that in Vancouver20:49
guitarzanyes sir20:49
geguileohemna: Will you be introducing a new Reading state for resources?20:49
guitarzansomeone was trying to argue about using a non acid db20:49
geguileohemna: I mean to be able to return VolumeIsBusy when the resource is being used for reading20:49
guitarzangeguileo: not specifically20:49
*** bswartz has quit IRC20:49
hemnageguileo, that will have to be part of it, but Nova needs to cope as well20:50
hemnawhich is the bigger hurdle IMO20:50
geguileohemna: Yes, I agree on the Nova part (pita)20:50
geguileohemna: I was just wondering on how you intend to remove the locks using just the status20:50
*** edmondsw has joined #openstack-cinder20:50
geguileohemna: Because when I've been looking at doing the locking with only the DB I've found that it's even more complicated20:51
guitarzangeguileo: why is that?20:51
hemnathe first step is to get nova updated to deal w/ VolumeIsBusy first.20:51
guitarzanisn't it easier to just fail any call where the thing isn't in the right state?20:51
geguileoguitarzan: Because we don't change the status of the resource when reading20:51
guitarzanit is much more difficult to figure out if something dies though20:51
guitarzangeguileo: I'm not sure I understand the concept of "reading"20:51
geguileoguitarzan: So in the API we have no idea when a resource is being used for reading (cloning a volume for example)20:51
geguileoguitarzan: Reading is when a resource is being used for another operation, but just for reading20:52
geguileoguitarzan: Like creating a volume with another as the source (cloning)20:52
geguileoguitarzan: Or creaing a volume from a snapshot20:52
guitarzanwhy does reading matter vs any other operation?20:52
hemnageguileo, if a volume is being cloned via cinder, then the volume's state should be busy20:52
geguileohemna: But we don't have that right now20:52
guitarzangeguileo: sure, I get it20:52
*** dustins has joined #openstack-cinder20:52
guitarzanso yeah, those are two good examples20:53
geguileohemna: That's why I was asking if you were going to introduce new statuses20:53
hemnaah I see.  yah that's a good point, and another portion of the solution I suppose.20:53
hemnasomething to talk about next week20:54
guitarzanI think the state only solution only has the problem of no heartbeats20:54
guitarzanother than that I think it'd work20:54
geguileoguitarzan: I think there's a way to fix that20:54
guitarzanI'm interested in hearing what that would look like20:54
geguileoguitarzan: But I want to give it a good thought together with the idea that thingee proposed earlier20:54
guitarzansure, you also need to go sleep :)20:55
geguileoguitarzan: First dinner  ;-)20:55
guitarzanno rush20:55
guitarzantrue, food is good20:55
geguileoSee you tomorrow guys20:55
*** ByPasS has quit IRC20:55
*** akerr has quit IRC20:56
*** ByPasS has joined #openstack-cinder20:56
*** rmcall has joined #openstack-cinder20:56
*** dustins has left #openstack-cinder20:56
*** Guest100_ has quit IRC20:56
*** dustins has joined #openstack-cinder20:56
*** haomaiwa_ has quit IRC21:01
*** 6JTABPDW5 has joined #openstack-cinder21:02
*** jordanP has joined #openstack-cinder21:07
*** mylu has quit IRC21:09
*** edtubill has left #openstack-cinder21:09
*** dustins has quit IRC21:09
*** mylu has joined #openstack-cinder21:10
*** ByPasS has quit IRC21:13
*** ekarlso has quit IRC21:16
*** ByPasS has joined #openstack-cinder21:18
mriedemnot helpful:
*** e0ne has quit IRC21:24
*** mylu has quit IRC21:25
*** mylu has joined #openstack-cinder21:25
*** mylu has quit IRC21:25
jgriffithmriedem: nope, not really21:26
openstackLaunchpad bug 1479530 in Cinder "emc scaleio uses drv_cfg exe which isn't in volume.filters" [Undecided,New]21:27
mriedemxyang: ^21:27
xyangmriedem: thanks:). I'll fix it21:27
*** cbader has quit IRC21:28
xyangmriedem: our Scaleio CI is in the middle of transition, that is why that link is empty21:34
mriedemxyang: ah21:36
mriedemmoving from china to the US?21:36
mriedemi only ask b/c that's a common theme21:36
xyangmriedem: moving from a contractor to internal21:37
mriedemoh that's fun too21:37
mriedemintel recently had some issues with their ci in china being blocked by the firewall21:37
mriedemwas down for about a month and no one realized :(21:37
xyangmriedem: let me find an older link, I see that we used our own filter21:37
*** akerr has joined #openstack-cinder21:38
*** akerr has quit IRC21:38
xyangmriedem: the contractor CI is still standing by and running os-brick CI21:38
xyangUntil end of August21:38
*** annashen has quit IRC21:39
xyangmriedem: go to Scaleio cinder CI and go to controller node21:40
*** rmcall has quit IRC21:41
*** jasondotstar has quit IRC21:41
xyangmriedem: ya, I see that , we need to fix that after it is added to volume filters21:42
*** annashen has joined #openstack-cinder21:43
*** ekarlso has joined #openstack-cinder21:44
*** BharatK has quit IRC21:48
hemnahrmm jungleboyj Storewize CI looks broken21:49
*** ByPasS has quit IRC21:51
*** bswartz has joined #openstack-cinder21:53
*** ByPasS has joined #openstack-cinder21:53
*** kvidvans has joined #openstack-cinder21:58
kvidvansthingee: Hi Mike, A BP i filed, needs your approval . Let me know if i need to add any other details to it for approval. Thanks22:00
*** 6JTABPDW5 has quit IRC22:01
*** markvoelker has quit IRC22:02
*** 16WAAEHWP has joined #openstack-cinder22:02
*** erlon has quit IRC22:03
*** BharatK has joined #openstack-cinder22:04
*** mriedem has quit IRC22:04
openstackgerritabhiram moturi proposed openstack/cinder: Fix Oracle ZFSSA iSCSI driver to return free space
*** annashen_ has joined #openstack-cinder22:07
openstackgerritabhiram moturi proposed openstack/cinder: Oracle ZFSSA iSCSI driver to return free space for non thin provisioned storage
jungleboyjhemna: Thanks.  Was working when I looked yesterday.  I will ping our person in a bit.22:10
*** daneyon has quit IRC22:11
hemnajungleboyj, ok thanks man.22:11
*** annashen_ has quit IRC22:12
jungleboyjGotta drop for a bit.22:12
*** hodos has quit IRC22:14
*** jungleboyj has quit IRC22:17
thingeekvidvans: done22:18
hemnawhen you have time...please :)22:21
hemnaI've had that one up for a while, including spec and code. fwiw.22:22
*** jasondotstar has joined #openstack-cinder22:28
thingeehemna: about that spec, the use case didn't make sense to me.22:29
*** jordanP has quit IRC22:29
thingeehemna: I remember jgriffith mentioning that the way we expect drivers to be doing exports is a bit of a mess.22:30
thingeerather not exports, but just doing attaches because of what some drivers end up doing. jgriffith can you speak on that again?22:30
jgriffiththingee: well... I was just pointing out that some drivers diverged and did "their own thing"22:31
jgriffiththingee: hemna haven't looked at the spec.... I'll look now22:32
hemnathingee, according to jgriffith's blog post about the subject and my research of live migration, we need that change.22:32
thingeesorry which blog post?22:33
hemnathe idea of create_export is to do just that, create the target exported on the array.  There are many drivers that cannot do that without having that connector object passed in to create_export.22:33
hemnaand is most likely why there are quite a bit of drivers that do nothing inside of create_export and do everything inside initialize_connection now.22:34
hemnainitialize_connection is supposed to be just collecting data to return.  not doing exports.22:34
*** akerr has joined #openstack-cinder22:35
*** ByPasS has quit IRC22:35
hemnaI documented the list of drivers that do exporting and other things inside of initialize_connection today, on the CinderNovaAPI etherpad.  I think there are about 30
*** ByPasS has joined #openstack-cinder22:37
jgriffithhemna: seems like that works... only question I would have which is what thingee was mentioning I think is as I mentioned the other week....22:38
jgriffithhemna: maybe the right answer would be to modify the drivers to all "do" the same thing22:38
jgriffithhemna: in other words, create_export may not really "do" much in your driver, and the initialize_connection is where you would update things and return a model-update etc22:39
jgriffithhemna: I'm not saying that's better/worse22:39
jgriffithhemna: just think that might be the comment I made that's in question.22:39
jgriffithhemna: and the reason I made that comment is not that I'm opposed to passing the connector in on export...22:39
hemnaI'd just like the api to be consistent22:39
hemnaand right now it's not22:40
jgriffithhemna: but I am sort of unhappy that there are different interpretations of what create_export does22:40
jgriffithhemna: yes... we're saying the same thing I think22:40
hemnabecause some drivers do stuff here create_export and others in init_conn,22:40
jgriffithhemna: but you're change doesn't fix that22:40
SwansonIf everything stays in initialize_connection you just need to make sure you can handle being called more than once without blowing up (I couldn't).  Then I wouldn't have to make any create_export tests.22:40
jgriffithhemna: it's still inconsistent no?22:40
hemnamy change allows driver developers to fix it22:40
hemnait's impossible otherwise22:40
jgriffithSwanson: coorect22:40
jgriffithhemna: well... I'm not saying i don't like your change.  Let's just be clear22:41
jgriffithhemna: but I am saying it's not really the only option22:41
hemnayah I get that totally.22:41
SwansonOn the flip side with hemna's change I can actually create the export in create_export.  Which I can't do now.22:41
hemnamaybe I can add that to the alternatives in the spec?22:41
jgriffithSwanson: right22:41
hemnaI just like the idea of folks using the API for what it's meant for22:41
jgriffithhemna: honestly I'm unclear on the need for a spec.. .this is pretty much a bug IMO22:41
jgriffithbut that's not up to me22:41
hemnaotherwise, why have create_export at all.22:42
jgriffithhemna: well... that's kinda the next question :)22:42
hemnaI figured since it was a change to the driver API, it required a spec22:42
thingeeyeah, I'm not sure a spec is needed22:42
jgriffithhemna: meh.. it sure doesn't hurt anything22:42
hemnawas just trying to do the right thing fwiw22:42
*** angela-s|away is now known as angela-s22:42
jgriffithhemna: so FWIW, I'm ok with your patch; that was pretty much my first thought on how to fix it as well.22:43
*** ByPasS has quit IRC22:43
thingeejgriffith: perhaps this is the first step for us to have drivers to what we expect.22:43
SwansonI'm offended that we have something called "initialize_connection" where we might not be initializing a connection but rather returning information about one.22:43
jgriffithhemna: but it doesn't fix the problem of inconsistent behaviors/expectations which is a bummer22:43
hemnathingee, +122:43
*** jaypipes has quit IRC22:43
*** ByPasS has joined #openstack-cinder22:43
jgriffiththingee: well.. it gives them an "either or" which is what lead us to problems to begin with.  And IMHO it will lead to more of that and likely cause a problem in the future22:44
jgriffithhemna: thingee again... not saying this isn't the correct answer for now22:44
hemnajgriffith, this is true.  I think it's a 2 phase approach.  1) get this in and 2) file bugs against drivers? and/or be more diligent with driver reviews to ensure the consistency.22:44
jgriffithhemna: thingee but I do believe there's deeper follow up work to be done22:44
hemnaI have plans to update our 4 drivers after this lands (if it lands) to use create_export as it was intended.22:44
jgriffithhemna: yeah, that's cool22:45
Swansonhemna: +1.  I'll move to create_export once it functions such that it can.22:45
*** esker has quit IRC22:46
thingeejgriffith: I'm unsure otherwise. Some target drivers need the connector info for initialize_connection22:46
*** dshaw_ has joined #openstack-cinder22:46
*** dshaw_ has quit IRC22:46
thingeejgriffith: iscsi for multipath check, lio for initiator key22:47
jgriffiththingee: right... so this change is ok because it doesn't break that22:47
thingeejgriffith: well I'm saying if we don't want the either'll be difficult to remove connector from initialize_connection22:47
jgriffiththingee: oh.. yeah, maybe22:47
hemnathingee, we don't want the connector removed from initialize_connection though.22:48
jgriffiththingee: I think it's doable though personally.  The other thing is that rather than modify create_export, the other option here was to modify intiialize_connection to receive a model update and update there if present22:48
hemnait contains some info that we need to find the right exported target22:48
thingeehemna: right, so I'm just saying I'm having a hard time of thinking how to enforce things via code instead of us being diligent in reviews.22:48
jgriffiththingee: which wouldn't require changing the signature... just an optional return (tuple) that could be None22:48
hemnawe have to be more diligent in reviews either way :)22:49
*** samueldmq has quit IRC22:49
hemnaso if we want to go the route of just doing everything in initialize_connection, then I don't see the need to even keep create_export.22:50
hemnait's confusing as heck22:50
*** lcurtis has quit IRC22:50
*** ericksonsantos has quit IRC22:50
jgriffithhemna: well.. more than half of the drivers are still using that :)22:50
*** abehl has quit IRC22:50
hemnadamned if we do.....damned if we don't.....22:50
patrickeastfwiw it would probably be beneficial either way to add more documentation into the base driver class explaining what the methods are actually for22:50
*** ByPasS has quit IRC22:50
jgriffithhemna: kinda22:50
jgriffithpatrickeast: where in the base driver?22:51
jgriffithpatrickeast: there are 4 instances of that method in the base driver now22:51
hemnapatrickeast, I added that docs in my cinder patch related to the spec fwiw22:51
jgriffithpatrickeast: which one gets documented/updated?22:51
thingeepatrickeast: you just opened a whole other can of worms.22:51
patrickeastat least one needs it22:51
jgriffithpatrickeast: hehe22:51
jgriffithpatrickeast: hemna thingee so IMO we just need to add a "driver-development" page to the docs22:52
jgriffithand remember it's there and update it :)22:52
jgriffithor merge my patch that has a single base driver :)22:52
*** ByPasS has joined #openstack-cinder22:52
jgriffithanyway.. thingee hemna I'm fine with adding the connector info to create_export22:52
hemnaa driver dev docs page would be nice to have.22:52
earlephilhowerhemna: +1000022:53
openstackgerritKurt Martin proposed openstack/cinder: Over subscription for HP LeftHand iSCSI driver
*** markvoelker_ has joined #openstack-cinder22:57
*** ByPasS has quit IRC22:59
*** 16WAAEHWP has quit IRC23:01
kmartinthingee, if you have a minute I added a driver BP that needs your approval
patrickeast^ do all driver ehancement type features like that need approved bp’s23:02
*** haomaiwang has joined #openstack-cinder23:02
kmartinpatrickeast, I think so, you added one for this feature23:02
thingeeI've done it for tracking purposes.23:03
patrickeastkmartin: haha yea, i have added them in the past and been told it wasn’t really needed23:03
* thingee has never changed his answer23:03
patrickeastthingee: in that case i might have a couple little ones real quick for the pure drivers23:03
kmartinwhat thingee said, helps for tracking and review of what's going in the release23:03
patrickeastyea that makes sense to me23:04
*** ByPasS has joined #openstack-cinder23:04
*** alexpilotti has quit IRC23:05
*** daneyon has joined #openstack-cinder23:09
*** Apoorva has quit IRC23:10
*** daneyon_ has joined #openstack-cinder23:10
patrickeastthingee: ok, sorry for the last minute request, if you could take a look and approve these it would be much appreciated23:13
patrickeastthey should be pretty small changes, just pending other features that are slipping into L-323:13
*** daneyon has quit IRC23:13
kmartinpatrickeast, driver only BPs do not have the strict deadline that general cinder BPs have if I remember correctly.23:15
kmartinthingee, ^ correct?23:15
patrickeasthaha yea, i’ve heard that before too23:15
kmartinme too23:16
patrickeastbut i’m not having a good day for remembering correct things :(23:16
patrickeastbetter safe than sending a FFE email23:16
thingeeyes and the point is to have a cut.23:16
thingeetake a look at the l-3 page. plenty to review soon23:17
*** diogogmt has quit IRC23:17
kmartinpatrickeast, you have me scared now, going to go add the 3PAR over subscription BP now :)23:17
dannywilsonkmartin: hah, did the phrase "FFE email" do it? :)23:18
kmartindannywilson, lol, we have to have a cut line at some point like Mike said.23:19
openstackgerritAnthony Lee proposed openstack/cinder: Add multipath support to 3PAR iSCSI driver
openstackgerritAnthony Lee proposed openstack/cinder: Add multipath support to 3PAR iSCSI driver
*** annegentle has joined #openstack-cinder23:28
*** cdelatte has joined #openstack-cinder23:28
*** annashen has quit IRC23:28
*** daneyon_ has quit IRC23:29
*** delattec has quit IRC23:30
leeanthothingee, when you have a moment can you approve this driver blueprint?
*** zhenguo has joined #openstack-cinder23:31
*** ByPasS has quit IRC23:32
thingeetbarron: ping23:32
*** jasondotstar has quit IRC23:33
*** ByPasS has joined #openstack-cinder23:34
*** IlyaG has quit IRC23:38
*** jamielennox is now known as jamielennox|away23:39
*** mugithi has joined #openstack-cinder23:41
*** Apoorva has joined #openstack-cinder23:45
*** jasondotstar has joined #openstack-cinder23:45
*** akshai has quit IRC23:46
*** mugithi has quit IRC23:47
*** rmcall has joined #openstack-cinder23:47
*** angela-s has quit IRC23:53
*** daneyon has joined #openstack-cinder23:53
*** amickus has quit IRC23:54
*** akshai has joined #openstack-cinder23:55

Generated by 2.14.0 by Marius Gedminas - find it at!