Monday, 2015-02-16

*** kragniz_ is now known as kragniz00:00
*** markvoelker has joined #openstack-cinder00:02
*** smoriya has joined #openstack-cinder00:02
*** markvoelker has quit IRC00:08
*** ebalduf has quit IRC00:10
*** Yogi1 has quit IRC00:16
*** ho has joined #openstack-cinder00:43
*** Lee1092 has joined #openstack-cinder00:46
*** ho has quit IRC00:47
*** ho has joined #openstack-cinder00:51
*** markvoelker has joined #openstack-cinder00:52
*** markvoelker has quit IRC00:57
*** liusheng has quit IRC01:02
*** Yogi1 has joined #openstack-cinder01:09
*** ebalduf has joined #openstack-cinder01:11
*** ebalduf has quit IRC01:16
*** Yogi1 has quit IRC01:26
openstackgerritwanghao proposed openstack/cinder: non-admin user to query volume filter by az
*** scottda_ has joined #openstack-cinder01:42
*** scottda_ has quit IRC01:46
*** markvoelker has joined #openstack-cinder01:54
*** markvoelker has quit IRC01:58
*** thrawn01 has quit IRC02:03
*** thrawn01 has joined #openstack-cinder02:07
*** ebalduf has joined #openstack-cinder02:12
*** ebalduf has quit IRC02:17
*** jamielennox is now known as jamielennox|away02:19
*** jamielennox|away is now known as jamielennox02:27
*** ebalduf has joined #openstack-cinder02:39
*** kaufer has joined #openstack-cinder02:39
*** ebalduf has quit IRC02:39
*** ebalduf has joined #openstack-cinder02:41
openstackgerritwanghao proposed openstack/cinder: Implement function of import/export snapshots
*** Yogi1 has joined #openstack-cinder02:42
*** markvoelker has joined #openstack-cinder02:55
*** mberlin has joined #openstack-cinder02:56
*** kaisers1 has joined #openstack-cinder02:56
*** kaisers has quit IRC02:57
*** mberlin1 has quit IRC02:58
*** Yogi1 has quit IRC03:00
*** markvoelker has quit IRC03:01
*** jaypipes has joined #openstack-cinder03:10
*** Yogi1 has joined #openstack-cinder03:24
*** harlowja_at_home has joined #openstack-cinder03:34
*** Yogi1 has quit IRC03:44
*** markvoelker has joined #openstack-cinder03:57
*** markvoelker has quit IRC04:02
*** harlowja_at_home has quit IRC04:07
openstackgerritwanghao proposed openstack/cinder: Implement function of import/export snapshots
openstackgerritSteven Kaufer proposed openstack/cinder: Replication status periodic task optimization
*** Mandell has joined #openstack-cinder04:12
*** kaufer has quit IRC04:21
*** ebalduf has quit IRC04:24
*** Teal has left #openstack-cinder04:39
*** pradipta has joined #openstack-cinder04:40
*** markvoelker has joined #openstack-cinder04:58
*** theanalyst has quit IRC05:03
*** markvoelker has quit IRC05:03
*** theanalyst has joined #openstack-cinder05:05
openstackgerritLiu Xinguo proposed openstack/cinder: Huawei driver check before associating LUN to a LUN group
*** Ilja has joined #openstack-cinder05:16
openstackgerritLiu Xinguo proposed openstack/cinder: Huawei driver fix problems under multipath
openstackgerritLiu Xinguo proposed openstack/cinder: Huawei driver remove LUN controller change
*** annashen has joined #openstack-cinder05:34
*** anshul has joined #openstack-cinder05:44
*** markvoelker has joined #openstack-cinder05:59
*** markvoelker has quit IRC06:04
*** sgotliv has joined #openstack-cinder06:06
*** ankit_ag has joined #openstack-cinder06:08
*** Mandell has quit IRC06:09
*** Mandell has joined #openstack-cinder06:19
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Imported Translations from Transifex
*** Mandell has quit IRC06:40
*** Murugan_ has joined #openstack-cinder06:42
*** annashen has quit IRC06:46
openstackgerritAmit Saha proposed openstack/cinder: Fixed typo
*** vnagpal has joined #openstack-cinder06:47
openstackgerritAmit Saha proposed openstack/cinder: Fixed typo
*** vilobhmm has joined #openstack-cinder06:58
openstackgerritPranali Deore proposed openstack/cinder: Snapshot of bootable volume goes in error state
*** markvoelker has joined #openstack-cinder07:01
*** nkrinner has joined #openstack-cinder07:02
openstackgerritAmit Saha proposed openstack/cinder: Fixed a couple of Typos
*** markvoelker has quit IRC07:06
*** theanalyst has quit IRC07:10
*** sgotliv has quit IRC07:12
openstackgerritAmit Saha proposed openstack/cinder: Fixed typo
*** theanalyst has joined #openstack-cinder07:13
*** Mandell has joined #openstack-cinder07:18
*** avishay has joined #openstack-cinder07:24
*** pradipta has quit IRC07:31
*** afazekas_ has joined #openstack-cinder07:35
*** junhongl has quit IRC07:40
*** zhipeng has joined #openstack-cinder07:45
*** chlong has quit IRC07:51
*** nellysmitt has joined #openstack-cinder07:53
*** vilobhmm has quit IRC07:57
*** markvoelker has joined #openstack-cinder08:02
*** avishay has quit IRC08:04
*** Mandell has quit IRC08:05
*** Mandell has joined #openstack-cinder08:06
*** markvoelker has quit IRC08:07
*** Mandell has quit IRC08:07
*** Mandell has joined #openstack-cinder08:07
*** jistr has joined #openstack-cinder08:08
*** takedakn has joined #openstack-cinder08:13
*** mberlin has quit IRC08:14
*** takedakn has quit IRC08:19
*** mberlin has joined #openstack-cinder08:25
*** dulek has joined #openstack-cinder08:40
*** jordanP has joined #openstack-cinder08:42
*** ronis has joined #openstack-cinder08:42
*** takedakn has joined #openstack-cinder08:42
*** jordanP has quit IRC08:44
*** jordanP has joined #openstack-cinder08:44
*** annashen has joined #openstack-cinder08:47
*** mberlin has quit IRC08:50
*** annashen has quit IRC08:52
*** deepakcs has joined #openstack-cinder08:53
*** takedakn has quit IRC08:54
*** nlevinki has joined #openstack-cinder08:56
*** openstackgerrit has quit IRC08:57
*** openstackgerrit has joined #openstack-cinder08:58
*** ndipanov has joined #openstack-cinder08:59
*** karimb has joined #openstack-cinder08:59
*** jistr has quit IRC08:59
*** Longgeek has joined #openstack-cinder09:00
*** markvoelker has joined #openstack-cinder09:03
*** markvoelker has quit IRC09:08
*** ho has quit IRC09:14
*** boichev has quit IRC09:21
*** Ilja has quit IRC09:29
*** dalgaaf has joined #openstack-cinder09:40
openstackgerritYuriy Nesenenko proposed openstack/cinder: Fix comments style according to the Hacking Rules
*** zhipeng has quit IRC09:42
*** Mandell has quit IRC09:43
*** theanalyst has quit IRC09:47
*** theanalyst has joined #openstack-cinder09:47
*** alecv has joined #openstack-cinder09:48
*** yuriy_n17 has joined #openstack-cinder09:56
*** takedakn has joined #openstack-cinder10:01
*** obutenko_ has quit IRC10:03
*** markvoelker has joined #openstack-cinder10:04
*** alecv has quit IRC10:04
*** markvoelker has quit IRC10:09
*** takedakn has quit IRC10:14
*** lpetrut has joined #openstack-cinder10:16
*** e0ne has joined #openstack-cinder10:20
*** rushiagr_away is now known as rushiagr10:23
*** pradipta has joined #openstack-cinder10:25
*** rushiagr is now known as rushiagr_away10:40
*** aix has joined #openstack-cinder10:44
openstackgerritAnton Arefiev proposed openstack/cinder: Import only modules: H302
*** annashen has joined #openstack-cinder10:48
*** annashen has quit IRC10:53
*** e0ne is now known as e0ne_10:57
openstackgerritJulien Danjou proposed openstack/cinder: Remove useless requirement on wsgiref
*** e0ne_ is now known as e0ne11:00
*** markvoelker has joined #openstack-cinder11:05
*** markvoelker has quit IRC11:11
*** casusbelli has joined #openstack-cinder11:20
*** aix has quit IRC11:27
*** sgotliv has joined #openstack-cinder11:30
*** Ilja has joined #openstack-cinder11:30
openstackgerritRonen Mesonzhnik proposed openstack/cinder: Provide backup service that uses ProtecTIER as the backend
*** Ilja has quit IRC11:35
openstackgerritMarc Koderer proposed openstack/cinder: Adapt volume manager to new driver structure
*** Ilja has joined #openstack-cinder11:36
*** Ilja1 has joined #openstack-cinder11:39
*** Ilja has quit IRC11:39
*** sgotliv has quit IRC11:40
openstackgerritRonen Mesonzhnik proposed openstack/cinder: Provide backup service that uses ProtecTIER as the backend
*** e0ne is now known as e0ne_11:45
*** boichev has joined #openstack-cinder11:48
*** sgotliv has joined #openstack-cinder11:54
*** e0ne_ has quit IRC11:55
*** e0ne has joined #openstack-cinder11:57
openstackgerritwang peter proposed openstack/cinder: CG Modification Support in EMC VNX Cinder Driver
*** markvoelker has joined #openstack-cinder12:08
*** smcginni1 is now known as smcginnis12:08
*** markvoelker has quit IRC12:12
*** smoriya has quit IRC12:25
*** aix has joined #openstack-cinder12:32
*** e0ne is now known as e0ne_12:35
*** IanGovett has joined #openstack-cinder12:36
*** kbyrne has joined #openstack-cinder12:37
*** e0ne_ is now known as e0ne12:39
openstackgerritYuriy Nesenenko proposed openstack/cinder: Fix comments style according to the Hacking Rules
*** annashen has joined #openstack-cinder12:49
*** humble_ has joined #openstack-cinder12:51
*** annashen has quit IRC12:54
*** takedakn has joined #openstack-cinder12:56
*** takedakn has quit IRC12:56
openstackgerritEdwin Wang proposed openstack/cinder: Add iSCSI protocol support for IBM FlashSystem
openstackgerritEdwin Wang proposed openstack/cinder: Add iSCSI protocol support for IBM FlashSystem
openstackgerritEdwin Wang proposed openstack/cinder: Add iSCSI protocol support for IBM FlashSystem
*** e0ne is now known as e0ne_13:04
*** takedakn has joined #openstack-cinder13:05
*** takedakn has quit IRC13:07
*** markvoelker has joined #openstack-cinder13:09
*** deepakcs has quit IRC13:14
*** markvoelker has quit IRC13:14
*** e0ne_ has quit IRC13:14
*** e0ne has joined #openstack-cinder13:16
*** ankit_ag has quit IRC13:19
*** EmilienM is now known as EmilienM|afk13:20
*** coolsvap- is now known as coolsvap13:27
*** Longgeek has quit IRC13:32
openstackgerritPh. Marek proposed openstack/cinder: DRBD for NOVA.
*** rushiagr_away is now known as rushiagr13:34
*** jcru has joined #openstack-cinder13:38
openstackgerritRonen Mesonzhnik proposed openstack/cinder: Provide backup service that uses ProtecTIER as the backend
*** e0ne is now known as e0ne_13:40
*** sgotliv has quit IRC13:44
*** Mandell has joined #openstack-cinder13:44
*** e0ne_ has quit IRC13:45
*** sgotliv has joined #openstack-cinder13:47
*** sgotliv has quit IRC13:47
*** sgotliv has joined #openstack-cinder13:47
*** Mandell has quit IRC13:49
*** e0ne has joined #openstack-cinder13:50
*** e0ne is now known as e0ne_13:51
*** e0ne_ is now known as e0ne13:51
*** e0ne is now known as e0ne_13:51
*** e0ne_ is now known as e0ne13:52
*** e0ne is now known as e0ne_13:52
*** sandywalsh_ has quit IRC13:52
*** e0ne_ is now known as e0ne13:52
openstackgerritMichal Dulko proposed openstack/cinder: Delay requests in scheduler to wait for metrics
*** e0ne is now known as e0ne_13:53
*** e0ne_ is now known as e0ne13:53
duleke0ne: Hi13:54
duleke0ne: You may want to take a look at my attempt to fix
openstackLaunchpad bug 1409012 in Cinder "Volume becomes in 'error' state after scheduler starts" [High,In progress] - Assigned to Michal Dulko (michal-dulko-f)13:54
e0nedulek: hi13:54
e0nedulek: sure, i'll do! thanks13:55
duleke0ne: I hope this idea doesn't share your problems with instability.13:55
dulekyour idea's problems ;)13:55
*** nlevinki has quit IRC13:57
*** boichev2 has joined #openstack-cinder13:57
*** anshul has quit IRC13:57
*** boichev has quit IRC13:59
*** Guest59076 is now known as dansmith14:05
*** dansmith is now known as Guest2845614:06
openstackgerritMichal Dulko proposed openstack/cinder: Delay requests in scheduler to wait for metrics
*** Guest28456 is now known as dansmith_14:09
*** markvoelker has joined #openstack-cinder14:10
*** markvoelker has quit IRC14:15
*** kaufer has joined #openstack-cinder14:18
*** Murugan_ has quit IRC14:20
*** vnagpal has quit IRC14:21
*** afazekas_ has quit IRC14:23
*** markvoelker has joined #openstack-cinder14:28
*** scottda_ has joined #openstack-cinder14:28
*** mriedem has joined #openstack-cinder14:31
*** Longgeek has joined #openstack-cinder14:33
*** markvoelker has quit IRC14:34
dulekDuncanT: What are the symptoms that Jenkins isn't fine with review 156219?14:35
*** scottda_ has quit IRC14:37
DuncanTdulek: I'm sorry, I entirely misread my screen there, it is the IBM CI system that isn't happy, jenkins hasn't got to it yet. Sorry for the false alarm14:38
*** primechuck has joined #openstack-cinder14:39
dulekDuncanT: np, thanks for clarifying, I've thought that there's some tool to check that I'm not aware of. :)14:39
*** Longgeek has quit IRC14:40
*** afazekas has joined #openstack-cinder14:40
*** Longgeek has joined #openstack-cinder14:40
DuncanTdulek: Nope, just me being blind. Thanks for looking at this, it is a nasty issue and we're unlikely to get a clean fix through any of the necessary hooks with oslo in time for Kilo14:41
dulekDuncanT: I've commented on the patchset, but I we can discuss here14:41
dulekIs it considered to make volume save metrics to the DB?14:41
dulekThis is how Nova is tackling this problem14:42
DuncanTI think we used to do that but we were trying to reduce the load on the DB (nova folks have also commented many times that the db approach has issues, the 'no db scheduler' proposal has floated and sunk many, many times)14:43
DuncanTI /think/ the right solution is for the scheduler to listen for incoming stats messages from drivers, but not for work, for the first thirty seconds that it is up14:44
dulekHm, compromise would be to save metrics but read it only on the start and then rely on the RPC casts14:44
dulekfor updates14:44
DuncanTThe volume-managers would still have to write to the db every update, since they don't know when scheduler(s) are starting14:45
dulekDuncanT: True14:45
DuncanTAnd it seems rather ugly to my mind to have two communications methods14:45
dulekDuncanT: You're right, when you have DB load on your mind then RPC solution seems to be the best14:46
DuncanTI suppose if it went in its own table, we can always back out the db scheduler cleanly if needed. The data coming back from the cinder-volume service is getting very large though, with pools, schema, help strings etc being in there - 10s of k is looking likely14:47
DuncanTI've been worrying for a while if MQ load isn't a problem too :-)14:47
DuncanTNeed to run some benchmarks though, not just guess, and I haven't time for that at the moment14:48
dulekI guess MQ is easier to scale than DB.14:48
DuncanTWe're trying to get a release out of the door this week, so I'm rather distracted by that TBH14:48
dulekWhat do you mean by out of the door?14:49
DuncanTWell, you can shard the db easily enough if you think hard about it, just as you can MQs, so they scale pretty much horizontally either way, though neither is easy to setup/manage14:49
DuncanTHP have a Helion version release nominally scheduled for this Friday14:50
*** annashen has joined #openstack-cinder14:50
*** jungleboyj has quit IRC14:50
DuncanTBy 'out the door' I mean we're in the final stages of the release process14:50
dulekAh, okay, thought of Cinder and that confused me14:51
*** tsekiyama has joined #openstack-cinder14:51
*** Ilja1 has quit IRC14:52
*** Ilja has joined #openstack-cinder14:52
DuncanTNo, if cinder was my only responsibility I'd be happier and more productive :-)14:53
*** annashen has quit IRC14:55
*** scottda_ has joined #openstack-cinder14:59
*** tsekiyam_ has joined #openstack-cinder15:00
*** sigmavirus24_awa is now known as sigmavirus2415:01
*** humble_ has quit IRC15:03
*** tsekiyama has quit IRC15:04
*** eharney has joined #openstack-cinder15:04
*** dansmith_ is now known as dansmith15:05
*** scottda_ has quit IRC15:05
*** scottda_ has joined #openstack-cinder15:10
*** patrickeast has joined #openstack-cinder15:11
openstackgerritMarc Koderer proposed openstack/cinder: Adapt volume manager to new driver structure
openstackgerritMarc Koderer proposed openstack/cinder: Port rbd driver to use new driver model
*** beekneemech is now known as bnemec15:13
*** scottda_ has quit IRC15:14
*** EmilienM|afk is now known as EmilienM15:16
*** thangp has joined #openstack-cinder15:16
*** scottda_ has joined #openstack-cinder15:19
openstackgerritRonen Mesonzhnik proposed openstack/cinder: Provide backup service that uses ProtecTIER as the backend
*** Ilja has quit IRC15:27
*** e0ne is now known as e0ne_15:28
*** thangp has quit IRC15:29
openstackgerritPatrick East proposed openstack/cinder-specs: Add DB table for driver private data
*** ebalduf has joined #openstack-cinder15:30
*** markvoelker has joined #openstack-cinder15:31
*** xyang1 has joined #openstack-cinder15:34
*** markvoelker has quit IRC15:35
*** patrickeast has quit IRC15:35
*** patrickeast_ has joined #openstack-cinder15:35
*** e0ne_ has quit IRC15:38
*** jungleboyj has joined #openstack-cinder15:39
openstackgerritJeegn Chen proposed openstack/cinder: Pool-aware scheduler support in EMC VNX Cinder driver
*** tsekiyam_ has quit IRC15:46
*** sandywalsh has joined #openstack-cinder15:47
*** haomaiwang has joined #openstack-cinder15:47
*** tsekiyama has joined #openstack-cinder15:50
*** nellysmitt has quit IRC15:51
*** Mandell has joined #openstack-cinder15:52
*** markvoelker has joined #openstack-cinder15:53
*** mtanino has joined #openstack-cinder15:53
*** rmesta has joined #openstack-cinder15:54
*** jistr has joined #openstack-cinder15:58
*** markvoelker has quit IRC15:58
*** david-lyle_afk is now known as david-lyle16:01
*** ronis has quit IRC16:01
*** scottda_ has quit IRC16:01
*** thangp has joined #openstack-cinder16:02
jgriffithhemna: gary-smith around?16:05
*** anshul has joined #openstack-cinder16:07
*** nkrinner has quit IRC16:08
*** pradipta has quit IRC16:09
*** hodos has joined #openstack-cinder16:14
openstackgerritTomoki Sekiyama proposed openstack/cinder: Limit volume copy bandwidth per backend
openstackgerritTomoki Sekiyama proposed openstack/cinder: Limit volume copy bandwidth per backend
*** xyang has joined #openstack-cinder16:16
*** scottda_ has joined #openstack-cinder16:18
*** scottda_ has quit IRC16:20
openstackgerritMikhail Khodos proposed openstack/cinder: Fix Nexenta NFS driver mounts
patrickeast_jgriffith: ping16:24
*** zigo has quit IRC16:26
jgriffithpatrickeast_: hey16:27
*** annashen has joined #openstack-cinder16:27
patrickeast_jgriffith: hey, just wanted to follow up on the replication thing about replicating to hosts outside of cinder16:28
jgriffithpatrickeast_: yeah..16:28
patrickeast_i think what you have works well, i actually lean towards not allowing that and making sure this doesn’t allow for it16:28
jgriffithpatrickeast_: so when you say outside of Cinder to me that implies you won't fail over to it either right?16:28
patrickeast_the current replication does and its a pain16:28
patrickeast_our driver then essentially has to manage multiple arrays16:28
patrickeast_no it does fail over16:29
jgriffithpatrickeast_: so it fails over in the driver only?16:29
jgriffithpatrickeast_: yeah, not a fan of that16:29
patrickeast_but the initialize_connection and all that needs to know which array to connect to any point after16:29
patrickeast_yea its gross16:29
patrickeast_caused a lot of issues (not sure if you saw the review our team put up)16:29
jgriffithpatrickeast_: FWIW my first iteration did that16:29
jgriffithpatrickeast_: I don't know that I did...16:30
jgriffithpatrickeast_: so my take on it would be Cinder needs to know about the device to do things properly16:30
jgriffithpatrickeast_: the "no_schedule" flag I think is a good thing to have16:30
*** delattec has joined #openstack-cinder16:30
patrickeast_jgriffith: ok cool, we’re on the same page then ;)16:31
jgriffithbut no knowledge of the target makes things ugly16:31
*** zigo has joined #openstack-cinder16:31
jgriffithpatrickeast_: awesome16:31
jgriffithpatrickeast_: and FWIW, my main thing about that is if you want to implement a replication target all in your driver with Cinder not having any real knowledge...16:31
jgriffithThat's fine, but you can do that today with extra-spec keys16:31
patrickeast_yea exactly16:32
*** jistr has quit IRC16:32
jgriffithI have a customer doing that with a "custom" driver on Icehouse today16:32
jgriffithno changes to Cinder at all16:32
jgriffithOk... cool16:32
*** delattec has quit IRC16:33
*** delatte has quit IRC16:33
patrickeast_jgriffith: so one other thing, not sure i saw it in the spec, are you thinking that the two drivers for the replicated backends will know about eachother? like for ours we need to know about their management ip’s and such16:33
patrickeast_i don’t think anywhere else in cinder a driver actually has access to values from another one16:34
jgriffithpatrickeast_: so that's the nice thing about having it in the conf and putting some sort of pairing together16:34
jgriffithpatrickeast_: the driver could use the conf file to obtain that info if it wants16:34
patrickeast_ooo right, gotcha16:34
jgriffithpatrickeast_: or ignore it if it doesn't care16:34
jgriffithpatrickeast_: although I don't know how it would ignore it16:34
jgriffithpatrickeast_: but regardless, it accomodates both scenarios16:34
jgriffithpatrickeast_: awesome... I'll get some more feedback from folks and update that spec later today16:35
jgriffithI'd like to make sure hemna, smcginnis, xyang, and jungleboyj inparticular have a good look at it16:36
xyangjgriffith: So Cinder will know only 1 volume, even though there are primary, secondary hosts16:36
jgriffithxyang: right16:36
jgriffithxyang: so rather than have multiple volumes we have multiple hosts for a volume16:37
jgriffithxyang: and that's what we change around instead of "volume swapping" magic16:37
xyangjgriffith: ok, I think that is easier to manage16:37
jgriffithxyang: yeah, I think it's a lot easier to manage16:37
patrickeast_^ i think that better represents what actually happens (at least for my array)16:37
*** e0ne has joined #openstack-cinder16:38
jgriffithxyang: my thoughts are that it also makes it more flexible and less intrusive to the Cinder code16:38
xyangjgriffith: we'll need to store primary lun id and secondar lun id somewhere and swap it when we do promote16:38
jgriffithand at the same time has solid structure around it16:38
jgriffithxyang: my thought was when you promote it just sends the initialize_con call to the "other" backend16:38
jgriffithxyang: so we avoid all of that tracking and swapping business altogether16:39
patrickeast_jgriffith: oh that reminds me, the plan is to still only support failover once a volume is detached and then requires a manual attach call, right?16:39
jgriffithpatrickeast_: Yes, I think it's more accurate for the replication use-case in general16:39
xyangjgriffith: will both primary and secondary hosts be active?16:39
*** Mandell has quit IRC16:39
*** sgotliv has quit IRC16:40
jgriffithxyang: depends on what you mean by that...  but I'd say "no"16:40
jgriffithxyang: it's an active/passive HA16:40
xyangjgriffith: ok, that's what I refer to16:40
xyangjgriffith: as active/active HA doesn't work for cinder yet16:40
jgriffithpatrickeast_: so TBH I think so at least first pass.  Only because without things like muti-attach I haven't figured otu a clever way to make it work yet :)16:40
jgriffithpatrickeast_: and that goes back to active/passive16:40
jgriffithxyang: right, and there's no way to get that without multi-attach16:41
patrickeast_ok cool16:41
jgriffithxyang: the good thing is my proposal does take it into consideration16:41
xyangjgriffith: so basically one cinder volume host will be managing a local-remote pair underneath a volume uuid16:41
jgriffithxyang: so that at some point in the future it could be added16:41
jgriffithxyang: well... no16:41
jgriffithxyang: actually kinda the opposite16:42
jgriffithxyang: one volume in Cinder16:42
jgriffithxyang: that can be managed/service by two hosts (backend-drivers)16:42
jgriffithxyang: it flips the model around16:42
xyangjgriffith: yes, one volume, and you only have one entry in db for each volume16:43
jgriffithxyang: but honestly, it's up to your implementation on exactly what that looks like16:43
jgriffithxyang: yes, exactly16:43
jungleboyjjgriffith: Which spec is this that you are talking about?16:43
e0nehi all! i know that is not targeted for K, but spec is merged and i'll be happy if anybody will have a time to review it16:44
jungleboyjWhoa, didn't know you had started this.16:44
jungleboyjjgriffith: Let me read through this.16:44
jgriffithxyang: so just saw your etherpad regarding CG's16:44
jungleboyjjgriffith: Thanks for getting it started.16:45
jgriffithxyang: I'll have to think through that a bit...16:45
xyangjgriffith: sure16:45
jgriffithxyang: but, my first thought is it's just an workflow16:45
jgriffithxyang: in other words, it just means as you say that CG's have to be on replicated volumes16:45
xyangjgriffith: I think it should work together, just need think thru it16:45
jgriffithxyang: depending on the backend that could be made implicit if required16:45
jgriffithxyang: FWIW, if it's required by the back-end I like the idea of making it implicit16:46
jgriffithxyang: the fewer moving parts for the cloud-admin the better16:46
xyangjgriffith: so the promote is currently on volume, we need to add one for group if we integrate the two16:46
patrickeast_jgriffith: xyang: i can only speak to my knowledge of how we have groups, but can’t we just treat the cg’s the same way we do volumes in your spec? Just let the commands specify a volume or cg id?16:47
jgriffithpatrickeast_: yeah.. interesting.  So the CG stuff I haven't really looked at with this.16:47
jgriffithpatrickeast_: xyang in my case I don't do CG's across devices currently16:47
jgriffithso that was a blind-spot for me16:47
jgriffithnot that I can't, I just don't :)16:47
*** xyang1 has quit IRC16:47
xyangjgriffith: patrickeast_: I think the scheduler can take care of that if you have both cg_support and replication_support as true16:47
jgriffithxyang: +116:48
*** xyang1 has joined #openstack-cinder16:48
jgriffithxyang: patrickeast_ so... just to make sure I'm clear16:48
jgriffithxyang: patrickeast_ do your backends require replication for CG's?16:48
jgriffithxyang: patrickeast_ or is that an "option"16:48
xyangjgriffith: there's one problem we need to make sure.  if volume is in CG, then user can't operate on it individually, so we could have a replication_cg_support flag16:49
xyangif you want both16:49
jgriffithxyang: makes sense16:49
jgriffithxyang: so it's an "option" that can be setup16:49
jgriffithxyang: I think we could do that using the CG API's16:49
patrickeast_so for us replication is on a cg level, we can do a single volume cg for volume level replication or a whole cg, its treated the same way16:49
jgriffithpatrickeast_: right... but the question is "can you do a CG on a single backend"16:50
xyangjgriffith, patrickeast_, ya, pure is the first to put the two together.  It is interesting16:50
jgriffithpatrickeast_: I've always assumed that answer to be "yes"16:50
patrickeast_jgriffith: yea we do cg’s on a single backend16:50
jgriffithpatrickeast_: so given that, I think we're good still16:51
patrickeast_things get weird for us in the case of trying to replicate a single volume that belongs to a larger group16:51
patrickeast_since if you fail it over the group is split16:51
xyangjgriffith: we can do both group level and volume level replication16:51
jgriffithpatrickeast_: xyang Ok, perfect.  That was my assumption16:51
jgriffithpatrickeast_: xyang so I think this works fairly well still16:51
*** dulek has quit IRC16:52
jgriffithpatrickeast_: xyang we may impose some limits to start with depending on how things look16:52
jgriffithpatrickeast_: xyang but IMO I'd rather do that and iterate on the implementation than try and solve every use case up front16:52
patrickeast_jgriffith: that sounds good to me, id rather have it be a bit more restrictive than have a weird support matrix of what works on each backend16:53
jgriffithpatrickeast_: xyang but I think this is def doable16:53
jgriffithpatrickeast_: +1 on the weird support matrix16:53
jgriffithpatrickeast_: that only ends badly16:53
xyangjgriffith: sure16:53
jgriffithxyang: patrickeast_ I'll get some stuff cleaned in the spec barring any huge concerns from HP, IBM or others16:54
xyangsounds good16:54
jgriffithand then start working on some prototypes that I can share with everyone16:54
jgriffithwe can all collaborate on this as it goes along is my thought16:54
*** markvoelker has joined #openstack-cinder16:54
jgriffithbeauty of github16:54
jgriffithand TBF the initial implementation was done the same way16:55
jgriffithbut I think many of us were too distracted with other things to give it as much attention as it deserved16:55
jgriffithespecially me :(16:55
*** e0ne is now known as e0ne_16:56
patrickeast_jgriffith: speaking of other things to distract from this… have you had a chance to look through my spec for the initiator db table meta data thingy?16:57
jgriffithI have as a matter of fact16:58
jgriffithpatrickeast_: for the most part I think it looks really sane16:58
jgriffithpatrickeast_: one question I had for you... that may be in the spec and I just missed it16:58
*** nileshb has joined #openstack-cinder16:58
jgriffithpatrickeast_: how does the driver access the info?16:59
*** markvoelker has quit IRC16:59
jgriffithpatrickeast_: ie, how does a driver know that an entry in that table belongs to it?16:59
*** annashen has quit IRC16:59
jgriffithpatrickeast_: FK?16:59
patrickeast_jgriffith: ah yea16:59
patrickeast_jgriffith: so that part is a little vauge16:59
patrickeast_jgriffith: my thought was the manager looks up data for the initiator and passes it to the driver, at which point the driver can decide what data it cares about17:00
patrickeast_the reasoning being that it would be much more restrictive/hard to have the manger/db query make that decision17:00
*** e0ne_ is now known as e0ne17:00
patrickeast_there is of course a performance hit17:01
patrickeast_if there is a ton of junk saved for some initiator17:01
patrickeast_but i’m not sure if thats really too big of a deal17:01
jgriffithpatrickeast_: so not to be a PITA but I'd like to avoid that if we could17:01
jgriffithpatrickeast_: let me look at something... just a sec17:01
patrickeast_i’m all for that, but i couldn’t think of a better solution that still worked for HA down the road17:02
jgriffithpatrickeast_: hmm... that's tricky17:03
jgriffithpatrickeast_: the only place that it would work is an FK in the service table17:03
jgriffithpatrickeast_: I think17:03
jgriffithpatrickeast_: as far as the HA piece, you can have the same info for both hosts no?17:03
patrickeast_jgriffith: hehe yea, but this goes back down the road of both drivers needing to be able to get the same value17:04
jgriffithpatrickeast_: oh... understood17:04
jgriffithpatrickeast_: but in HA I'd say just like replication...17:05
jgriffithpatrickeast_: create some association between the two that the driver knows about17:05
jgriffithpatrickeast_: build on that foundation17:05
jgriffithpatrickeast_: so iterating through items in a DB return has bit us in the past17:05
jgriffithpatrickeast_: it's just really inefficient IMO and it's arbitrary17:06
patrickeast_jgriffith: yea, I agree17:06
jgriffithpatrickeast_: I like explicit when possible17:06
jgriffithpatrickeast_: so I think the FK in the Service table might be a good way to do this17:06
patrickeast_jgriffith: i kind of like the idea of the driver being able to specify a query parameter, almost like a vendor prefix17:06
patrickeast_jgriffith: so like you said they can be linked through the config files17:06
jgriffithpatrickeast_: and when we do get to HA we can easily put something in the conf file to let two backends know they're linked17:07
jgriffithpatrickeast_: yeah... that sounds good to me17:07
patrickeast_jgriffith: its a bit more free-form then the service name17:07
jgriffithpatrickeast_: honestly, the way that shapes up it goes back to the possibility of just "custom" KV pairs for driver usage being possible17:07
jgriffithpatrickeast_: ohhh...17:08
jgriffithpatrickeast_: interesting17:08
jgriffithpatrickeast_: so you're saying use a label in the conf file17:08
jgriffithpatrickeast_: skip the FK altogether and query on that label17:08
patrickeast_jgriffith: yep17:08
jgriffithpatrickeast_: hmm... that's interesting17:08
jgriffithpatrickeast_: I kinda like that17:09
jgriffithpatrickeast_: and default for that label could be backend_name17:09
jgriffithpatrickeast_: haha17:09
patrickeast_jgriffith: yea, thats even where i started on that path17:09
jgriffithpatrickeast_: actually, if you do that you're covered17:09
jgriffithpatrickeast_: if it's backend name you're going to have the same entry no matter how many backends you have17:10
jgriffithpatrickeast_: in other words... second Pure backend comes online, checks the table and says "hey, I've already got the initiator info, move along"17:10
jgriffithpatrickeast_: let's do that :)17:10
patrickeast_jgriffith: sounds good to me17:10
jgriffithpatrickeast_: awesome!17:10
patrickeast_jgriffith: ill update the spec with that17:11
jgriffithpatrickeast_: cool, thanks17:11
patrickeast_jgriffith: no no, thank you17:11
jgriffithpatrickeast_: the only other thing worth considering is going back to the "generic" data17:11
patrickeast_solving the hard parts for me ;)17:11
jgriffithpatrickeast_: but I don't want to throw that wrench into the works17:11
jgriffithpatrickeast_: it's likely to get all sorts of scrutiny :)17:11
patrickeast_jgriffith: isn’t the general concensous that generic data is bad?17:11
jgriffithpatrickeast_: and there's my point :)17:11
jgriffithpatrickeast_: just depends on who you ask17:12
jgriffithpatrickeast_: so go with what you have17:12
jgriffithpatrickeast_: I'm in a different camp on the generic data, BUT the problem is abuse17:12
jgriffithpatrickeast_: and no way to prevent it other than review which is WAY too subjective17:12
jgriffithpatrickeast_: update that spec and I'm happy to approve it17:13
patrickeast_jgriffith: yea i agree, its kind of a scary can of worms17:13
jgriffithcan o'worms, slippery slope, road to hell17:13
jgriffithpopular quotes that have come up on the topic17:13
patrickeast_haha yep17:14
patrickeast_jgriffith: i’ll post an update a little later this afternoon… going to go enjoy some sunshine while i can17:14
jgriffithhaha!  Rare moments in Seattle eh17:14
*** coolsvap is now known as coolsvap_17:14
*** jdurgin has joined #openstack-cinder17:21
*** _cjones_ has joined #openstack-cinder17:23
openstackgerritSean McGinnis proposed openstack/cinder: Fix logging guideline violations in volume/
*** _cjones_ has quit IRC17:24
*** _cjones_ has joined #openstack-cinder17:24
*** anshul has quit IRC17:25
*** pck has quit IRC17:25
*** scottda has quit IRC17:25
*** tellesnobrega has quit IRC17:25
*** number80 has quit IRC17:25
*** rmstar has quit IRC17:25
*** mkoderer has quit IRC17:25
*** mtreinish has quit IRC17:25
*** strictlyb has quit IRC17:25
*** Anticimex has quit IRC17:25
*** guitarzan has quit IRC17:25
*** gugl2 has quit IRC17:25
*** SamYaple has quit IRC17:25
*** anish has quit IRC17:25
*** SergeyLukjanov has quit IRC17:25
*** eikke has quit IRC17:25
*** gpocentek has quit IRC17:25
*** anshul has joined #openstack-cinder17:31
*** pck has joined #openstack-cinder17:31
*** scottda has joined #openstack-cinder17:31
*** number80 has joined #openstack-cinder17:31
*** tellesnobrega has joined #openstack-cinder17:31
*** rmstar has joined #openstack-cinder17:31
*** mkoderer has joined #openstack-cinder17:31
*** mtreinish has joined #openstack-cinder17:31
*** 64MABZ0GB has joined #openstack-cinder17:31
*** Anticimex has joined #openstack-cinder17:31
*** guitarzan has joined #openstack-cinder17:31
*** gugl2 has joined #openstack-cinder17:31
*** SamYaple has joined #openstack-cinder17:31
*** anish has joined #openstack-cinder17:31
*** SergeyLukjanov has joined #openstack-cinder17:31
*** eikke has joined #openstack-cinder17:31
*** gpocentek has joined #openstack-cinder17:31
*** ronis has joined #openstack-cinder17:35
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
*** harlowja_at_home has joined #openstack-cinder17:37
jungleboyjeharney: Ping17:38
eharneyjungleboyj: hi17:38
jungleboyjeharney: Hey, long time no chat.17:38
eharneyjungleboyj: indeed :)17:39
jungleboyjeharney: Have you done anything with RHEL 7 and sharing out volumes via iSCSI?17:39
eharneyjungleboyj: well i know a bit about LVM+LIO on it17:39
*** mkerrin has quit IRC17:39
*** nellysmitt has joined #openstack-cinder17:40
jungleboyjeharney: Ok, you are further than me.  I have a HyperV node that seems to be failing because I think CHAP is enabled, but I don't see anywhere that stuff like that is cmnfigured.17:40
eharneyjungleboyj: so when you create an export in the LVM driver, chap credentials are created for the LIO target17:41
eharneyjungleboyj: Cinder does this with the rtstool script in Cinder, as a user you'd look at these things w/ targetcli17:41
jungleboyjeharney: Ok, so that is happening by default.  That was what it looked like to me.17:41
jungleboyjeharney: Right.17:41
*** nileshb has quit IRC17:42
*** karimb has quit IRC17:42
eharneyjungleboyj: from the point of view of the compute node, it shouldn't be different from tgtd AFAIK17:43
jungleboyjWas tgt setting chap credentials by default?17:43
jungleboyjI don't know that it was.17:44
eharneywell, i thought...17:45
jgriffithjungleboyj: what version of cinder are you working with?17:45
jungleboyjeharney: Regardless, there isn't an option to disable setting up CHAP?17:45
eharneyjungleboyj: nope17:45
jungleboyjjgriffith: This is Kilo.17:46
jungleboyjLooks like a late January build though.17:46
jgriffithjungleboyj: put a debug in volume/targets/lio at the iscsi target create17:47
jgriffithjungleboyj: you can see the chap setting that way17:47
jgriffithjungleboyj: or instrument the code with a log statement17:47
jungleboyjjgriffith: Ok.17:47
jgriffithjungleboyj: you need to make sure you have the iqn's set up from initiator too17:47
jgriffithbut eharney knows way more than I do about LIO17:48
jungleboyjjgriffith: Well, from the debug I was getting it looked like CHAP was being set up.  Which explains why I am getting an error from the HyperV node.  It can't do CHAP.17:49
jgriffithjungleboyj: hehe17:49
jgriffithsorry... so your question was if you can disable use of chap17:50
jungleboyjjgriffith: Yes, I wasn't seeing an option for that.  Was trying to confirm.17:50
jungleboyjjgriffith: And you wouldn't laugh if you had to deal with HyperV.  :-)17:50
eharneyjungleboyj: you'll see in cinder/cmd/rtstool that it sets the "authentication" attribute to 1, so it's always required17:50
jgriffithjungleboyj: eharney we should clean things up a bit, as the conf file has an entry to set chap to None17:52
jgriffithCHAP username to use for iSCSI Targets17:52
*** briancline has quit IRC17:52
*** briancline has joined #openstack-cinder17:52
jgriffithadded for scst I think (at my request)17:53
jgriffithalthough just documenting some drivers always use chap is probably fine too17:54
*** markvoelker has joined #openstack-cinder17:56
eharneyjgriffith: hmm, seems like a good idea to clean things up for sure.  I think LIO is a little different in that it uses different CHAP creds for every target, too17:56
jgriffitheharney: yeah, there are other drivers that do that, but I think it's unique in the "volume/targets" space17:57
eharneyer, yeah, i meant per-volume actually17:57
jgriffitheharney: yeah, I'm with ya17:57
jgriffitheharney: my driver does the same thing... but that's completely different17:58
jgriffithanyway, just a potential clean up17:58
*** markvoelker has quit IRC18:01
jungleboyjeharney: jgriffith Thanks for the pointers.  Makes sense to have CHAP enabled by default.  HyperV is just limited there.18:04
*** e0ne has quit IRC18:05
*** yuriy_n17 has quit IRC18:07
*** EmilienM is now known as EmilienM|afk18:07
*** rushiagr is now known as rushiagr_away18:11
openstackgerritPatrick East proposed openstack/cinder-specs: Add DB table for driver private data
*** afazekas has quit IRC18:13
*** coolsvap_ is now known as coolsvap18:25
casusbellijgriffith: Hi! Do you have a min regarding sos-ci setup questions?18:29
casusbellijgriffith: (no is a valid answer. :) )18:30
openstackgerritThang Pham proposed openstack/cinder: Snapshot object
openstackgerritThang Pham proposed openstack/cinder: Cinder objects base
*** patrickeast_ has quit IRC18:34
*** MasterPiece has joined #openstack-cinder18:38
jgriffithcasusbelli: hey... sure18:39
*** jordanP has quit IRC18:40
*** ebalduf has quit IRC18:40
*** Mandell has joined #openstack-cinder18:45
*** ybathia has joined #openstack-cinder18:48
ybathiaHi, I have a small query.. Why are quotas expected to be in gigabytes in this file Can the quotas not be smaler or way bigger than gigabytes?18:51
jgriffithybathia: the resolution of a volume size is Gigabytes18:51
jgriffithybathia: so it sort of seems natural18:51
jgriffithybathia: so no smaller, but sure on bigger (104857600000 Gigabytes)18:52
jgriffithybathia: it's just the default unit of measure18:52
ybathiaOk..Got it. Submitted a patch in horizon and got asked this question. So, wanted to know the reason from the cinder guys..18:53
jgriffithybathia: sure18:54
*** e0ne has joined #openstack-cinder18:56
*** markvoelker has joined #openstack-cinder18:57
*** annashen has joined #openstack-cinder19:00
*** markvoelker has quit IRC19:02
casusbellijgriffith: ick, was afk, sry.19:04
casusbellijgriffith: I'm doing a sos-ci setup but i'm a bit confused about which OpenStack, especially neutron / networks setup it expects.19:05
*** annashen has quit IRC19:05
casusbellijgriffith: What setup is your default scenario. sos-ci running on a controller? In my setup the VMs require a floating ip to be contactable from the ci but that does not seem to be the case in your default setup...19:06
casusbellijgriffith: questions questions questions... :)19:07
*** Yogi1 has joined #openstack-cinder19:07
jgriffithcasusbelli: so look in templates19:08
jgriffithcasusbelli: that the local.conf file I send over and use for stack.sh19:08
jgriffithcasusbelli: it doesn't use neutron in mine19:08
jgriffithcasusbelli: but you could modify it to do so if you like of course19:08
jgriffithcasusbelli: you're correct, I don't require floating IP's either19:09
jgriffithcasusbelli: and as such I don't register those as a variable19:09
jgriffithcasusbelli: you can modify that pretty easily in the "boot_openstack_instance.yml" task19:09
casusbellijgriffith: yep, already tampered with the yml scripts19:10
jgriffithcasusbelli: either add a task to assing a floating ip, or set your nova.conf on the cloud you're running against to always assing floating IP and just parse itout19:10
jgriffithit out19:10
jgriffithcasusbelli: the networking info in the instance object is kind of atrocious though19:10
casusbellijgriffith: yep, that's why i wanted to ask. :-D19:10
jgriffithcasusbelli: you'll want to run the task by itself and dump the object to figure out how to parse it19:11
casusbellijgriffith: hmm, ok19:11
jgriffithcasusbelli: hehe.. yeah, it's somewhat madening19:11
casusbellijgriffith: How do you avoid this in your setup?19:11
jgriffithcasusbelli: I don't need floating IP's for anything19:11
*** Yogi1 has quit IRC19:11
*** anshul has quit IRC19:11
jgriffithcasusbelli: everything is run "on" the machine I have my cloud deployed on19:11
jgriffithcasusbelli: devstack inside of devstack19:11
jgriffithso to speak19:11
casusbellijgriffith: that's what i gathered. Are the private IPs of your VMs reachable from the sos-ci node?19:11
*** aix has quit IRC19:11
jgriffithso private IP's are all I need19:12
jgriffithcasusbelli: yeah, so the sos-ci node IS the devstack node that's running my CI19:12
casusbelliJgriffith: hmm, irc hickup19:12
jgriffithcasusbelli: I "used" to have it split out and running on a VM in my OpenStack cloud19:12
casusbellijgriffith: aaah, ok19:12
jgriffithcasusbelli: that would solve the problem for you as well19:12
jgriffithcasusbelli: ie instead of an sos-ci node, you have an sos-ci instance19:13
jgriffithcasusbelli: which honestly for a number of reasons was actually a "better" way to go19:13
jgriffithcasusbelli: but I changed it a while back while working on some things and never switched it back :)19:13
jgriffithcasusbelli: so what's nice about using instances other than the whole "IP access" thing19:14
jgriffithis you can have several of them doing different things19:14
casusbellijgriffith: hmm, ok. That give's me some food for thought...19:14
*** Yogi1 has joined #openstack-cinder19:14
jgriffithI had one running LVM, one with LVM thin and one with SolidFire19:15
jgriffithcasusbelli: I highly recommend the instance model FWIW19:15
jgriffithcasusbelli: plus you can can do things like "experimental" setups easily that way19:15
jgriffithrunning things in parallel19:15
jgriffithcasusbelli: not sure if that answered your questions or not19:16
casusbellijgriffith: when running a sos-ci node, where does it access the nova client? I tried that but the ansible nova module didn't work mit RDO based controller19:16
jgriffithcasusbelli: I should probably draw out the infrastructure I'm using in teh README19:16
casusbellijgriffith: THat's definitely a good idea19:16
jgriffithcasusbelli: it *should* because if you look I'm just specifying the rc file19:16
jgriffithcasusbelli: should work with RDO although I haven't tried it19:17
jgriffithcasusbelli: my setup is an all in one OpenStack deployment on a server19:17
casusbellijgriffith: i was surprised, too. It works when run against localhost but not when run from a VM instance vs. the controller node19:17
jgriffithcasusbelli: I set all of the openstack creds in the sos-ci code (keystone endpoint, login, user, tenant)19:17
jgriffithcasusbelli: OH!19:17
jgriffithcasusbelli: did you install the clients and everything?19:18
jgriffithcasusbelli: and are your endpoints accessible from the instance?19:18
casusbellijgriffith: on the vm? no19:18
casusbellibut on the controller? yep19:18
jgriffithcasusbelli: there's your problem19:18
jgriffithcasusbelli: it's just "openstack" at that point19:18
casusbellijgriffith: Well, normally the point of an ansible module is to be executable on the remote host, not on the local one. and i ran the ansible scripts with the controller as the remote hsot19:19
jgriffithcasusbelli: wherever sos-ci is running, you have to have the clients installed and be able to access with your openstack creds19:19
casusbellijgriffith: *host19:19
casusbellijgriffith: OH19:19
jgriffithcasusbelli: no, no... that's what I'm saying you don't *have* to be on localhost19:19
*** ebalduf has joined #openstack-cinder19:19
jgriffithcasusbelli: but you have to enable it with the tools you need to do it's thing19:19
jgriffithcasusbelli: and I'd argue about the localhost part19:19
casusbellijgriffith: ok, i see that now19:20
jgriffithcasusbelli: so localhost is great for doing things like calling "nova boot --image-id xxxx....."19:20
jgriffithcasusbelli: localhost is where your python-novaclient is etc19:20
jgriffithcasusbelli: remote tasks are installing devstack on the resultant instance, configuring/running tempest and gathering results19:21
casusbellijgriffith: ok, i'll review the whole setup with this info.19:21
casusbellijgriffith: Thanks for the help!!19:21
jgriffithcasusbelli: np19:21
jgriffithlemme know how it goes19:21
casusbellijgriffith: will do!19:21
jgriffithcasusbelli: ya know.. maybe I'll write an ansible task to "deploy" and setup an sos-ci machine19:22
jgriffithcasusbelli: that way it will do all of this "for you"19:23
jgriffithlunchtime.. bbl19:23
casusbellijgriffith: i'm working on that, actually19:23
casusbellijgriffith: bon appetite19:23
*** coolsvap is now known as coolsvap_19:31
*** lpetrut has quit IRC19:41
*** kaufer has quit IRC19:48
*** xyang has quit IRC19:48
openstackgerritxing-yang proposed openstack/cinder: Kilo Consistency Group API update
*** MasterPiece has quit IRC19:49
*** Lee1092 has quit IRC19:50
*** Mandell has quit IRC19:53
*** markvoelker has joined #openstack-cinder19:58
*** diemt has joined #openstack-cinder19:58
*** kaufer has joined #openstack-cinder19:59
*** markvoelker has quit IRC20:03
*** EmilienM|afk is now known as EmilienM20:04
e0nejgriffith: hi! one small question about API and cinder client from me20:06
e0neany thouts why do we have 'detatiled=True' by default in client code:
openstackgerritMitsuhiro Tanino proposed openstack/cinder: Clear migration_status from a destination volume if migration fails
guitarzane0ne: probably because the non detail view is very boring20:12
guitarzane0ne: and you'd just want to make a detail call after that anyway20:12
e0neguitarzan: in such case, having 2 api methods don't make sence20:13
guitarzanoh, I'm not going to get into any arguments about whether "show" is useful :)20:13
guitarzanor index I suppose20:14
e0neguitarzan: my question is related not only for 'volume get' method20:14
*** _cjones_ has quit IRC20:15
guitarzanhmm, create returns the summary view as well?20:16
guitarzannah, I must be reading something wrong20:17
*** xyang has joined #openstack-cinder20:19
openstackgerritAnish Bhatt proposed openstack/cinder: Add support for chiscsi iscsi helper
*** lpetrut has joined #openstack-cinder20:22
*** Mandell has joined #openstack-cinder20:23
anishjgriffith: e0ne slight change in code, can I get the +2's back please :)20:23
*** patrickeast has joined #openstack-cinder20:24
e0neanish: lets wait for Jenkins20:24
anishyes, o'course20:26
jgriffithanish: sorry... that was my last one20:29
jgriffithanish: e0ne think this is good to go no?20:29
jgriffithminor changes, was pretty much done prior to them20:30
anishyes, it reused more code that mtanino added instead of duplicating and moved a debug line around20:30
mtaninoanish: Thanks!20:30
jgriffithyeah, I'll +2/A after jenkins passes20:31
e0nejgriffith: +2 from me20:31
*** harlowja_at_home has quit IRC20:31
e0nejgriffith: did you see my question about client and api above?20:32
jgriffithe0ne: no... sorry, just got back from lunch20:32
* jgriffith scrolls back20:32
e0nejgriffith: np. i'm confused a bit how we use our api20:33
jgriffithe0ne: hmmm...20:34
jgriffithe0ne: well that's kinda lame20:34
e0nejgriffith: i hope, i've missed something. looks like we've got only 'detailed views'20:34
jgriffithe0ne: yeah, it's a pointless option it appears20:36
jgriffithoh... wait20:36
jgriffithlooking at the wrong thing20:36
jgriffithe0ne: nope, I wasn't20:36
jgriffithe0ne: it's always True20:36
jgriffithe0ne: suppose you could set it if you were using something other than the client20:37
e0ne"grep -r 'detailed=False' . | wc -l" - 6 lines found20:38
jgriffithie curl20:38
jgriffithe0ne: yeah, but there's no mechanism in the client to set it that I can see20:38
guitarzananyone sane will use their own client ;)20:38
jgriffithguitarzan: ouch20:38
jgriffithguitarzan: so you're saying our client is no good?20:39
guitarzancan't paste in the curl debugged commands anymore :(20:39
e0nejgriffith: i want to added in scope of
guitarzanjgriffith: nah, just griping on the keystoneclient "security" features that keep getting added20:39
jgriffithguitarzan: ahhh20:39
e0nejgriffith: but not sure that it is good K20:39
guitarzanthose folks obviously never have to use openstack20:39
jgriffithguitarzan: there's a lot of that going on in OpenStack these days IMO20:39
jgriffithguitarzan: including in Cinder :(20:39
guitarzanI was trying to see the /index vs /detail but can't just use cinder --debug to try it20:40
e0nejgriffith: thanks for clarification it. i'm a sad a bit: we've got a lot of unused code in client:(20:40
jgriffithe0ne: so that spec you have I'm in favor of, think you and I discussed this before20:41
jgriffithe0ne: only difference is I said show, you say detailed20:41
jgriffithe0ne: which if we make detailed actually "be" something is good with me20:41
e0nejgriffith: yes. but for now we use only 'detailed' views.20:42
jgriffithe0ne: right20:42
e0nejgriffith: so changing it requires changes both in server and client APIs20:42
jgriffithe0ne: yeah, as awful as this sounds... maybe we change the default, expose detailed and move on?20:44
e0nejgriffith: sounds reasonable to leave current behaviour as default and move everything else to detailed20:45
*** dulek has joined #openstack-cinder20:45
jgriffithe0ne: yeah, that sounds good since it's what we expose anyway20:47
jgriffithe0ne: the only catch I tought of is people like guitarzan that have their own clients20:47
jgriffithe0ne: where detailed can actually be used for something ;)20:47
*** Longgeek has quit IRC20:48
anishhuh, something up with the third party CIs ?20:50
anishnormally I'm flooded within the first 20 min20:50
anishtoday, nothing20:50
guitarzanjgriffith: nah, don't worry about me, I don't really use my own client for much :)20:51
jgriffithguitarzan: lol... you just use Jedi Mind Trick for Cinder calls :)20:52
openstackgerritPatrick East proposed openstack/cinder-specs: Add DB table for driver private data
guitarzanjgriffith: my "client" for that is called requests :)20:52
guitarzanalso, TIL that the volume manager exits if there are no periodic tasks20:53
guitarzanI didn't really expect that20:53
guitarzanjgriffith: doing the "maintenance mode" signal handler thing is really easy20:53
guitarzanjgriffith: but I can't cancel the periodic tasks or the launcher keeps restarting the service :)20:54
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
e0nejgriffith, guitarzan: "3rd party clients" are always risky until API freezed20:55
*** patrickeast has quit IRC20:55
sigmavirus24guitarzan: I'm glad to hear someone likes requests =P20:56
jgriffithguitarzan: you might be on to something :)20:56
*** _cjones_ has joined #openstack-cinder20:56
*** dulek has quit IRC20:56
guitarzanjgriffith: I'll have a simple POC diff for you and DuncanT to look at :)20:56
*** patrickeast has joined #openstack-cinder20:56
guitarzanI'm curious if he thinks I'm crazy with this idea20:56
openstackgerritAnish Bhatt proposed openstack/cinder: Add support for chiscsi iscsi helper
jgriffithguitarzan: I'll be curious to see it.  He'll be a better person to ask if it's crazy or not20:57
*** hodos has quit IRC20:58
jgriffithanish: my +2/A is now withdrawn since you added copyright20:58
*** markvoelker has joined #openstack-cinder20:59
anishthis is the part where I really wish I knew if you were joking :)20:59
jgriffithanish: don't worry, I was joking :)21:01
anishfwiw, I have zero idea what copyright + reserved + apache even implies21:01
*** annashen has joined #openstack-cinder21:01
* anish is following directions from above21:01
guitarzananish: "looks like all the other files" == good :)21:01
*** devlaps has joined #openstack-cinder21:01
anishguitarzan: I'm pretty sure that's what "above" went with as well21:02
*** patrickeast has quit IRC21:02
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
*** markvoelker has quit IRC21:04
*** annashen has quit IRC21:06
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
e0nejgriffith: is any change to get spec ^^ approved for K?21:13
anishhaha. you should see nova right now, it's a stream of "spec not approved for K"21:14
anishand here..21:14
jgriffithe0ne: you're going to have to talk to thingee and other cores on that21:16
jgriffithe0ne: IMO I'd be ok with it21:16
jgriffithe0ne: because it's more of a "fix" and "improvement" on something we have currently that isn't quite right21:16
e0nethingee, hemna, jungleboyj, winston-d, xyang: could you please take a look on
e0nejgriffith: agree. we spent a lot of time with it while fixint a 'bug' instead of implemented it like a bp21:18
*** xyang has quit IRC21:18
*** xyang has joined #openstack-cinder21:19
openstackgerritxing-yang proposed openstack/python-cinderclient: Kilo Consistency Group CLI update
jgriffithe0ne: minor point on that spec if you don't mind21:21
*** _cjones_ has quit IRC21:22
jgriffithe0ne: I'd like it if we were very explicity about what happens to the existing views vs the new views etc21:22
jgriffithe0ne: make sense?21:22
*** jwcroppe has joined #openstack-cinder21:22
e0nejgriffith: absolutely, i'll add it21:23
*** _cjones_ has joined #openstack-cinder21:23
*** xyang has quit IRC21:23
jgriffithe0ne: great!21:24
jgriffithe0ne: thanks!21:24
*** xyang has joined #openstack-cinder21:31
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
*** annashen has joined #openstack-cinder21:36
e0nejgriffith: thanks for very detailed and reasonable feedback21:37
*** lpetrut has quit IRC21:38
*** jungleboyj has quit IRC21:42
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
jgriffithe0ne: for sure, thanks for the great spec21:46
*** xyang has quit IRC21:47
*** Mandell has quit IRC21:47
e0nejgriffith: to be honest: it's only summarizing of irc and gerrit discussions21:48
*** xyang has joined #openstack-cinder21:50
*** annashen has quit IRC21:52
jgriffithe0ne: IMO those are the best specs21:53
jgriffithe0ne: and somebody following through and taking on the work is the hard part21:53
jgriffithe0ne: especially when it comes to writing specs, that's not *easy* IMHO21:53
e0nejgriffith: imo, it means that we had a good community work21:54
e0nejgriffith: and maybe it is because nobody likes to write specs:)21:54
jgriffithe0ne: haha... agree on both points21:55
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
kaufere0ne: I'm reading your spec, time for a question on it?21:59
e0nekaufer: sure. but i'll be available about 15-20min more today22:00
*** zzzeek has joined #openstack-cinder22:00
kauferDid you see my question in the Alternatives section about adding a --minimal param?22:00
kauferI'm just curious if you considered that22:00
openstackgerritJon Bernard proposed openstack/cinder: Add support for generic volume migration
zzzeekjbernard / eharney : I’m running cinder-volume from a fresh devstack, and this is my current console output:  it doesn’t fail at all across mulitple runs, can you give me a clue what dependencies are needed to cause the failure seen in   OS, plugins, cinder.conf, existing state ?22:01
openstackLaunchpad bug 1417018 in Cinder "Cinder encounters dbapi error: NoSuchColumnError: "Could not locate column in row for column '%(140070887032400 anon)s.volumes_id'"" [Undecided,New]22:01
*** markvoelker has joined #openstack-cinder22:01
*** Mandell has joined #openstack-cinder22:02
*** aix has joined #openstack-cinder22:02
e0nekaufer: i'm sorry, i missed  it22:02
*** jamielennox is now known as jamielennox|away22:03
*** primechuck has quit IRC22:03
e0nekaufer: i didn't added anything about '--minimal' to spec because not sure that is is needed right now22:04
e0nekaufer: i mean, i haven't got real use-cases where it is needed22:04
kaufere0ne: Is the goal of the spec to execute non-detailed queries when the caller knows that they do not need all of the information?22:05
*** markvoelker has quit IRC22:05
e0nekaufer: yes22:06
e0nekaufer: and i don't want to break api compatibility with a current master22:06
kaufere0ne: So are you proposed that the user would need to add --detailed when they needed all of that information?  And if it did not supply it then it would not make a detailed query?22:07
e0nekaufer: w/o --detailed flag user will got the same data as it is available in the current implementation22:08
e0nekaufer: for now, i'm planning to add only snapshots info in a detailed view22:08
*** thangp has quit IRC22:08
*** Longgeek has joined #openstack-cinder22:08
kaufere0ne: So then on the server the non-detailed view handler would become the detailed view handler? And then detailed view handler would add in the 'child_snapshots' field?22:10
e0nekaufer: detailed and default views are mixed now22:10
anishe0ne: that was fast. do you get notified when jenkins passes ?22:10
e0nekaufer: on the server-side, current detailed view will be moved to default and detailed will be extended with 'child_snapshots'22:12
jbernardeharney: did you hit the same issue, or was it just me? (nosuchcolumn error)22:12
kaufere0ne: By default view you mean the summary function here?
eharneyjbernard: i haven't hit it22:14
*** jamielennox|away is now known as jamielennox22:14
*** kaufer has quit IRC22:15
e0nee0ne: yes, it isn't used via cinderclient now and should be safe22:15
jbernardzzzeek: maybe it's just my hacked up tree… i can give you my local.conf, a patch to apply, and an operation to perform and you should see it22:16
e0nekaufer: ^^22:16
zzzeekjbernard: ahha.  OK, please do :)22:16
*** kaufer has joined #openstack-cinder22:16
*** primechuck has joined #openstack-cinder22:16
e0nekaufer: also there is no reference to it in a api docs:
zzzeekjbernard: the reporter of the issue was someone other than yourself, correct?  so more than one person is getting it22:16
e0nekaufer: (coped message) yes, it isn't used via cinderclient now and should be safe22:17
*** diemt has quit IRC22:17
kaufere0ne: Sorry, I have a bad connection, what isn't used via cinderclient?22:17
jbernardzzzeek: yes, but im not completely sure they are the same issue22:18
e0nekaufer: "the summary function" that you referenced22:18
jbernardzzzeek: this local.conf:
zzzeekjbernard: OK.   yes your issue is something different symptom wise22:18
jbernardzzzeek: this patch:
zzzeekjbernard: undo that patch and the error goes away ?22:19
jbernardzzzeek: i haven't had time to bisect or try different commit yet22:19
zzzeekjbernard: OK :)22:19
kaufere0ne: The server will use the view handler to format any non-detailed API query, which is what the client would invoke22:19
*** ronis has quit IRC22:19
zzzeekjbernard: if i could have a clueless moment with devstack, for this local.conf do I have to run again ?22:20
jbernardzzzeek: it doens't appear to happen in the gate for that patch, so it could very well be a problem with my setup22:20
jbernardzzzeek: yep, you'll have to rstack22:20
e0nekaufer: didn't get yout point in the last message22:20
zzzeekjbernard: OK, but you are hitting one of those creepy mysql driver issues, woudl be good to know22:20
kaufere0ne: The cinderclient flow uses that function, that's all22:21
jbernardzzzeek: i hit two differtn (but posibly related errors) 1: no such column error and 2: database connection reset error22:21
zzzeekjbernard: yeah22:21
jbernardzzzeek: every now and then :)22:21
e0nekaufer: how does cinderclient use it?22:22
jbernardzzzeek: this particular vm is ubuntu 14.04, that may be relevant22:22
zzzeekjbernard: it could be22:22
e0nekaufer: it calls /volumes/vol_id rest api method22:23
e0nekaufer: wich renders detailed view on the server-side22:24
kaufere0ne: Oh, for a show, yes22:24
kaufere0ne: I thought that you meant for a list when removing the detail suffix from the API call22:25
e0nekaufer: list works good on the server-side22:26
e0nekaufer: but client calls it with 'detail' suffux everytime22:26
*** jcru has quit IRC22:26
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Volume get and list should be able get detailed views
e0nekaufer: so what is the benefit to leave it as is? are there a lot of people that use API directly w/o client?22:34
e0nekaufer: i believe that only non-python users do it22:34
kauferTBH, I'm not sure, I was just surprised to see that the non-detailed view would be going away22:35
*** Yogi1 has quit IRC22:36
kaufere0ne: That is what you are proposing, right?22:36
e0nekaufer: i'm also don't have stats for it but it has backward compatibility and hasn't performance impact on the server side22:37
e0nekaufer: yes, it is22:37
*** nellysmitt has quit IRC22:37
*** nellysmitt has joined #openstack-cinder22:38
kaufere0ne: So, to back up, you are trying to define a framework where more details (ie, child_snapshots) can be retrieved for a volume, right?22:39
e0nekaufer: i'm really sorry. must go offline and sleep for a while. i'll be available tomorrow22:39
kaufere0ne: sure, np, I'll collect my thoughts and put something into the spec review, g'night!22:40
e0nekaufer: i didn't thanks that it is a framework, but it is something like you said22:40
*** annashen has joined #openstack-cinder22:41
*** Longgeek has quit IRC22:41
*** nellysmitt has quit IRC22:42
*** jdurgin has quit IRC22:43
*** ybathia has quit IRC22:44
*** mriedem has quit IRC22:46
*** afazekas has joined #openstack-cinder22:47
*** e0ne has quit IRC22:50
*** sdague has joined #openstack-cinder22:51
*** kaufer has quit IRC22:56
*** tsekiyama has quit IRC22:58
*** afazekas has quit IRC23:01
sdagueso... what do I need to change in default devstack to make this go away - ?23:02
*** markvoelker has joined #openstack-cinder23:02
*** tsekiyama has joined #openstack-cinder23:03
*** markvoelker has quit IRC23:07
*** annashen has quit IRC23:08
*** chlong has joined #openstack-cinder23:09
*** annashen has joined #openstack-cinder23:10
*** ebalduf has quit IRC23:12
openstackgerritxing-yang proposed openstack/python-cinderclient: Add support to incremental backups in cinder
*** xyang has quit IRC23:15
openstackgerritJay Bryant proposed openstack/cinder: Sync policy module from oslo-incubator
*** primechuck has quit IRC23:20
*** xyang has joined #openstack-cinder23:25
*** afazekas has joined #openstack-cinder23:30
*** annashen has quit IRC23:42
*** ebalduf has joined #openstack-cinder23:46
*** scottda_ has joined #openstack-cinder23:52
*** Mandell has quit IRC23:52
*** Mandell has joined #openstack-cinder23:54
*** scottda_ has quit IRC23:58
*** Mandell has quit IRC23:59

Generated by 2.14.0 by Marius Gedminas - find it at!