Wednesday, 2015-07-01

*** pv_ has quit IRC00:02
*** laughterwym has joined #openstack-cinder00:03
*** tsekiyam_ has joined #openstack-cinder00:12
*** IlyaG_ has quit IRC00:14
*** DericHorn-HP has joined #openstack-cinder00:15
*** mtanino has quit IRC00:15
*** DericHorn-HP has quit IRC00:16
*** tsekiyama has quit IRC00:16
*** tsekiyam_ has quit IRC00:17
*** logan2 has quit IRC00:19
*** chlong has joined #openstack-cinder00:21
*** logan2 has joined #openstack-cinder00:22
*** julim has quit IRC00:24
*** cebruns has quit IRC00:27
*** cebruns has joined #openstack-cinder00:28
*** anish has quit IRC00:29
*** jasondotstar has joined #openstack-cinder00:30
*** dims has joined #openstack-cinder00:32
*** leeantho has quit IRC00:33
*** annashen has joined #openstack-cinder00:34
*** breitz has quit IRC00:34
*** anish has joined #openstack-cinder00:35
*** dims_ has joined #openstack-cinder00:35
*** breitz has joined #openstack-cinder00:37
*** nikeshm has joined #openstack-cinder00:37
*** annashen has quit IRC00:38
*** dims has quit IRC00:39
*** IlyaG has joined #openstack-cinder00:46
*** zhenguo has joined #openstack-cinder00:47
*** IanGovett has quit IRC00:48
*** _cjones_ has quit IRC00:49
*** _cjones_ has joined #openstack-cinder00:50
*** Yogi1 has joined #openstack-cinder00:51
*** _cjones_ has quit IRC00:55
*** dannywilson has quit IRC00:55
*** annegentle has joined #openstack-cinder01:01
*** boris-42 has quit IRC01:02
*** Rockyg has joined #openstack-cinder01:03
*** tellesnobrega_ has joined #openstack-cinder01:05
*** IlyaG has quit IRC01:09
*** lixiaoy1 has joined #openstack-cinder01:10
*** r-daneel has quit IRC01:11
*** liuxg has joined #openstack-cinder01:13
*** IlyaG has joined #openstack-cinder01:14
*** bswartz has quit IRC01:14
*** Yogi1 has quit IRC01:17
*** changbl has joined #openstack-cinder01:19
*** davechen has joined #openstack-cinder01:20
openstackgerritTeruaki Ishizaki proposed openstack/cinder: sheepdog: change create and delete operation
*** Apoorva has quit IRC01:22
*** dannywilson has joined #openstack-cinder01:23
*** patrickeast has quit IRC01:24
*** liuxg has quit IRC01:25
*** liuxg has joined #openstack-cinder01:25
*** Rockyg has quit IRC01:28
*** asselin_ has joined #openstack-cinder01:33
*** annashen has joined #openstack-cinder01:35
*** IlyaG has quit IRC01:36
*** tellesnobrega_ has quit IRC01:38
*** annashen has quit IRC01:39
*** Lee1092 has joined #openstack-cinder01:40
*** liuxg has quit IRC01:42
*** annegentle has quit IRC01:42
*** liuxg has joined #openstack-cinder01:43
openstackgerritDave Chen proposed openstack/cinder: set/unset volume image metadata
*** liuxg has quit IRC01:47
*** heyun has joined #openstack-cinder01:48
*** liuxg has joined #openstack-cinder01:48
*** liuxg has quit IRC01:57
*** jwcroppe has quit IRC02:02
*** jwcroppe has joined #openstack-cinder02:02
openstackgerritDave Chen proposed openstack/python-cinderclient: Add commands for modifying image metadata
*** IlyaG has joined #openstack-cinder02:05
*** jwcroppe has quit IRC02:07
*** jasondotstar has quit IRC02:07
*** annegentle has joined #openstack-cinder02:08
*** Longgeek_ has joined #openstack-cinder02:10
*** Longgeek has quit IRC02:12
*** coolsvap|away has quit IRC02:13
*** haomaiwa_ has joined #openstack-cinder02:14
*** coolsvap|away has joined #openstack-cinder02:15
*** mriedem has quit IRC02:16
*** mutoulbj has joined #openstack-cinder02:18
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritwanghao proposed openstack/cinder-specs: Support query volume filter by glance metadata
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
*** xyang1 has quit IRC02:23
*** tobe has joined #openstack-cinder02:29
*** mutoulbj_ has joined #openstack-cinder02:30
*** mutoulbj has quit IRC02:34
*** annashen has joined #openstack-cinder02:36
*** haomaiwa_ has quit IRC02:39
*** Longgeek_ has quit IRC02:39
*** haomaiw__ has joined #openstack-cinder02:40
*** Longgeek has joined #openstack-cinder02:41
*** annashen has quit IRC02:41
*** winston-d_afk has quit IRC02:49
*** winston-d has joined #openstack-cinder02:50
*** annegentle has quit IRC02:53
openstackgerritRakesh H S proposed openstack/cinder: Return 404 if volume type encryption is not found
*** annegentle has joined #openstack-cinder03:04
*** haomaiw__ has quit IRC03:09
*** haomaiwang has joined #openstack-cinder03:10
*** liuxg has joined #openstack-cinder03:11
*** Longgeek has quit IRC03:14
openstackgerritLisaLi proposed openstack/cinder: Replace missed basestring by six for python3 compatability
*** liuxg has quit IRC03:16
*** mutoulbj_ has quit IRC03:18
openstackgerritwanghao proposed openstack/cinder: query volume detail support volume_glance_metadata
*** mutoulbj has joined #openstack-cinder03:21
*** mutoulbj has quit IRC03:23
*** garthb_ has joined #openstack-cinder03:24
*** garthb has joined #openstack-cinder03:24
*** breitz has quit IRC03:31
*** annegentle has quit IRC03:31
*** mutoulbj has joined #openstack-cinder03:32
*** bill_az has quit IRC03:34
*** yamada-h has joined #openstack-cinder03:36
*** annashen has joined #openstack-cinder03:37
*** annegentle has joined #openstack-cinder03:38
*** harlowja_at_home has joined #openstack-cinder03:40
*** annashen has quit IRC03:41
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
*** Longgeek has joined #openstack-cinder03:44
*** hemna has joined #openstack-cinder03:45
*** BharatK has joined #openstack-cinder03:46
openstackgerritLiu Xinguo proposed openstack/cinder: Refactor Huawei Volume driver
*** annegentle has quit IRC03:48
*** Longgeek has quit IRC03:49
*** Liu has joined #openstack-cinder03:51
*** Liu has quit IRC03:53
*** hemna has quit IRC03:53
*** MentalRay has joined #openstack-cinder03:56
*** heyun1 has joined #openstack-cinder03:56
*** heyun has quit IRC03:57
*** jwcroppe has joined #openstack-cinder04:00
*** dims_ has quit IRC04:03
*** davechen has quit IRC04:07
*** davechen has joined #openstack-cinder04:08
openstackgerritLiu Xinguo proposed openstack/cinder: Support SmartX for Huawei volume driver
*** coolsvap|away is now known as coolsvap04:14
*** harlowja_at_home has quit IRC04:15
*** davechen1 has joined #openstack-cinder04:17
*** davechen has quit IRC04:18
*** hemna has joined #openstack-cinder04:20
*** davechen has joined #openstack-cinder04:20
*** hemna has quit IRC04:21
*** dannywilson has quit IRC04:21
*** dannywilson has joined #openstack-cinder04:21
*** davechen1 has quit IRC04:22
*** morganfainberg is now known as caerbannograbbit04:24
*** caerbannograbbit is now known as CaerbannogRabbit04:24
openstackgerritVincent Hou proposed openstack/cinder: Implement the update_migrated_volume for the drivers
*** tobe has quit IRC04:26
*** deepakcs has joined #openstack-cinder04:33
*** Longgeek has joined #openstack-cinder04:34
*** annashen has joined #openstack-cinder04:37
*** avishay has joined #openstack-cinder04:40
*** shyama has joined #openstack-cinder04:41
*** nkrinner has joined #openstack-cinder04:41
*** annashen has quit IRC04:42
*** annegentle has joined #openstack-cinder04:49
*** MentalRay has quit IRC04:53
*** annegentle has quit IRC04:55
*** nihilifer has joined #openstack-cinder04:58
*** skraynev has joined #openstack-cinder05:02
*** chlong has quit IRC05:03
*** Nikolay_St has joined #openstack-cinder05:06
*** dims has joined #openstack-cinder05:07
*** dims has quit IRC05:12
*** chlong has joined #openstack-cinder05:21
*** tobe has joined #openstack-cinder05:26
*** heyun1 has quit IRC05:30
*** heyun has joined #openstack-cinder05:31
*** haomaiwang has quit IRC05:33
*** gaurang has joined #openstack-cinder05:36
*** haomaiwa_ has joined #openstack-cinder05:38
*** annashen has joined #openstack-cinder05:39
*** kaisers1 has quit IRC05:43
*** annashen has quit IRC05:43
*** gaurang1 has joined #openstack-cinder05:44
*** gaurang has quit IRC05:45
*** gaurang1 has left #openstack-cinder05:46
*** gaurang1 has joined #openstack-cinder05:46
*** gaurang1 has left #openstack-cinder05:49
*** jdurgin1 has joined #openstack-cinder05:51
*** gaurangt has joined #openstack-cinder05:52
*** dannywilson has quit IRC05:53
openstackgerritShay Halsband proposed openstack/cinder: XtremIO support for iscsi discovery auth
*** gaurangt has left #openstack-cinder05:56
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritDave Chen proposed openstack/cinder: set/unset volume image metadata
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
*** gaurangt has joined #openstack-cinder05:56
*** jdurgin has quit IRC05:57
*** garthb_ has quit IRC05:57
*** garthb has quit IRC05:57
*** tobe has quit IRC05:59
*** e0ne has joined #openstack-cinder05:59
*** jdurgin1 has quit IRC06:00
*** chlong has quit IRC06:01
*** chlong has joined #openstack-cinder06:01
*** BharatK has quit IRC06:02
openstackgerritLiu Xinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
*** e0ne has quit IRC06:11
openstackgerritLiu Xinguo proposed openstack/cinder: Add CHAP support for Huawei driver
*** BharatK has joined #openstack-cinder06:17
*** gaurangt has left #openstack-cinder06:18
*** chlong has quit IRC06:19
openstackgerritLiu Xinguo proposed openstack/cinder: Add multiple pools support for Huawei driver
*** chlong has joined #openstack-cinder06:34
*** skraynev has quit IRC06:35
openstackgerritLisaLi proposed openstack/cinder: Fix backup.rpcapi to pass object backup
*** annashen has joined #openstack-cinder06:39
openstackgerritMichal Dulko proposed openstack/cinder: Mock socket.gethostbyaddr in test_v7000_fcp
*** annashen has quit IRC06:44
*** agarciam has joined #openstack-cinder06:50
*** annegentle has joined #openstack-cinder06:51
*** tobe has joined #openstack-cinder06:53
*** belmoreira has joined #openstack-cinder06:54
*** skraynev has joined #openstack-cinder06:56
*** annegentle has quit IRC06:56
*** anshul has joined #openstack-cinder06:56
*** ndipanov has quit IRC06:57
*** sgotliv has quit IRC06:59
*** jdurgin has joined #openstack-cinder07:07
*** avishay has quit IRC07:12
*** avishay_ has joined #openstack-cinder07:12
openstackgerritDave Chen proposed openstack/cinder: set/unset volume image metadata
*** chlong has quit IRC07:24
winston-ddavechen: ping07:28
*** gaurangt has joined #openstack-cinder07:32
*** zhenguo has quit IRC07:32
*** zhenguo has joined #openstack-cinder07:33
*** annashen has joined #openstack-cinder07:40
*** annashen has quit IRC07:45
*** jistr has joined #openstack-cinder07:45
*** avishay_ has quit IRC07:49
*** avishay_ has joined #openstack-cinder07:50
openstackgerritXi Yang proposed openstack/cinder: Fix getting out-of-date volume operation state issue for VNX
*** e0ne has joined #openstack-cinder07:56
*** jordanP has joined #openstack-cinder07:57
openstackgerritTom Fifield proposed openstack/cinder: Fix typo in solidfire driver option
*** afazekas is now known as __afazekas07:59
openstackgerritAnton Arefiev proposed openstack/cinder: Add entry create and cast tasks to manage workflow
openstackgerritAnton Arefiev proposed openstack/cinder: Fix lvm manage existing volume
davechenwinston-d: hi, zhiteng08:04
davechendavechen: sorry, just see your message.08:04
davechenwiston-d: sorry, just see your message.08:04
davechenwinston-d: still around?08:05
winston-ddavechen: hi08:05
openstackgerritPeter Wang proposed openstack/cinder: Add white list support for target ports in VNX driver
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
winston-dthe enum34 lib isn't part of cinder requirement yet.08:06
davechenwinston-d: but it's in global-req, just as you said.08:06
winston-dyou will have to sync it from global-requirement before you can use it. otherwise jenkins will puke08:06
davechenwinston-d: I am confused, so we cannot rely on global-req?08:07
e0newinston-d: hi. i answered your questions accroding to
davechenwinston-d: how to sync with it, can you show me?08:08
winston-ddavechen: in short, you need to explicitly add 'enum34' to cinder/requirements.txt before you can use it08:08
davechenwinston-d: got it.08:08
davechenwinston-d: will update it shortly.08:08
*** bluex-pl has joined #openstack-cinder08:09
davecheneOne: thanks for your review on the patches about the policy protection.08:09
winston-de0ne: thx, what about the configurations for two setups?08:09
davechenwinston-d: anything else do you found with that patch?08:09
winston-ddavechen: i'm good with the rest08:10
e0newinston-d: i added it to the comment: "I tested it on VM with Ubuntu 14.04.1 LTS, 1 CPU, 1GB RAM, SSD. Rally: commit e4ee07f755f32a7962fb01dc0658e14d7ba622b7 Keystone: commit a70b514ed9aa1ccb4084e85fb7d342daabd337ed Keystone client: the latest version from PyPI."08:10
winston-de0ne: i saw that.08:10
winston-dthat's hardware config and software version.08:10
winston-di'm asking for detailed keystone configurations08:11
davechenwinston-d: And, Sean has already vote +2 on the patch of cinder client.
e0newinston-d: it's devstack configuration w/o changes08:11
winston-ddavechen: i'll wait for your server-side change to land.08:11
openstackgerritLiu Xinguo proposed openstack/cinder: Add CHAP support for Huawei driver
e0newinston-d: i just stopped apache and started 'keystone-all' manually08:11
davechenwinston-d: okay, many thanks, will ping you when that patch pass the Jenkins.08:12
winston-de0ne: can you paste keystone.conf and httpd.conf ?08:12
e0newinston-d: sure, 1min please08:13
winston-de0ne: thx08:13
*** sgotliv has joined #openstack-cinder08:13
davechenwinston-d: quite not understand why Jenkins passed with sync with global-req.08:15
davechenwinston-d: anyway, I will add it the requirment.txt08:16
openstackgerritLiu Xinguo proposed openstack/cinder: Add iSCSI multipath support for Huawei driver
e0newinston-d: apache config:
e0newinston-d: keystone config w/o commented lines:
*** __afazekas is now known as afazekas08:20
openstackgerritDave Chen proposed openstack/cinder: Role based properties protection
openstackgerritDave Chen proposed openstack/cinder: set/unset volume image metadata
openstackgerritDave Chen proposed openstack/cinder: Policies based properties protection
haypohemnafk: hi! i'm victor stinner08:23
winston-de0ne: is the same keystone.conf be used in both tests?08:23
e0newinston-d: yes08:23
winston-dhaypo: 1am for hemnafk, so ...08:23
haypowinston-d: i will wait for him ;)08:24
winston-de0ne: ok, so for keystone-apache, it would ignore 'admin/public_worker' option, right?08:24
e0newinston-d: yes, it will use WSGIDaemonProcess settings08:25
winston-de0ne: and for your apache config, 'keystone-public processes=5 threads=1' means you configured apache to use 5 processes, right?08:25
haypohey! os-brick got its (non-voting) py34 check job, cool ;)08:26
winston-de0ne: but for evenlet option, 'admin_workers = 2'08:26
winston-d2 vs 5.08:26
winston-dthat's not fair game.08:26
e0newinston-d: my fail08:26
winston-de0ne: that's why i need detailed config. :)08:26
e0newinston-d: i'll fix it later today08:26
e0newinston-d: sure, you're right08:26
winston-dhow many vCPUs does your VM have?08:26
e0newinston-d: i didn't verify that there are different numbers of workers08:27
e0ne1 CPU08:27
winston-de0ne: ok. so next time, when you re-do the test, please make sure the workers = processes. and paste the config to gerrit.08:28
e0newinston-d: thanks for pointing me where i'm wrong!08:28
winston-de0ne: np, thx for gathering the data.08:28
e0newinston-d: i'll update my tests results whith configs today before our meeting08:29
winston-de0ne: ok. sounds good.08:29
davechenwinston-d: do you know where Sean based?08:29
e0neneed to head to office. bb in ~1.5 hour08:29
winston-dSean is in US too, so basically he's sleeping.08:29
openstackgerritLisaLi proposed openstack/cinder: Fix backup.rpcapi to pass object backup
davechendamn it, gate-cinder-requirements failed after adding enum34 into req.txt.08:34
winston-ddavechen: you didn't do it right.08:35
*** mutoulbj_ has joined #openstack-cinder08:36
*** liudong has joined #openstack-cinder08:36
openstackgerritLisaLi proposed openstack/cinder: Fix backup.rpcapi to pass object backup
*** e0ne is now known as e0ne_08:36
davechenemm, how?08:36
*** liudong has left #openstack-cinder08:36
winston-ddavechen: the actual line in global-requirements looks like this:
winston-ddavechen: it must be exactly the same when adding it to cinder.08:37
winston-ddavechen: 'sync'08:37
*** gaurangt has quit IRC08:37
*** gaurangt has joined #openstack-cinder08:38
*** e0ne_ is now known as e0ne08:38
*** mutoulbj has quit IRC08:39
davechenwinston-d: oh, may local code is out of date.08:40
davechenwinston-d: how do you know all of these things. :)08:40
*** lpetrut has joined #openstack-cinder08:40
*** e0ne has quit IRC08:41
*** annashen has joined #openstack-cinder08:41
openstackgerritDave Chen proposed openstack/cinder: set/unset volume image metadata
*** avishay_ has quit IRC08:45
*** avishay__ has joined #openstack-cinder08:45
winston-ddavechen: hmm, there is no unit test for db api change and volume.api change08:45
*** annashen has quit IRC08:46
openstackgerritYuriy Nesenenko proposed openstack/cinder: Checking the existence of volume
*** laughterwym has quit IRC08:47
*** aix has joined #openstack-cinder08:48
*** jasondotstar has joined #openstack-cinder08:48
*** afazekas is now known as __afazekas08:50
openstackgerritVictor Stinner proposed openstack/cinder: Port dothill to Python 3
hayposmcginnis: hum, i had to rebase to fix a simple conflict in tox.ini. you may approve it again08:52
*** liudong has joined #openstack-cinder09:08
*** liudong has left #openstack-cinder09:08
*** dims has joined #openstack-cinder09:08
*** dims has quit IRC09:14
davechenwinston-d: can we address this in a seperated patch?09:18
davechenwinston-d: I am worried that testcase will bloat the patch again.09:19
*** haomaiwa_ has quit IRC09:21
davechenwinston-d: seems there is no new method added but some refactoring instead.09:21
*** afazekas has joined #openstack-cinder09:24
*** yrabl has quit IRC09:29
avishay__Anyone here familiar with the LVM driver?  Specifically LIO vs TGT?09:29
*** yrabl has joined #openstack-cinder09:29
*** avishay__ is now known as avishay09:29
openstackgerritwanghao proposed openstack/cinder: Query volume detail support volume_glance_metadata
*** aix has quit IRC09:31
*** yamada-h has quit IRC09:31
*** e0ne has joined #openstack-cinder09:35
*** laughterwym has joined #openstack-cinder09:37
openstackgerritYuriy Nesenenko proposed openstack/cinder: Checking the existence of volume
*** ociuhandu has quit IRC09:39
*** annashen has joined #openstack-cinder09:42
*** davechen has left #openstack-cinder09:42
*** annashen has quit IRC09:46
*** vincent_hou has joined #openstack-cinder09:47
vincent_hougeguileo; Hey, dude.09:48
geguileovincent_hou: Hi09:48
vincent_hougeguileo: I add the explanation at the bottom.09:48
geguileovincent_hou: Is it about my downvote for the StorPool?09:48
vincent_houYes. and my opinion about the rename.09:48
geguileovincent_hou: So you say that even if that throws an exception if will be Ok?09:49
vincent_houI think we do not have to throw an exception.09:49
geguileovincent_hou: No, we don't have to09:49
geguileovincent_hou: But if the rename raises the exception09:49
vincent_houBut need to handle _name_id well09:49
avishaygeguileo: hey09:50
geguileoavishay: Hi09:50
geguileoavishay: Any progress with your tgt?09:50
avishaygeguileo: Yes09:50
geguileoavishay: Awesome!!09:51
geguileoavishay: What was it?09:51
avishaygeguileo: The first problem is that the centos package installs some default conf file that places iscsi_helper=lioadm in the [DEFAULT] section09:51
geguileoavishay: That's why I asked you to look at all configurations  ;-)09:52
vincent_houI will drop off now. Any concerns, just leave your comment there.09:52
geguileoavishay: in /usr/share09:52
vincent_houSee you, both.09:52
geguileovincent_hou: My concern is what I said09:52
geguileovincent_hou: If the rename raises an exception09:52
geguileovincent_hou: We should catch it09:52
avishaygeguileo: I added iscsi_helper = tgtadm to the backend section, and I think maybe sometimes it was using LIO and sometimes TGT09:52
geguileoavishay: You need to have DEFAULT set to tgtadm and backend set to tgtadm as well09:53
avishaygeguileo: Further, I see that scsi-target-utils is no longer included in RH709:53
geguileoavishay: Or just remove it from both and it will use tgt by default09:53
*** lixiaoy1 has quit IRC09:53
avishaygeguileo: Well that's broken - it's a driver setting, not a general setting09:53
vincent_hougeguileo: Ah, I see your point.09:54
geguileoavishay: That's how Openstack configuration works09:54
geguileoavishay: Defaults are not inherited  :(09:54
vincent_houSorry, I missed a little.09:54
geguileovincent_hou: Sometimes I don't explain myself very well, sorry   XD09:54
avishaygeguileo: So basically if I want something supported in RH I need to use the LIO driver, which is broken and has no CI09:54
vincent_hougeguileo: Yes, I am with you geguileo09:54
geguileoavishay: No, no, tgt should work as well09:55
avishaygeguileo: tgt is not supported in RH709:55
geguileoavishay: There's already a patch upstream to fix broken LIO09:55
avishaygeguileo: it was moved to EPEL09:55
geguileoavishay: And I'll start working on a CI pretty soon09:55
vincent_houC U.09:55
*** vincent_hou has quit IRC09:55
avishaygeguileo: i won't run anything without a CI - LIO should have been removed09:56
geguileoavishay: LIO is not a driver09:56
avishaygeguileo: LVM+LIO is a driver09:56
geguileoavishay: The driver is iSCSI  ;-)09:56
avishaygeguileo: you know that's BS as well as I do09:57
openstackgerritwanghao proposed openstack/cinder: Incremental backup improvements for L
geguileoavishay: It's not complete BS, but I see your point09:57
geguileoavishay: That's why I'll work on the CI  :-D09:57
avishaygeguileo: it's 100% BS.  Just like HP has iSCSI and FC CIs, LVM needs CIs for all target types.09:58
*** aix has joined #openstack-cinder09:58
geguileoavishay: iSCSI and FC are protocols09:58
*** dims has joined #openstack-cinder09:59
geguileoAnyway, no reason to argue09:59
geguileoSince we both agree there should be a CI  :)09:59
geguileoI'll keep you posted on the progress09:59
avishaygeguileo: I will suggest in the Cinder meeting that all non-tgt backends should be removed until they have a CI09:59
avishaygeguileo: Semantic loopholes are not interesting to me when I waste 2-3 days of work on a broken driver10:00
avishaygeguileo: Sorry I'm a bit ticked off, but RH really dropped the ball on this one10:01
*** BharatK has quit IRC10:01
avishaygeguileo: Nothing personal10:01
geguileoavishay: I know it's not personal, don't worry10:01
geguileoavishay: And the problem you had is not specific with LIO10:01
e0newinston-d: are you around?10:01
geguileoavishay: The problem is with oslo configuration10:01
*** yamada-h has joined #openstack-cinder10:01
geguileoavishay: Since backends don't inherit from default value the iscsi_helper10:02
avishaygeguileo: that's the latest problem when i tried moving away from LIO10:02
geguileoavishay: Oh, ok10:02
avishaygeguileo: LIO is better than TGT in my opinion, and i would have been happy to use it if the driver worked and was tested10:02
*** wanghao has left #openstack-cinder10:03
*** mutoulbj_ has quit IRC10:03
geguileoavishay: Well, now you are making me feel bad, because I'm the one who broke it in Juno  :(10:03
sgotlivavishay, so I guess we have to work to make it working10:03
avishaygeguileo: You don't have to feel bad, it happens, and that's what CI is for10:03
*** wanghao has joined #openstack-cinder10:03
*** wanghao has left #openstack-cinder10:03
*** wanghao has joined #openstack-cinder10:04
avishaygeguileo: I think jgriffith had some issues with CI, where LIO behaved differently on RH and Ubuntu, but I don't know the details10:04
geguileoavishay: I guess I'll find out once I start working on it :)10:05
avishaygeguileo: anyway, i hope you can get it running soon10:05
avishaygeguileo: thank you10:05
geguileoavishay: He mentioned that the main problem is that Ubuntu does not support LIO, you need some kernel changes and whatnot10:05
avishaygeguileo: could be, like i said, i don't know the details10:06
geguileoavishay: I'll figure it out, no problem10:06
*** yamada-h has quit IRC10:06
avishaygeguileo: you have a fix for LIO now?  tested somehow?10:07
geguileoavishay: The patch is waiting review:
geguileoavishay: But it's not passing the gates because of an issue with the gate10:07
geguileoavishay: So I had to write a new elastic-recheck fingerprint10:08
avishaygeguileo: OK, if I don't make progress with tgt soon I may go back to lio, and I hope to have a CI in the near future10:09
e0newinston-d: i've updated review requests with new rally results10:09
geguileoavishay: Ok, if you move back to LIO you just need to remove a couple of lines for it to work  ;-)10:09
avishaygeguileo: thanks a lot10:09
avishaygeguileo: i'm worried that it's broken in other places because it's not tested10:10
geguileoavishay: I can tell you that there are some potential issues in Juno that are fixed in Kilo10:12
geguileoavishay: But the most common got fixed:
avishaygeguileo: yes applied that one10:12
openstackgerritMichal Dulko proposed openstack/cinder: Service object
openstackgerritMichal Dulko proposed openstack/cinder: Fix saving tz aware datetimes in Versioned Objects
*** lpetrut has quit IRC10:24
geguileodulek: Why did you update a patch to my Versioned object tz fix patch?10:29
*** BharatK has joined #openstack-cinder10:32
*** ndipanov has joined #openstack-cinder10:36
*** e0ne is now known as e0ne_10:39
*** e0ne_ is now known as e0ne10:39
*** DericHorn-HP has joined #openstack-cinder10:42
*** annashen has joined #openstack-cinder10:43
*** annashen has quit IRC10:48
*** sgotliv has quit IRC10:52
*** sgotliv has joined #openstack-cinder10:53
*** mutoulbj has joined #openstack-cinder10:53
*** zhenguo has quit IRC10:54
*** BharatK has quit IRC10:57
*** heyun has quit IRC11:02
*** ociuhandu has joined #openstack-cinder11:03
dulekgeguileo: I needed to rebase it to make my Service object patch dependent on it.11:03
dulekgeguileo: Don't worry, votes are still there, because I haven't changed anything. :)11:04
dulekgeguileo: Authorship info also seems to be intact.11:05
geguileodulek: Yes, only LP bug has changed.11:06
geguileodulek: Not cool updating my patch. ;-)  (kidding)11:07
dulekgeguileo: Hm? Oh, I see, patch topic. You can update it from Gerrit UI. Basically it doesn't mean anything, LP tracks patches by Closes-Bug flags in commit messages.11:08
dulekgeguileo: I've could prevented that, sorry.11:08
geguileodulek: And now you are the assignee for this bug in LP11:09
*** IanGovett has joined #openstack-cinder11:09
*** avishay has quit IRC11:10
*** BharatK has joined #openstack-cinder11:10
dulekgeguileo: Fixed.11:11
geguileodulek: XD, there was no need to change it11:12
dulekgeguileo: Well, statistics count. ;)11:12
*** ankit_ag has joined #openstack-cinder11:14
*** Vikash_cz has joined #openstack-cinder11:16
*** anshul has quit IRC11:16
*** anshul has joined #openstack-cinder11:17
openstackgerritYuriy Nesenenko proposed openstack/cinder: Filter cgsnapshots data on the DB side
*** avishay has joined #openstack-cinder11:20
*** yuriy_n17 has joined #openstack-cinder11:21
*** e0ne is now known as e0ne_11:24
*** e0ne_ is now known as e0ne11:24
*** DericHorn-HP has quit IRC11:25
*** gaurangt has left #openstack-cinder11:28
*** yrabl has quit IRC11:28
*** yrabl has joined #openstack-cinder11:29
*** jasondotstar has quit IRC11:35
*** jasondotstar has joined #openstack-cinder11:36
*** jasondotstar has quit IRC11:41
*** annashen has joined #openstack-cinder11:44
*** tobe has quit IRC11:46
*** avishay has quit IRC11:46
*** annashen has quit IRC11:48
geguileodulek: ping - HA11:56
*** Longgeek has quit IRC11:59
*** lpetrut has joined #openstack-cinder12:05
*** abhishekk has joined #openstack-cinder12:08
*** ociuhandu has quit IRC12:09
*** markvoelker has quit IRC12:14
*** markvoelker has joined #openstack-cinder12:14
*** eharney has joined #openstack-cinder12:14
*** ociuhandu has joined #openstack-cinder12:21
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
openstackgerritErickson Filipe Guedes dos Santos proposed openstack/cinder: Nested Quota: Set default values to subproject
*** Longgeek has joined #openstack-cinder12:25
*** deepakcs has quit IRC12:26
*** juzuluag has joined #openstack-cinder12:26
*** zigo has quit IRC12:28
*** mutoulbj has quit IRC12:28
*** zigo has joined #openstack-cinder12:29
*** mutoulbj has joined #openstack-cinder12:29
*** BharatK has quit IRC12:29
*** mutoulbj has quit IRC12:30
*** mutoulbj has joined #openstack-cinder12:30
openstackgerritVincent Hou proposed openstack/cinder: Implement the update_migrated_volume for the drivers
*** Longgeek has quit IRC12:31
*** nihilifer has quit IRC12:32
*** edmondsw has joined #openstack-cinder12:32
*** kjnelson has joined #openstack-cinder12:34
*** BharatK has joined #openstack-cinder12:35
*** annegentle has joined #openstack-cinder12:37
*** ankit_ag has quit IRC12:38
*** tellesnobrega_ has joined #openstack-cinder12:40
*** e0ne is now known as e0ne_12:41
*** e0ne_ is now known as e0ne12:42
*** thangp has joined #openstack-cinder12:42
*** annashen has joined #openstack-cinder12:45
*** tellesnobrega__ has joined #openstack-cinder12:47
*** annegentle has quit IRC12:48
*** annegentle has joined #openstack-cinder12:49
*** tellesnobrega_ has quit IRC12:49
*** annashen has quit IRC12:49
*** Longgeek has joined #openstack-cinder12:49
*** amoturi has joined #openstack-cinder12:53
*** e0ne is now known as e0ne_12:53
*** Lee1092 has quit IRC12:55
*** tellesnobrega__ has quit IRC12:55
*** Ctina__ has joined #openstack-cinder12:56
*** e0ne_ is now known as e0ne12:57
*** annegentle has quit IRC12:57
*** annegentle has joined #openstack-cinder12:58
*** flip214 has quit IRC13:00
*** kbyrne has quit IRC13:03
*** yamada-h has joined #openstack-cinder13:04
*** yamada-h has quit IRC13:08
winston-de0ne: around?13:10
winston-de0ne: saw the new result.13:10
e0newinston-d: yes13:10
winston-de0ne: thx. i remember you said the VM has only one vCPU.13:11
e0nekvm instance with 1 cpu13:11
winston-de0ne: is it possible to do the test on a multi-core VM? 2 is fine, 4 would be ideal13:11
openstackgerritPetrut Lucian proposed openstack/cinder: RemoteFS: Fix the offline snapshot delete operation
e0newinston-d: i'll test with 2 cpu13:12
e0newinston-d: need to find env to setup 4 cpu per instance13:12
*** bluex-pl has quit IRC13:13
winston-dif we can have 4vCPU VM, we can collect 3 data point for both test cases: 1 worker/1 process; 2 workers/2 processes; 4 workers/4 processes. then we can draw a line for each to see how they scale13:14
*** Longgeek has quit IRC13:14
e0newinston-d: how many workers should i setup for 2 vCPU?13:14
*** Longgeek has joined #openstack-cinder13:14
*** subscope has joined #openstack-cinder13:15
*** bill_az has joined #openstack-cinder13:15
winston-dif you can find 2vCPU VM, then we can collect 2 data points for each  setup: 1worker/1process and 2 workers/2 processes.13:15
winston-de0ne: what's the difference between 'load duration' and 'full duration'?13:17
*** avishay has joined #openstack-cinder13:19
*** ociuhandu has quit IRC13:19
winston-de0ne: does keystone have osprofiler support?13:19
*** ociuhandu has joined #openstack-cinder13:20
winston-de0ne: it'd be interesting to see the latency breakdown13:20
e0newinston-d: afaik, it doesn't support osprofiler13:21
winston-de0ne: it's fine.13:21
*** dtynan1 has joined #openstack-cinder13:22
*** kbyrne has joined #openstack-cinder13:22
openstackgerritVincent Hou proposed openstack/cinder-specs: Volume migration improvement for Liberty version
winston-de0ne: so you new/old data showed that changing # of workers/processes in a single CPU environment doesn't affect keystone peformance, which is expected.13:22
*** dtynan has quit IRC13:22
e0newinston-d: i think, load duration is duration w/o rally invokes. trying to veryfy it from rally team13:22
winston-de0ne: but apache sems to have smaller overhead than eventlet, so the data is still valuable.13:23
openstackgerritVincent Hou proposed openstack/cinder-specs: Volume migration improvement for Liberty version
winston-d2vCPU VM data will be more interesting, it should tell us more.13:24
dulekgeguileo: pong, what's up?13:25
dulekgeguileo: I need "I survived another meeting sticker". ;)13:25
geguileodulek: XD13:25
geguileodulek: That's a good idea, I would love one too  ;-)13:26
*** coolsvap is now known as coolsvap|away13:26
geguileodulek: I have added HA to the meeting's agenda13:26
geguileodulek: And I didn't know if you wanted to lead as well13:26
geguileos/as well/too13:27
*** ociuhandu has quit IRC13:27
*** mriedem has joined #openstack-cinder13:27
dulekgeguileo: I'll participate today, if you need backup - I can cover. :)13:27
geguileodulek: Ok13:28
geguileodulek: The main idea is to get a list of Cinder-Nova issues stopping HA progress13:28
dulekgeguileo: To be honest - I'm a little confused on that now.13:29
geguileodulek: Someone must know about that one, right?13:29
e0newinston-d: and
e0ne2 workers per cpu13:29
geguileodulek: The other one is locking deletion from resources used for reading (allowing multiple readers)13:30
geguileodulek: And finally get a fine grained list of jobs and who's going to be doing them13:30
dulekgeguileo: Okay, I'm familiar with that one.13:30
geguileodulek: Because I think we are not making much progress13:30
e0newinston-d: meeting:(. i'll test with 2 cpu and 1 worker per cpu a bit later13:30
dulekgeguileo: And that's needed also. :)13:30
*** ociuhandu has joined #openstack-cinder13:30
*** dtynan1 has quit IRC13:31
dulekgeguileo: My perspective on Nova->Cinder is that attach has reserve and detach has begin_detaching.13:31
geguileodulek: Do you know what point 3 in action items is referring to?
dulekgeguileo: Yeah, it's about locking deletion from resources used for reading.13:32
dulekgeguileo: We somehow figured it out. ;)13:32
geguileodulek: Oh, that's "" changes?13:32
*** avishay has quit IRC13:32
dulekgeguileo: Mhm.13:32
dulekgeguileo: The way it's written is hard to understand, that's why this item is there.13:33
geguileodulek: I thought it was referring to atomic state changes   XD13:33
*** dtynan has joined #openstack-cinder13:33
geguileodulek: We definitely need better naming in action items  XD13:34
*** avishay has joined #openstack-cinder13:34
dulekgeguileo: It is related. Once atomic state changes are done we won't need these locks (in most cases I think).13:34
e0newinston-d: also, i could test with concurrency >10. it could be interesting13:34
geguileodulek: No we'll need some of them, the ones locking resources that are used only for reading13:35
avishaygeguileo: I've decided to stick with LIO since I'd rather have a less stable driver than deprecated target, hope to see a CI soon :)13:35
geguileoavishay: Cool13:35
dulekgeguileo: Okay, so an example would be creating volume from snapshot - we need to lock on snapshot for reading.13:36
geguileoavishay: We'll discuss it in today's meeting :)13:36
avishaygeguileo: Have you seen this error?   rtslib.utils.RTSLibError: This _Backstore already exists in configFS13:36
geguileodulek: Yes, and there could be multiple volumes creating from that one13:36
geguileoavishay: Yes13:36
dulekgeguileo: We can probably change state of the snapshot in the db to used_for_volume_creation and allow multiple creations but no deleting13:36
geguileoavishay: I fixed that one in L and backported it to K13:36
geguileodulek: Then you can only have 1 reader13:37
avishaygeguileo: link to kilo backport please?  i'd like to apply it to our juno13:37
geguileodulek: I sent my proposal to the mail list and I refer to it in the agenda13:37
geguileoavishay: 1 sec13:37
dulekgeguileo: I think depending on API call you can allow multiple operations or not. But you would need to maintain a list of operations already pending... Okay, this is getting messy.13:38
*** abhishekk has quit IRC13:38
winston-de0ne: and maybe we can increate the rally iterations to 500 or even 1,00013:38
avishaygeguileo: thanks a lot!13:38
geguileoavishay: You may be also interested on the persistence one13:39
avishaygeguileo: sure13:39
geguileodulek: That's why I added it to today's meeting, because we have to decide how to do it  XD13:39
dulekgeguileo: Okay, fine for me.13:40
*** julim has joined #openstack-cinder13:42
geguileoavishay: I don't know if the persistence one is in Kilo:
geguileoavishay: And that introduces a bug if you don't have targetcli installed, so there was this other patch later:
*** Ctina__ is now known as ctina13:43
geguileoavishay: Hope it all goes well now  :-)13:44
avishaygeguileo: awesome thank you!13:45
geguileoavishay: No problem13:45
*** annashen has joined #openstack-cinder13:45
*** dtynan has quit IRC13:46
*** dtynan has joined #openstack-cinder13:47
*** jungleboyj has quit IRC13:49
*** annashen has quit IRC13:50
*** ociuhandu has quit IRC13:51
e0newinston-d: times 300, concurrency 50 -
*** Vikash_cz has quit IRC13:53
*** ociuhandu has joined #openstack-cinder13:53
winston-de0ne: cool. the link text should change, they are all eventlet13:56
e0neyep. fixed13:56
e0netimes 1000, concurrency 50 - in process13:57
*** dims has quit IRC13:59
*** dims has joined #openstack-cinder14:00
*** andrey-mp has joined #openstack-cinder14:01
*** mtanino has joined #openstack-cinder14:02
*** subscope has quit IRC14:03
winston-de0ne: thx. the data is already very promising.14:03
*** jungleboyj has joined #openstack-cinder14:03
*** avishay has quit IRC14:04
*** avishay has joined #openstack-cinder14:05
*** dalgaaf has joined #openstack-cinder14:06
openstackgerritxhzhf proposed openstack/cinder: VMware driver: compatible to vc 6.0
*** lpetrut has quit IRC14:07
*** e0ne is now known as e0ne_14:08
*** lcurtis has joined #openstack-cinder14:08
*** lpetrut has joined #openstack-cinder14:09
*** avishay has quit IRC14:10
andrey-mpDuncanT, hemnafk, please see my answers in review (Avoid race condition at snapshot deletion stage). We were stuck because of this bug. Gating of EC2 project is failed very often...14:11
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
DuncanT@andrey-mp I'll take a look now14:11
DuncanT@andrey-mp So I still don't see why we need to call snapshot_metadata_get for each snap in the list - that is just broken behaviour14:14
*** zhenguo has joined #openstack-cinder14:14
lpetrutHi guys. I have a few questions: can we still send blueprints targeting Liberty?14:15
DuncanT@andrey-mp If there's a list codepath that hits this code, that codepath is totally broken and should be fixed14:15
andrey-mpDuncanT, I think that code doesn't know about join in DB layer - so I left snapshot_metadata_get for such cases. But in this case bug can come back...14:15
lpetrutWe'd like to support volume backup/migration for the Windows iSCSI and SMB drivers. One step would be adding Windows support in os-brick14:16
andrey-mpSo you mean that I should remove snapshot_metadta_get at all and be sure that all queries have join clause?14:16
DuncanT@andrey-mp I'd rather this code was removed - it is dangerous - somebody with many snapshots doing a snapshot list can bring the db to a grinding halt14:17
DuncanT@andrey-mp Ideally you'd only do the join in codepaths that need it, but that is hard to get right, particularly now we've moved to objects, so I think always joining is better than this14:18
andrey-mpIn this case I have to rework many unit tests that rely on this feature. is it ok?14:19
DuncanTTo me, yes14:20
DuncanTI'm not sure I understand to comment "First change removes getting metadata from DB for each found snapshot.14:21
DuncanTInstead of this it copies metadata from query result. This query14:21
DuncanTis already joined with metadata table."14:21
DuncanTIt looks like you're adding a db query, not removing one14:22
DuncanTI might be failing to understand, however14:22
andrey-mpok, I will rewrite commit message ) It means that first change tries to copy metadata from db object (as in code)14:23
DuncanTOh, I think I see14:24
DuncanTOk, sorry, I think this change can probably go in as-is - we just need to do an audit to make sure all of the API list snapshot paths do a joined load14:25
*** jungleboyjiPad has joined #openstack-cinder14:26
andrey-mpOuch ) Now i'm sure that all snapshot queries have join clause. But snapshot code calls snapshot_metadata_get if metadata is absent in db object.14:27
DuncanTChanged my vote to 0 for now, I'll go read Walt's comments and maybe I'll change it to +2 - certainly nobody should feel like I strongly object anymore, I'd like to see those unit tests cleaned up, but that is future work14:28
DuncanTMost of our unit tests, if written correctly, shouldn't care IMO14:29
andrey-mpmost of unit tests now mock snapshot_metadata_get and use fake snapshot db object without metadata. But live cloud uses snapshot db object with metadata. It is confusing...14:31
DuncanTandrey-mp: Could we keep track of whether we did a joined load or not and make it an exception to have metadata in the required fields if you didn't do a joined load?14:32
DuncanTexpected_attrs, not required fields14:33
DuncanT(I'm not saying instead of your patch, I'm asking for future work)14:34
andrey-mphm, it make sense. I will do this. (And rework most unit tests also)14:35
andrey-mpI think that I can do this in this patch. It will be more correct.14:35
*** ociuhandu_ has joined #openstack-cinder14:37
*** jwang__ has joined #openstack-cinder14:37
*** serverascode_ has joined #openstack-cinder14:39
*** Longgeek has quit IRC14:40
*** cppforlife__ has joined #openstack-cinder14:40
*** guitarza1 has joined #openstack-cinder14:42
*** thingee_ has joined #openstack-cinder14:43
openstackgerritKendall Nelson proposed openstack/cinder: Enable cinder-manage to delete old services from the database
*** EmilienM_ has joined #openstack-cinder14:45
*** mtreinish_ has joined #openstack-cinder14:45
*** rwall_ has joined #openstack-cinder14:45
openstackgerritVictor Stinner proposed openstack/cinder: Fix Python 3 issues in Hitachi HNAS tests
openstackgerritVictor Stinner proposed openstack/cinder: Port remotefs driver to Python 3
*** ociuhandu has quit IRC14:46
*** juzuluag has quit IRC14:46
*** marcusvrn has quit IRC14:46
*** mtreinish has quit IRC14:46
*** Swanson has quit IRC14:46
*** rwall has quit IRC14:46
*** rushiagr_away has quit IRC14:46
*** andymccr has quit IRC14:46
*** git-harry has quit IRC14:46
*** rhe00_ has quit IRC14:46
*** jwang_ has quit IRC14:46
*** EmilienM has quit IRC14:46
*** cppforlife_ has quit IRC14:46
*** rex_lee has quit IRC14:46
*** thingee has quit IRC14:46
*** serverascode has quit IRC14:46
*** guitarzan has quit IRC14:46
*** EmilienM_ is now known as EmilienM14:46
*** juzuluag_ has joined #openstack-cinder14:46
*** strictlyb is now known as sb14:46
*** ociuhandu_ is now known as ociuhandu14:46
*** mtreinish_ is now known as mtreinish14:46
*** rwall_ is now known as rwall14:46
*** annashen has joined #openstack-cinder14:46
*** annegentle has quit IRC14:47
*** nkrinner has quit IRC14:47
*** MentalRay has joined #openstack-cinder14:47
*** rhe00_ has joined #openstack-cinder14:48
*** thingee_ is now known as thingee14:48
*** DericHorn-HP has joined #openstack-cinder14:48
*** sb has quit IRC14:51
*** annashen has quit IRC14:51
*** git-harry has joined #openstack-cinder14:51
*** strictlyb has joined #openstack-cinder14:51
*** cppforlife__ is now known as cppforlife_14:53
*** marcusvrn has joined #openstack-cinder14:54
*** Swanson has joined #openstack-cinder14:55
*** serverascode_ is now known as serverascode14:56
DuncanTandrey-mp: That's great, thanks! Sorry to have to cause you more work, but I think it will make things better in the end14:56
andrey-mpsure :)14:56
*** subscope has joined #openstack-cinder14:57
*** rmesta has joined #openstack-cinder14:58
jordanPwith the LVM driver, I can't extend a volume that has snapshots. It fails with "Snapshot origin volumes can be resized only while inactive: try lvchange -an"15:00
jordanPis that expected ?15:00
*** jordanP has quit IRC15:01
*** xyang1 has joined #openstack-cinder15:02
*** jordanP has joined #openstack-cinder15:02
*** andrey-mp has quit IRC15:02
eharneyjordanP: sounds like a bug15:04
jordanPeharney, will submit a bug report then15:04
jordanP*i will15:04
*** rex_lee has joined #openstack-cinder15:05
*** guitarza1 is now known as guitarzan15:06
*** andymccr has joined #openstack-cinder15:06
*** Longgeek has joined #openstack-cinder15:07
*** e0ne_ is now known as e0ne15:07
*** Longgeek has quit IRC15:07
*** rmesta has left #openstack-cinder15:08
*** Longgeek has joined #openstack-cinder15:08
lpetruteharney: Hi. I was thinking about adding volume backup support for the RemoteFS based drivers. I think we could easily support backing up volumes having snapshots as well by mounting them, using qemu-nbd for example.15:09
*** Apoorva has joined #openstack-cinder15:09
*** belmoreira has quit IRC15:10
*** Longgeek has quit IRC15:12
*** Longgeek has joined #openstack-cinder15:14
*** blmartin has joined #openstack-cinder15:15
openstackLaunchpad bug 1470558 in Cinder "LVM driver: volume with snaps can't be extended" [Undecided,New]15:15
DuncanTlpetrut: Are there still security concerns around qemu-nbd on user-controlled volumes? They've certainly been raised before, though I don't know how substantial they are15:15
jordanPas an ops, I don't like qemu-nbd. I've seen weird stack traces in my kernel.log files when we used to no use libguestfs15:16
jordanPlpetrut, so if you want to go forward with qemu-nbd, I recommend to code a switch to be able to completely disable the call to qemu nbd15:17
jordanP(and loose some feature of course)15:17
*** liewegas has joined #openstack-cinder15:18
*** e0ne is now known as e0ne_15:19
*** e0ne_ is now known as e0ne15:20
e0newinston-d: results for 1000 runs with concurrency 50 are ready:
*** aarefiev_home has joined #openstack-cinder15:24
*** afazekas has quit IRC15:26
*** IlyaG has quit IRC15:26
openstackgerritVictor Stinner proposed openstack/os-brick: Cleanup Python 3 changes
*** CaerbannogRabbit is now known as morganfainberg15:30
hayposmcginnis: "Huawei CI shows a failure, but its logs are not accessible." hum, i just schedule a recheck15:32
lpetrutthanks for the heads up guys. DuncanT: you are reffering to the image format autodetection issue, right? Well, we can easily specify the expected format15:32
hayposmcginnis: i wrote patches per driver because of all these Cinder CIs :)15:32
smcginnishaypo: Yeah, don't worry about it. I'm pretty sure it's an issue on their end and not your patch.15:32
smcginnishaypo: I appreciate that. Good to see the individual confirmations from the CIs. When they actually work. :)15:33
*** coolsvap|away is now known as coolsvap15:33
hayposmcginnis: only Jenkins is allowed to vote -1, or some vendor CIs are also allowed to vote -1?15:33
smcginnishaypo: I think one of the VMware CIs has voting, but we've been holding of for stability before letting them all.15:33
*** xyang has joined #openstack-cinder15:33
hayposmcginnis: yeah, it's really hard to get a stable CI :-( i know that15:34
smcginnishaypo: Yeah, I'm in the same boat. Mostly stable, but still not as much as I would like it to be.15:34
haypoi put my py3 patches depending on os-brick into WIP. i'm now working with walter on - i hope that a new version of os-brick with these py3 fixes will be released quickly15:35
eharneylpetrut: there is a patch out now related to backup of qcow2:
*** tsekiyama has joined #openstack-cinder15:36
haypohemnafk: ping again for ;)15:36
smcginnishaypo: Awesome, was just looking at hemnafk's patch and read your comment.15:36
smcginnishaypo: Would be good to combine them IMO.15:36
lpetruteharney: thanks for the link, checking it out15:37
hayposmcginnis: i don't like "hijacking" a patch. i hate when someone does that on my patches :)15:37
eharneylpetrut: it depends on some other work which hasn't landed yet so it's not quite ready15:37
*** jaypipes has quit IRC15:37
eharneylpetrut: there were also some concerns about the idea in general which i'm not sure have been sorted out adequately15:37
*** Apoorva has quit IRC15:39
e0neeharney: just fyi. i'm testing this simple patch
e0neeharney: from the first look - it works. i'm waiting results from the load tests15:41
*** liuxg has joined #openstack-cinder15:42
*** dtynan has quit IRC15:43
openstackgerritJon Bernard proposed openstack/cinder: Add support for generic volume migration
*** coolsvap is now known as coolsvap|away15:44
*** dtynan has joined #openstack-cinder15:44
eharneye0ne: isn't that what we had previously?15:47
*** annashen has joined #openstack-cinder15:47
e0neeharney: no, i leave tpool to connect to RDB and removed in _connect_to_rados method15:48
*** adurbin_ has joined #openstack-cinder15:48
*** laughterwym has quit IRC15:48
e0neeharney: to decrease cross-threading communications15:48
*** annashen has quit IRC15:52
*** dannywilson has joined #openstack-cinder15:52
*** hemnafk is now known as hemna15:55
e0nehemna: evening:)15:55
*** patrickeast has joined #openstack-cinder15:55
*** flip214 has joined #openstack-cinder15:57
*** erlon has joined #openstack-cinder15:57
*** deepakcs has joined #openstack-cinder15:58
hemnaso, I don't want to remove the testenv:py34 section in tox.ini15:59
hemnait causes runs of tox -epy34 to fail15:59
*** zongliang has joined #openstack-cinder16:00
openstackgerritAlberto Murillo proposed openstack/cinder: Update expected error message from lvs
*** alau2 has joined #openstack-cinder16:01
*** rmesta has joined #openstack-cinder16:04
*** aarefiev_home has quit IRC16:04
*** garthb has joined #openstack-cinder16:04
*** garthb_ has joined #openstack-cinder16:04
smcginnishemna: When I did a side-by-side I thought there were a few things haypo had in his patch that you didn't. Not related to tox.ini.16:04
smcginnishemna: Did you take a look at that?16:04
*** lpetrut has quit IRC16:04
*** aarefiev_home has joined #openstack-cinder16:05
*** Vikash_cz has joined #openstack-cinder16:05
*** lpetrut has joined #openstack-cinder16:05
*** lpetrut has quit IRC16:05
*** leeantho has joined #openstack-cinder16:06
hemnahis follow up/dep patch had tox.ini changes (removing the section) and removed the changes16:06
hemnaI can try removing the changes and see if the tests pass again16:06
hemnaI just -1'd his follow up patch16:06
hemnaremoving that section breaks running tests16:06
smcginnishemna: Oh, I didn't see a the follow up patch.16:06
*** Longgeek has quit IRC16:06
smcginnisSo we've been whitelisting tests for now. Maybe not needed once it is all compatible?16:07
hemnaI think there is some problem with testr16:07
*** asselin_ has quit IRC16:07
hemnait pukes running tox -epy34 standalone w/o that section16:07
hemnanova does it this way, as does cinder16:07
*** Longgeek has joined #openstack-cinder16:08
*** jistr has quit IRC16:08
hemnathat's what you get when you remove that section and run the tests16:08
hemnajust vomits every time.16:08
smcginnishemna: Not good.16:09
*** Vikash_cz has quit IRC16:10
hemnaand forcing folks to do rm -rf .testrepository && tox -epy34 is not acceptable to me16:10
smcginnishemna: Definitely. That would be very confusing.16:10
*** anshul has quit IRC16:11
*** Longgeek has quit IRC16:12
haypohemna: hey. "removing that section breaks running tests" that's strange. i'm unable to reproduce your issue16:13
haypohemna: "db type could not be determined" ah you missed my comment :) you must run py34 *before* running py2716:14
*** Apoorva has joined #openstack-cinder16:14
haypohemna: otherwise, .testrepository is incompatible with python 3. stupid testr bug16:14
hemnaI'm running tox -epy3416:14
*** avishay has joined #openstack-cinder16:14
haypohemna: this issue will not affect the check jobs which always start from a fresh checkout16:15
hemnafails every time16:15
haypohemna: try: rm -rf .testrepository16:15
hemnaI don't want people to have to do that16:15
haypohemna: you should write py34,py27 in tox.ini16:15
*** DericHorn-HP has quit IRC16:15
hemnathat's not right16:15
openstackgerritKyrylo Romanenko proposed openstack/python-cinderclient: Add tests for python-cinderclient
haypohemna: people will have not to do that :)16:15
smcginnisDoh! Meeting.16:15
hemnaI changed it to py34,py27...16:15
haypohemna: this issue is a bug in testr, it's unrelated to cinder or os-brick. all projects are affected16:15
hemnasame problem16:15
hemnasame failure16:15
haypohemna: first, remove .testrepositry16:16
*** DericHorn-HP has joined #openstack-cinder16:17
openstackgerritBharat Kumar Kobagana (BharatK) proposed openstack/cinder: GlusterFS: support extending a volume that has snapshots
hemnahaypo, ok py27 fails now for the same reason16:18
*** garthb_ has quit IRC16:18
*** garthb has quit IRC16:18
hemnaok yah, this doesn't work16:18
smcginnishaypo: Even if it's in the right order, wouldn't existing ones fail?16:19
smcginnishaypo: Everyone would have to know to delete that after pulling down the latest.16:19
*** harlowja_at_home has joined #openstack-cinder16:20
haypohemna: "ok py27 fails now for the same reason" ah? "No module named gdbm" is a different error16:20
hayposmcginnis: it doesn't break py27, it's just an annoying issue when running tox -e py3416:21
haypoi'm hitting this issuse since more than one year...16:21
haypohemna: your python 2.7 doesn't have the gdbm module?16:21
haypoon Fedora 22, "tox -e pep8,py34,py27" works well, and python 2.7 has the gdbm module16:22
smcginnishaypo: But that's my point. If someone already is working on things and has run py27 before, when they pull down the latest code and see python3 support is now available, it won't be obvious to them they need to delete anything before they are able to run it.16:22
haypogdbm is part of the Python stdlib...16:23
hayposmcginnis: i don't know how to advertise on this bug16:23
hayposmcginnis: i documented it at :)16:23
smcginnishaypo: Well, it sounds like the way hemna has it you won't need to.16:23
mriedemthat compares the rally failures to the check-tempest-dsvm-full job failures over the last 6 months16:23
hayposmcginnis: again, it's unrelated to os-brick. and it would be a shame to not use testr for this reason :-/16:23
smcginnishaypo: So if we don't remove that section it can just work.16:24
e0nemriedem: thanks16:24
e0nemriedem: now, we need to find why they failed16:24
hayposmcginnis: yes16:24
smcginnishaypo: Can't we still use testr though?16:24
mriedeme0ne: from you can drill into jobs and add stats to graphs via stats - zuul - pipeline - check - job16:24
mriedemthen pick jobs16:24
mriedemi used check queue since that's what the gate-rally-dsvm-cinder job is running in right now16:24
mriedemDuncanT: ^16:24
hayposmcginnis: testr relies on testrepository. testrepository has an annoying bug if you run py27 first16:25
haypoi ported or helped to port +20 projects, all of them are now using testr on python 316:25
haypoi'm surprised that it doesn't work on hemna host :-/16:25
e0nemriedem: we can't compare fails count w/o analyzing why it fails16:26
hemnaI'm trying again16:26
*** rushiagr_away has joined #openstack-cinder16:26
DuncanTmriedem: expanding the jobs list took our my browser :-) I'll try later when there's no meeting. Thanks for the pointer16:26
hemnahaypo, rm -rf .tox && rm -rf .testrepository && tox16:26
hemnawe'll see if it works16:26
haypohemna: i guess that you will hit the same error on gdbm16:26
haypohemna: is it a regular Ubuntu?16:26
haypoah, it'a a package:
smcginnisMy py3 on 14.04 does not have gdbm.16:27
smcginnispy2 does.16:27
hemnaok py27 puked again16:27
*** Vikash_cz has joined #openstack-cinder16:27
haypohemna: try: sudo apt-get install -y python-gdbm16:28
mriedeme0ne: well, if the rally job trends about the same with the jenkins job, then you could consider it fairly stable16:28
mriedeme0ne: if there were big mismatches, then it would be unstable16:28
mriedemsince jenkins is the baseline16:28
mriedemwell, check-tempest-dsvm-full is the baseline16:28
e0nemriedem: fair enugh16:28
mriedemit looks fairly consistent over 6 months16:28
haypohemna: ah, it's explicitly installed in DevStack for testr: files/apts/general:20:python-gdbm # needed for testr16:29
hemnawhy has it never failed before then16:30
* hemna is confused16:30
hayposmcginnis: gdbm was renamed to dbm.gnu in python 3 ;)16:30
hemnahaypo, so I just did another test16:30
hemnaI removed py34 from tox.ini16:31
hemnarm -rf .testrepository && rm -rf .tox && tox16:31
hemnapy27 worked16:31
smcginnisThere it is!16:31
*** Vikash_cz has quit IRC16:31
haypohemna: testr is fun :)16:31
haypohemna: it looks like depending on the python version, it uses a different dbm format16:31
*** tsbot_ has joined #openstack-cinder16:31
haypohemna: gdbm is you first run tox -e py3416:31
haypondbm is you first run tox -e py2716:32
haypoyou can check using: file .testrepository/times.dbm16:32
haypoi'm checking again. tox -e py27 first: "Berkeley DB (Hash, version 9, native byte-order)". tox -e py34 first: "GNU dbm 1.x or ndbm database, little endian, 64-bit"16:34
haypoit looks like python 3 is unabled to open "Berkeley DB (Hash, version 9, native byte-order)"16:34
*** BharatK has quit IRC16:34
haypowhere python 2 is able to open "GNU dbm 1.x or ndbm database ..."16:35
*** vilobhmm has joined #openstack-cinder16:35
*** harlowja_at_home has quit IRC16:37
*** yamada-h has joined #openstack-cinder16:37
*** jwcroppe_ has joined #openstack-cinder16:38
*** ociuhandu has quit IRC16:38
*** jwcroppe has quit IRC16:40
*** yamada-h has quit IRC16:42
hemnaso with the py34 section in place it all just works16:43
hemnaevery time.16:43
hemnaso I'm not removing it.16:43
haypohemna: i also understand that tests pass with testr (without the py34) section if you remove .testrepository and you install gdbm16:44
openstackgerritTom Swanson proposed openstack/cinder: Dell SC: Add support for consistency groups
hemnaif cinder and nova did it w/o the py34 testenv section, then I'd say fine, lets do it.  but both nova and cinder do exactly what my patch does.16:45
hemnait's something that can be revisited later16:45
haypohemna: the final goal is to remove the py34 section16:45
hemnafor now, I'd rather get this in, as it just works.16:45
*** Longgeek has joined #openstack-cinder16:45
haypohemna: i wrote the py34 section of cinder and nova because we cannot port them at once. it will require many weeks of work16:45
hemnaso we can get py34 validation in os-brick.16:46
haypofor os-brick, we are very lucky: the whole port can be done in a single commit!16:46
*** deepakcs has quit IRC16:46
hemnalets just do this for now.  it works.16:46
*** BharatK has joined #openstack-cinder16:48
*** annashen has joined #openstack-cinder16:48
e0newinston-d: could you take a look on new prerformance tests after the meeting or tomorrow, please16:49
haypohemna: no problem. i will probably a following patch to drop the py34 section :) it will tested by Jenkins ;)16:50
hemnahaypo, but if it fails locally, then what good is it ?16:51
hemnait simply doesn't work16:51
hemnaanyway, I'm over it.16:51
*** sgotliv has quit IRC16:52
*** annashen has quit IRC16:53
*** rmesta1 has joined #openstack-cinder16:53
haypohemna: you should fix your setup (install gdbm) and it should work16:53
haypoi know how these testr bugs are annoying16:53
haypobut i didn't choose testr :)16:53
*** rmesta has quit IRC16:54
*** aix has quit IRC16:55
*** belmoreira has joined #openstack-cinder16:55
*** bswartz has joined #openstack-cinder16:56
e0nethingee, jgriffith: one topic left from me in an agenda: Move check-cinderclient-dsvm-functional job from experimental queue to non-voting job17:00
e0neand others: ^^17:00
e0newhat do you think about this idea?17:00
*** greghaynes has joined #openstack-cinder17:00
smcginnise0ne: Seems like a good thing.17:00
smcginnise0ne: How have the experimental runs gone?17:00
geguileowinston-d: So that wip is not still up for review, right?17:00
thingeeso geguileo I'm confused. you asked earlier about those three patches solving ha?17:01
thingeeForce iSCSI disconnect after timeout:
winston-dgeguileo: sorry, it's not, it's on my local branch17:01
geguileothingee: No, not solving HA   XD XD17:01
thingeeRollback if attach_volume timesout:
e0nesmcginnis: it's successful after my fix was merged :)17:01
geguileothingee: Helping us solve Cinder/Nova interactions that block our HA work17:01
thingeeDetach and terminate conn if Cinder attach fails:
thingeethis list is so big17:01
geguileowinston-d: Ok17:01
e0nesmcginnis: tbh, it has only a very and very simple tests17:01
*** Longgeek has quit IRC17:02
geguileowinston-d: And with that patch, what else do we need to do before we can change Cinder?17:02
thingeegeguileo, winston-d hemna it would be good to know which I can help push through in nova17:02
dulekhemna: There's a lock on snapshot when creating a volume - so it won't get deleted during the procedure.17:02
dulekhemna: But I don'17:02
dulekhemna: I don't thing it's related to TaskFlow stuff. ;)17:02
winston-de0ne: hey, i've already read your new numbers, it's constantly showing apache has 50% peformance gain over eventlet, i'm convinced.17:02
MentalRaySomeone a POC of Kilo+Ceph up. We have a weird snapshot issue and curious to see if the issue is local to our setup or not.17:02
*** rajinir has joined #openstack-cinder17:02
winston-dthingee: will let you know once my patch is up.17:02
*** patrickeast_ has joined #openstack-cinder17:02
*** patrickeast has quit IRC17:03
e0newinston-d: thanks for pointing me when i was wrong!17:03
e0newinston-d: could you remove your -1 now?17:03
winston-de0ne: sure17:03
thingeeMentalRay: meet jbernard17:04
*** jordanP has quit IRC17:04
*** hemna is now known as hemnafk17:04
liuxgthingee: Huawei CI's problem have been solved now, please check it17:04
MentalRayyes, waiting to see if he can reproduce on his end17:04
MentalRaywill wait for him17:04
geguileoSo once Nova accepts VolumeBusy, what else do we need to do to be able to do changes to Cinder?17:04
*** lpetrut has joined #openstack-cinder17:04
winston-dgeguileo: what change?17:05
geguileowinston-d: Changes like atomic status change17:05
geguileowinston-d: Removing some write locks using status on DB17:05
geguileowinston-d: Try to get some kind of shared locks to allow multiple reading of the same resource while we prevent deletions17:06
*** patrickeast has joined #openstack-cinder17:06
geguileowinston-d: Move drivers to use a DLM17:06
winston-dDuncanT, hemnafk: any idea about geguileo's question?17:06
geguileoBecause if we don't know what we need to do for HA to advance it's going to be hard to get there17:06
geguileo(Or if only some people know)17:07
winston-dgeguileo: I can't think of any, but DuncanT and hemnafk they know better.17:07
geguileoDuncanT: mentioned that we couldn't simple change Cinder because of the clients17:07
*** patrickeast_ has quit IRC17:07
winston-dplus, it seems we haven't agreed on using DLM to solve the problem, yet, do we?17:07
geguileowinston-d: Ok, then lets call it driver locking17:08
geguileoSo we need to:17:08
geguileo1- Allow Nova to accept VolumeBusy exceptions17:08
jgriffithDuncanT: I get what you're saying "kinda"17:08
thingeeliuxg: I already spoke about this a while back on the ML and today's meeting
thingeeliuxg: please stop asking me17:08
jgriffithDuncanT: my first answer was "ok, just spawn a new one" but you also want to pick up tasks it was in the middle of17:09
geguileo2- Other stuff we don't know about before changing Cinder17:09
DuncanTjgriffith: spawning a new one means knowing for sure that the old one is dead, which means node fencing and pacemaker, which has a whole bunch of issues of its own17:09
dulekjgriffith: Oh, picking tasks is harder. I was able to do that with taskflow persistence patches, but we know whole taskflow story.17:10
openstackgerritBharat Kumar Kobagana (BharatK) proposed openstack/cinder: GlusterFS: Using 'fallocate' instead of 'dd' to create raw volumes
geguileo3- Change locks for resource writting (possible using status and atomic state changes)17:10
*** vilobhmm has quit IRC17:10
*** annashen has joined #openstack-cinder17:10
thingeeliuxg: if this not for new drivers, I will be check improvements to existing drivers when I get the chance.17:10
geguileo4- Shared locks for reading and preventing deletion while there's readers17:10
*** vilobhmm has joined #openstack-cinder17:10
geguileo5- Drivers locks17:10
jgriffithdulek: :)17:10
DuncanTgeguileo: Doing locking without changing the API contract is way harder than I thought... changign the contract with Nova is much easier than changing it with tenant facing APIs17:10
dulekDuncanT, jgriffith: Oh, that's right, it needs to fence it - we don't know if it isn't running and talking with backend.17:10
jgriffithdulek: DuncanT You could start simple....17:11
geguileoDuncanT: Any insights on how we can work on those APIs?17:11
jgriffithdulek: DuncanT turns "hosts" into a list.. if manager can't reach one, try the other17:11
jgriffithdulek: DuncanT that would at least solve DuncanT 's bad port last night17:11
dulekjgriffith: Still - you need to distributely lock.17:12
jgriffithdulek: DuncanT in progress task may have failed still of course, but at least user coudl ecover17:12
jgriffithdulek: I dont' think you do17:12
dulekjgriffith: For example RemoteFS patches cannot do snapshot in parallel.17:12
jgriffithdulek: primary/secondary17:12
dulekjgriffith: RemoteFS drivers.17:12
DuncanTjgriffith: you still need to know the the primary has stopped17:12
dulekjgriffith: So you need to fence primary on fail.17:12
jgriffithsecondary takes over IFF primary isn't able to be contacted17:12
dulekjgriffith: You need to be 100% sure.17:12
jgriffithDuncanT: yes, and that's simple17:12
jgriffithDuncanT: Scheduler can't communicate with it when it tries to send a request17:13
DuncanTjgriffith: 'can't be contacted' isn't enough, you need to fence (pull the power, worse case)17:13
dulekjgriffith: But you don't know if link died or whole server.17:13
liuxgthingee: thanks, this is not for new drivers, just for existing drivers17:13
jgriffithDuncanT: We can disagree all day :)17:13
DuncanTjgriffith: The scheduler does fire and forget17:13
dulekjgriffith: So maybe server is alive and able to contact backend - and messing with it.17:13
jgriffithdulek: but you can't solve that anyway17:13
*** aarefiev_home has quit IRC17:13
thingeeliuxg: ok great. I will remove the -2's after I have time to review the CIs...17:13
jgriffithdulek: that's my point17:13
jgriffithabout "in progress" things17:14
jgriffithat least at first pass.. who cares17:14
DuncanTjgriffith: So start a snap of a big volume, link goes down, another snap request for that volume comes in, goes to the second c-vol, boom, corruption17:14
dulekjgriffith: Ah, okay, I see. Link dies, distributed lock drops, you need fencing anyway.17:14
jgriffithyou fail those tasks, but user can re-issue and clean up17:14
liuxgthingee: Thanks thingee :)17:14
jgriffithDuncanT: no corruption17:14
jgriffithDuncanT: new snap, new UUID17:14
*** liuxg has quit IRC17:15
DuncanTjgriffith: Not if it is file backed...17:15
openstackgerritBharat Kumar Kobagana (BharatK) proposed openstack/cinder: RemoteFS: Reporting configured reserved_percentage in _update_volume_stats
jgriffithDuncanT: sounds like a problem with the choice of backends?17:15
avishaycan we please not not not not not add distributed locks?  the solution is to make things simpler, not more complex.17:15
jgriffithDuncanT: or at least impl of snaps in the backend17:15
jgriffithavishay: +117:15
*** IlyaG has joined #openstack-cinder17:15
DuncanTjgriffith: We have the backends we have17:15
avishaybackends should have the requirement of idempotent operations, then life becomes simpler17:16
jgriffithDuncanT: fair enough I guess, but maybe you could fix the impl in their snaps in the Cinder driver that's public?17:16
dulekavishay: Well, using tooz isn't hard actually. But you probably meant fact it's there.17:16
DuncanTjgriffith: I'm at least trying to push the locks down into just the drivers that need them... they can be cleaned up case by case then if anybody really cares17:16
jgriffithDuncanT: which driver?  I can look and see if I understand better17:16
geguileoavishay: We need distributed locks17:16
avishaydulek: yes, it's another thing to administer, it's another thing that can fail17:16
geguileoavishay: Be it using DB fields as locks17:17
geguileoavishay: Or an actual DLM17:17
avishaygeguileo: i am not convinced17:17
jgriffithDuncanT: I'm not all that familiar with f-based backends, but I can look through the driver17:17
DuncanTjgriffith: e.g. gluster17:17
dulekavishay: But it may default to using good old file locks. Configure it if you want it.17:17
geguileoavishay: Then let me ask you17:17
avishayi'm not convinced that cinder needs ANY locks17:17
geguileoavishay: How do you prevent some node from deleting the snapshot you are using to create a volume?17:17
jgriffithDuncanT: dulek geguileo So is all of this just about solving issues with Gluster?17:18
avishaygeguileo: garbage collect snapshots/volumes offline, and only when there is no current operation on them17:18
dulekjgriffith: Here's how to reproduce problems with running two c-vol A/A on gluster:
jgriffithor more generally file-based backends?17:18
DuncanTjgriffith: They seem to be the worst ones17:18
jgriffithdulek: DuncanT so I'm realizing this is mostly a Gluster discussion now17:18
geguileojgriffith: Gluster?17:18
DuncanTjgriffith: I seem to remember some backends don't like it either, but I guess they can poll in the driver17:18
geguileojgriffith: For me it's about getting Cinder to a point were I'm not ashamed of the races we have17:19
jgriffithdulek: thanks for the link17:19
DuncanTjgriffith: File based backends have the worst problems with races for sure17:19
*** e0ne is now known as e0ne_17:19
avishaythere are more complex systems that don't have any locks, just requires some more thought and creativity17:19
*** patrickeast has quit IRC17:19
winston-davishay: +100 for idempotent17:19
dulekjgriffith: I think all RemoteFS drivers share the issues.17:19
jgriffithDuncanT: well... I guess I just don't understand how the corruption happens17:19
jgriffithDuncanT: "most" backends you say 'create snapshot' it creates a snapshot and it's independent17:19
geguileoavishay: And how do you know there is no operations on them?17:20
jgriffithI'm not trying to think about resuming tasks etc17:20
dulekjgriffith: But hemna once told that FC drivers suffer similar issues if I recall correctly.17:20
avishaynovel idea - maybe block storage management shouldn't be implemented on file systems17:20
*** e0ne_ is now known as e0ne17:20
DuncanTavishay: That boat has sailed17:20
winston-ddulek: they said using taskflow isn't hard17:20
eharneycinder isn't block storage management.  it's block storage for instances.17:20
*** Yogi1 has joined #openstack-cinder17:20
jgriffithdulek: data corruption?17:20
avishayeharney: i am working on a solidfire driver for manila as we speak BTW17:20
avishaygeguileo: you can check if there is a volume has the snapshot as a source and is in "creating" state17:20
jgriffithDuncanT: dulek hemnafk so let me see if I got this right....17:21
dulekwinston-d: :D It isn't. ;)17:21
*** lpetrut has quit IRC17:21
geguileoavishay: So you block the whole table to checkk all volumes at the same time?17:21
jgriffithDuncanT: dulek hemnafk what you're saying is that you have some backend devices that can't do things like more than one call to a single resource at a time?17:21
DuncanTjgriffith: Yes17:21
jgriffithDuncanT: dulek hemnafk without corrupting "something"17:21
jgriffithDuncanT: I have NO sympathy17:21
*** patrickeast has joined #openstack-cinder17:21
jgriffiththat's stupid17:22
jgriffithand they don't belong in a cloud17:22
jgriffitheven LVM can handle that17:22
avishaygeguileo: why block the whole table?  once a snapshot is in "deleted" state you can't create from it.  so the amount of volumes being created from it after it is deleted should go down with time.17:22
jgriffithwhich everybody likes to talk badly about :)17:22
avishaygeguileo: just scan periodically and delete when the number is 017:22
*** jaypipes has joined #openstack-cinder17:22
geguileoavishay: Because you need to check no volume is in creating state17:22
jgriffithDuncanT: you inparticular over the years have often said "I don't care, your device doesn't belong in a cloud then"17:22
avishaygeguileo: no, you need to check no volume with that particular source is in creating state17:23
geguileoavishay: And lock the whole table so no one starts using it17:23
jgriffithI don't know why a devices inability to be "used" in a modern data center becomes a Cinder problem to work around17:23
avishaygeguileo: nobody can create a volume from a deleted source17:23
jgriffithkinda like lipstick on a pig... it's still a pig :)17:23
jgriffithfix it in the driver IMHO17:23
geguileoavishay: Oh, so you have volumes being created from a deleted source?17:24
*** dannywilson has quit IRC17:24
winston-ddulek: you have been creative to fix 'allocated_capacity' bug with taskflow foo, you think it's simple?17:24
DuncanTjgriffith: We can, I think, with a bit of effort, push all of the mutual exclusion down into the drivers17:24
* jgriffith now understands why some people put locks everywhere :)17:24
geguileoavishay: So we'll have now a limbo status?17:24
jgriffithDuncanT: frankly that's where it belongs don't you think?17:24
DuncanTavishay: Delayed operations cause chaos17:24
avishaygeguileo: i have no idea what you mean17:24
dulekwinston-d: There will be a functionality now which makes it easier. :)17:25
geguileoavishay: They are being used, therefore they exist17:25
DuncanTavishay: Sorry, meant to say 'quota chaos'17:25
avishayDuncanT: no, delayed operations are the best way of garbage collection17:25
geguileoavishay: But they are deleted17:25
geguileoavishay: And cannot be used in the meantime for new operations17:25
avishayOK, let's start again17:25
geguileoavishay: Even though there are current operations that can used them17:25
DuncanTavishay: No, they aren't. We tried that.17:25
geguileoavishay: So any operation that would do a retry would probably not be able to find the resource17:26
eharneyso we just had some discussion in the meeting about NFS snapshots17:26
DuncanTjgriffith: I agree we should push this down into the drivers. Coming up with a sane & safe mechanism to do so is proving tricky17:26
avishayI have snapshot A, and create volumes 1, 2, and 3 from it.  While they are being created, i delete snapshot A.  now i have a STATELESS background process that says "let's find all the deleted snaphots that i can delete right now"17:26
eharneywe aren't going to implement that and then have folks yelling about it not belonging in the cloud afterward, right?17:26
avishayand then it finds snapshot A and says "oh, this has 3 volumes being created from it, i can't delete it now"17:27
dulekjgriffith: Apart from drivers locking on for example snapshot when creating a volume is needed. But that may be solved by atomic state changes.17:27
DuncanTavishay: But it is still messing up your quota17:27
avishayand then after 20 minutes it finds snapshot A again and says "oh boy, no volumes being created from snapshot A, this is the best day of my life - delete"17:27
dulekjgriffith: Only problematic thing would be when c-vol will die during such creation - who will drop the status of a snapshot back to normal?17:27
DuncanTavishay: You get users who do long chains of snap & delete and have to squash, which causes I/O spikes17:27
avishayDuncanT: no, the quota gets update as soon as the delete happens17:27
DuncanTavishay: The API delete? But the space is still in use (potentially indefinitely) on your backend17:28
avishayno, it will be deleted once active operations on it finish17:28
*** yuriy_n17 has quit IRC17:28
jgriffithYou guys are all crazy :)17:29
avishayOK fine, add distributed locks, add states in the DB, just write the entire volume to the DB and be done with it17:29
geguileojgriffith: +117:29
*** e0ne is now known as e0ne_17:29
winston-davishay: mysql vol driver for cinder? :)17:29
jgriffithDuncanT: sorry :)17:30
jgriffithwinston-d: stop that right now!17:30
avishaywinston-d: yes, why not?17:30
DuncanTSo we kick most of the drivers out of cinder and rewrite it to just support a couple of backends, because engineering is hard?17:30
avishaywinston-d: make one for manilla too :)17:30
jgriffithDuncanT: that's hardly what anybody is saying here17:30
jgriffithDuncanT: I'm certainly not saying that17:30
*** IlyaG has quit IRC17:30
geguileoDuncanT: XD  +117:30
DuncanTjgriffith: I kind of is what you said17:30
eharneyi kinda hear what DuncanT said when people start going on about things not belonging in the cloud17:31
geguileoDuncanT: Sounds like a plan  ;-)17:31
jgriffithDuncanT: just a sec, phone call17:31
DuncanTjgriffith: I actually think there might be a space for a 'cinder++' blue sky codebase, but that is totally different to fixing the problems in the current one17:31
winston-dyou know what, i remember the room when Nova people talked about adding cell to solve scaling problem.17:31
DuncanTcells still suck...17:32
jgriffithDuncanT: you keep throwing out all sorts of things so I have no idea what to think about17:32
*** e0ne_ is now known as e0ne17:32
winston-dsome people said, if it has scaling issue, let's fix it, not work around it.17:32
DuncanTjgriffith: Saying 'that doesn't belong in the cloud' to 30% of out drivers is, to my mind, rather unhelpful at this point17:33
jgriffithwinston-d: +117:33
avishayDuncanT: I would say 95% :-P17:33
*** BharatK has quit IRC17:33
winston-dquestion to us now is: is introducing DLM fixing the problem or just working around it?17:34
DuncanTavishay: I'd suggest you guys would be better writing an api-compatible re-write, but still....17:34
jgriffithDuncanT: avishay Really?17:34
avishaybut seriously, i think with atomic state changes, and much fewer states, we can get a recoverable system with no distributed locks17:34
DuncanTatomic state changes *are* a form of distributed lock17:34
jgriffithWe have that many drivers that you can't send two API calls to at the same time without them falling over?17:34
dulekjgriffith: I honestly don't understand why it shouldn't be an option in Cinder if it may solve some of the problems people have. It's completely optional the way we're designing it.17:34
avishayDuncanT: i would love to, but then i have no drivers17:34
jgriffithdulek: which?17:34
*** dtynan has quit IRC17:34
DuncanTjgriffith: All of the (many) file based ones.... that's a third of the drivers17:35
jgriffithDuncanT: I'm so confused17:35
*** dtynan has joined #openstack-cinder17:35
jgriffithbut ok17:35
DuncanTjgriffith: A third of our drivers or so are file based I think17:35
avishaywell we can say that those are not HA-ready, and tough cookies17:35
dulekjgriffith: DuncanT has problems. ;) I'm not running any setup.17:35
*** e0ne has quit IRC17:35
jgriffithDuncanT: I guess if you and dulek implement tooz or something that would be cool17:35
DuncanTjgriffith: My math might be a bit out, but it is not a small number17:35
winston-ddulek: nova cell is optional, but people get tricked into using it and hard to get away from it.17:36
jgriffithDuncanT: well when you have 60 to choose from I don't know about 33%17:36
jgriffithlet alone avishay 95% :)17:36
DuncanTjgriffith: I'd love to push all the locks down into just the drivers that need them... I considered proposing we just push the current file locks down into *all* the driver, then the ones that don't need them can remove them17:36
avishayDuncanT: do you really want to administer tooz as well?17:36
dulekwinston-d: Hm, probably clear documentation on that would be helpful...17:36
jgriffithI obviously don't "get it" because I don't see the problem in LVM or any of the other things that I focus on17:36
jgriffithso I should leave it to you and others that seem to have a really good handle on the problem17:37
avishayi think we can remove that ONE lock from the manager.  if a driver has its own locks, then the documentation says it can't be run active-active.  end of story.17:37
DuncanTavishay: I already have to administer rabbit, mysql, backends, horizon, load ballancers, rate limiters... adding redis isn't going to make it that much harder17:37
avishayDuncanT: in that case, i have a few ideas to make it more complex and fragile :)17:38
DuncanTjgriffith: Or at least wait until we have code before you complain... I think you'll be not too unhappy with the end result, just hard to get there17:38
*** belmoreira has quit IRC17:38
avishayOK I'm done for tonight, bye all, was a pleasure :)17:38
DuncanTavishay: Adding DLM might make things less fragile, not more. Atomic ops in the DB are painful to the DB... you can't do reader/writer with 'update where' as far as I  can tell17:39
dulekavishay: bye! :)17:39
*** lcurtis has quit IRC17:39
DuncanTFuck, I've just snapped the power lead on my laptop :-(17:39
avishayDuncanT: maybe SQL was the wrong choice for a distributed management service then ;)17:39
dulekDuncanT: You cannot with tooz in current state tooz. ;)17:39
geguileoavishay: Bye17:39
*** avishay has quit IRC17:40
jungleboyjDuncanT: Yikes.  That sucks!17:40
winston-dDuncanT: really? unless you already have someone who's familiar with redis or other nosql stuff, otherwise, getting someone one to operate one new stuff they don't familiar with prove to be ...17:40
jgriffithDuncanT: that's fair... but FTR I wasn't "complaining" I was just asking to better understand the issue17:40
jgriffithwhich I "kinda" do now17:40
dulekwinston-d: Well, tooz can use even MySQL as locking backend. It's not too reliable, but it's a possibility. ;)17:41
*** ndipanov has quit IRC17:41
winston-ddulek: can you quantatively measure 'how not too reliable' it is?17:41
geguileoAnd tooz is only a thin abstraction layer17:41
DuncanTdulek: I don't actually care if we use tooz or not... if it gives us what we need, great, if not we can do something else17:41
geguileoDuncanT: +117:42
dulekDuncanT: Right, I'm getting locked on library, which isn't needed here.17:42
greghaynesYes - do not use the sql DB for distributed locks - this really breaks down into deployment-specific cases even and it super hard to make it work at all17:42
*** flaper87 has joined #openstack-cinder17:43
DuncanTRight, it's time for me to bale. I'll try to write up some of the discussion from tonight, there's been some food for thought.17:43
*** marcusvrn1 has joined #openstack-cinder17:43
greghaynes(e.g. galera vs single node mariadb, etc)17:43
DuncanTgreghaynes: Having it as an option is nice for POC deploys17:43
greghaynesSure, so if you want that kind of flexibility then definitely use tooz, but there should be a big red flag of "this is not for prod"17:43
DuncanTgreghaynes: Particularly if you can do it with 'update... where' - that has to work on all backends17:44
*** flaper87 has left #openstack-cinder17:44
greghaynesYes, youre going to turn all your reads into writes and poll on failures ;)17:44
DuncanTgreghaynes: 'This will suck for performance above 4 nodes' is a perfectly fair warning label to put on it, sure17:44
dulek - you may even use RabbitMQ. ;)17:44
xyangDuncanT: hi, can you take a look of the non disruptive backup patch again?  Addressed your comments.17:45
*** thangp is now known as thangp_afk17:45
DuncanTxyang: Sure, it'll be a while though, I need to clear my head17:45
xyangDuncanT: sure, thanks17:45
DuncanTxyang: If it isn't done by tomorrow, please shout at me17:46
xyangDuncanT: ok:)17:46
*** marcusvrn has quit IRC17:46
*** lcurtis has joined #openstack-cinder17:46
*** vilobhmm has quit IRC17:47
*** winston-d is now known as winston-d_zZZ17:48
*** rwsu has quit IRC17:50
*** hemnafk is now known as hemna17:53
hemnajgriffith, I don't believe I ever said that.17:53
jgriffithhemna: yeah, I dunno... was just going off what DuncanT and others said17:53
hemnajgriffith, we removed all of our locks on purpose because our backends can do more than one operation17:53
jgriffithhemna: yeah, I'm not the one who said that17:54
jgriffithhemna: scroll back and read, don't blame me :)17:54
hemnabut the problem is that you can't run multiple c-vol servicing the same backend right now in cinder because of the local file locks in the manager.17:54
hemna"jgriffith> DuncanT: dulek hemnafk what you're saying is that you have some backend devices that can't do things like more than one call to a single resource at a time?"17:54
hemnaI was replying to that one.17:54
*** rwsu has joined #openstack-cinder17:54
jgriffithhemna: and scroll back17:54
jgriffithhemna: they invoked your name saying "this is true for FC devices as well according to hemna "17:55
hemnaremoving the local file locks in the volume manager is a good goal to have IMHO17:55
*** dtynan has quit IRC17:55
*** DericHorn-HP has quit IRC17:55
hemnait's not my/nor Cinder's problem if a particular vendor's driver has local file locks.17:55
*** dtynan has joined #openstack-cinder17:55
hemnabut Cinder itself shouldn't have them IMHO17:56
jgriffithhemna: I'd agree with that probably17:56
hemnaand if Cinder needs locks, they shouldn't be local file locks, but that's a separate discussion I thinks.17:56
*** lpetrut has joined #openstack-cinder17:56
jbernardmtanino: heya, re live migration17:56
hemnaso, we can do something in the mean time to help remove those local file locks in the volume manager17:56
jbernardmtanino: i need to decouple terminate connection from remove export17:57
hemnaby 1) getting Nova to expect and handle VolumeIsBusy exceptions coming back from Cinder.17:57
jbernardmtanino: so that they can be called individually from the context of a migration17:57
jbernardmtanino: else, live migration will suffer17:57
hemna2) put 'ing' checks in each of the cinder API calls, and return VolumeIsBusy.17:57
jbernardmtanino: is that accurate?17:57
*** IlyaG has joined #openstack-cinder17:58
jgriffithhemna: yes, and IMHO those things you list solves at least 80% of the problem areas for us I think17:58
hemnaso whenever 'HA' discussions come up, I always point to that first17:58
jgriffithhemna: but as I mentioned to DuncanT and others, I don't fully understand the challenges they have in their envs17:59
hemnaand folks get bogged down in locking mechanism talks seems pointless to me.17:59
hemnatooz vs. whatever vs. sucksslesslocks17:59
*** Yogi1 has quit IRC18:00
jgriffithhemna: we should write a sucksslesslocks lib18:02
*** lpetrut has quit IRC18:03
*** DericHorn-HP has joined #openstack-cinder18:03
*** delatte has quit IRC18:04
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** cdelatte has joined #openstack-cinder18:05
jgriffithhemna: thingee BTW, I traced down the issue with my CI system...18:05
jgriffithhemna: thingee it turns out that "somewhere" port 3260 is being blocked on that entire rack18:05
*** cdelatte has quit IRC18:06
thingeejgriffith: glad to hear you figured it out!18:06
*** cdelatte has joined #openstack-cinder18:06
jgriffiththingee: yeah... now if I could just figure out exactly what was done and by whom18:06
jgriffiththingee: in the meantime I'm building a new cloud in a different lab :)18:06
jgriffiththingee: It's a bummer, because I just hit 275 consecutive succesful runs :(18:07
jgriffithnow I have to start over18:07
thingeea new cinder ci record probably!18:07
thingeesetting the bar high there for us :D18:08
jgriffithwell to be fair that doesn't count anything that Jenkins failed18:08
mtaninojbernard: hi18:09
mtaninojbernard: Correct "so that they can be called individually from the context of a migration, else, live migration will suffer"18:10
xyangjgriffith: hi, here is a fix to the VNX unit test problem you ran into earlier:   can you take a look?18:11
openstackgerritMarcus V R Nascimento proposed openstack/cinder: Fix HBSD horcm driver with oslo.concurrency 2.1.0
*** zongliang has quit IRC18:12
jgriffithxyang: oh.. awesome!18:13
xyangjgriffith: thanks18:13
jgriffithxyang: I'd just +2/A it but I will get scolded (again) so just a +218:14
xyangjgriffith: sure:) thanks!18:14
*** thangp_afk is now known as thangp18:14
jgriffithOMG IR SO DUMB18:16
openstackgerritMarcus V R Nascimento proposed openstack/cinder: Fix HBSD horcm driver with oslo.concurrency 2.1.0
hemnajgriffith, doh, well that's a good find then.  those are hard to find.18:19
jgriffithhemna: annoying18:20
jgriffithhemna: I realized after seeing 12 in a row that it was def my env18:20
*** jungleboyj has quit IRC18:22
*** thangp has quit IRC18:23
*** e0ne has joined #openstack-cinder18:23
*** e0ne is now known as e0ne_18:23
*** e0ne_ is now known as e0ne18:23
*** e0ne has quit IRC18:23
*** e0ne has joined #openstack-cinder18:24
*** thangp has joined #openstack-cinder18:28
*** Yogi1 has joined #openstack-cinder18:30
dulekhemna: Sorry for this FC-drivers locks confusion, I've either misunderstood you or haven't remembered correctly.18:30
hemnano worries.18:30
hemnaI'm not worried about it.  I'd just like to see some more folks working on known problems instead of adding new features.18:31
*** Nikolay_St has quit IRC18:31
*** dannywilson has joined #openstack-cinder18:33
*** Nikolay_St has joined #openstack-cinder18:33
*** e0ne is now known as e0ne_18:34
*** e0ne_ is now known as e0ne18:37
*** Lee1092 has joined #openstack-cinder18:37
openstackgerritxing-yang proposed openstack/cinder: EMC VMAX Modify CG
*** xyang has quit IRC18:37
*** Rockyg has joined #openstack-cinder18:41
*** juzuluag_ has quit IRC18:42
*** xyang has joined #openstack-cinder18:43
*** mutoulbj has quit IRC18:46
e0neeharney: tested rbd patch with cuncorrency and high load: it looks good18:49
e0neeharney: i'll make review request to get some feedback from you and community18:49
*** julim has quit IRC18:51
*** Yogi1 has quit IRC18:52
*** hodos has joined #openstack-cinder18:52
*** rajinir has quit IRC18:52
*** rajinir has joined #openstack-cinder18:53
*** dtynan has quit IRC19:00
*** dtynan has joined #openstack-cinder19:01
*** rajinir has quit IRC19:02
*** rajinir has joined #openstack-cinder19:03
eharneye0ne: ok i'll check it out19:04
e0neeharney: - the second item19:04
openstackgerritPatrick East proposed openstack/cinder: WIP generic image cache
openstackgerritIvan Kolodyazhny proposed openstack/cinder: Fix block eventlet threads on rbd calls
patrickeastjgriffith: hey, not sure if you had time to mess with that image cache poc code or not, but i’ve got it mostly working pretty well now… i did however realize i overlooked something important, it only works for public images19:08
patrickeastjgriffith: i’m trying to figure out a nice way to have it work for non-public ones… but the only ways i’m seeing involve some ninja moves with the contexts and/or cinder database :(19:09
*** erlon has quit IRC19:10
*** DericHorn-HP has quit IRC19:11
mriedemtbarron: xyang: fyi
mriedemthingee: jgriffith: DuncanT: hemna: ^19:13
mriedemper the cinder meeting today on encrypted volume drivers19:13
*** boris-42 has joined #openstack-cinder19:13
*** reed_ has joined #openstack-cinder19:15
*** blmartin has quit IRC19:16
*** garthb has joined #openstack-cinder19:20
*** garthb_ has joined #openstack-cinder19:20
* morganfainberg waves at the cinder-type people cause... it's that time of day to wander through IRC channels.19:22
*** zhenguo has quit IRC19:23
*** DericHorn-HP has joined #openstack-cinder19:24
ericksonsantosmorganfainberg, o/19:24
*** rajinir has quit IRC19:25
*** thangp has quit IRC19:26
*** rajinir has joined #openstack-cinder19:26
*** IlyaG has quit IRC19:28
*** annashen has quit IRC19:29
*** Apoorva has quit IRC19:29
*** dansmith has quit IRC19:30
*** Apoorva has joined #openstack-cinder19:30
*** annashen has joined #openstack-cinder19:30
*** dansmith has joined #openstack-cinder19:30
*** rajinir has quit IRC19:32
jbernardmtanino: got it, thank you.  i should have an update soon19:33
*** rajinir has joined #openstack-cinder19:33
mtaninojbernard: Sure. thank you for your understanding.19:33
*** lpetrut has joined #openstack-cinder19:34
*** annashen has quit IRC19:34
*** Rockyg has quit IRC19:35
*** ociuhandu has joined #openstack-cinder19:36
jgriffithpatrickeast: hey19:36
jgriffithpatrickeast: I had some things I wanted to talk to you about on that19:36
jgriffithpatrickeast: go a minute now?19:37
*** e0ne has quit IRC19:38
thingeejgriffith: hey did you get my message yesterday about helping me with the scheduler part of the capability spec?19:38
jgriffiththingee: I did not19:38
ericksonsantosI'm new to openstack and I've been helping vilobh with nested quota driver for a while.19:39
thingeejgriffith: since you have that more figured out, would you mind updating my spec with that part? I'm on my phone so I can't pull the link at the moment.19:39
ericksonsantosthis may be a dumb question, but...19:39
jgriffiththingee: haha19:39
jgriffithI may not get to it til a bit later19:39
ericksonsantosdoes someone know when to use 'cinder quota-class-update' instead of 'cinder quota-update'?19:39
*** yamada-h has joined #openstack-cinder19:39
jgriffithbut I can have it by lunch tomorrow at the latest if that's ok?19:40
thingeeThat's fine. That'll be faster than me going through revisions.19:40
openstackgerritTom Swanson proposed openstack/cinder: Dell SC: Add support for consistency groups
thingeejgriffith: appreciate it19:40
*** nkrinner has joined #openstack-cinder19:40
jgriffiththingee: no problem19:41
ericksonsantoson cinderclient19:42
*** yamada-h has quit IRC19:44
*** mutoulbj has joined #openstack-cinder19:47
patrickeastjgriffith: hey, sry was afk19:47
jgriffithpatrickeast: hey19:48
jgriffithpatrickeast: so here's the thing... IIUC19:48
*** ctina_ has joined #openstack-cinder19:48
jgriffithpatrickeast: this still requires the bulk of the impl in the driver via "clone_image"19:48
jgriffithpatrickeast: but adds a whole mess of things on top to try and keep records of things right?19:49
jgriffithpatrickeast: so what if we just steppe back and made it super simple....19:49
patrickeastjgriffith: uhh no, it shouldn’t need clone_image at all19:49
patrickeastjgriffith: it only runs this if clone_image didn’t do anything19:49
jgriffithpatrickeast: ummm... well then I absolutely can't get it to work :)19:49
patrickeastjgriffith: haha, so i was wondering how it would play with your drivers cache stuff… its very possible i broke that19:50
patrickeastjgriffith: because you do implement clone_image19:50
jgriffithpatrickeast: well, I turned my stuff off19:50
patrickeastjgriffith: it works great on mine which doesn’t implement it19:50
patrickeastjgriffith: ah ok, something else then19:50
jgriffithso it is a noop19:50
patrickeastjgriffith: any chance you could send/paste a log from c-vol when it should be using it?19:51
jgriffithpatrickeast: but I could not get it to create the second volume in any case19:51
patrickeastjgriffith: hmm thats not good19:51
jgriffithpatrickeast: well.. yesterday I could've... today not so much :(19:51
jgriffithpatrickeast: my lab is officially *dead*19:51
jgriffithI'm trying to get everything back up, but stuck waiting for IT to figure out what they did to the network19:51
jgriffithit started with just a single port.. then all ports19:52
jgriffiththen pooof!19:52
patrickeastoh geez19:52
jgriffithand the machine running your code is a VM in that lab19:52
*** ctina has quit IRC19:52
*** Rockyg has joined #openstack-cinder19:52
patrickeastjgriffith: haha, well, assuming that the code doesn’t need to use clone_image, were there any other concerns?19:53
*** ctina_ has quit IRC19:53
patrickeastjgriffith: and ignoring the whole not working thing19:53
*** mutoulbj has quit IRC19:54
*** lpetrut has quit IRC19:56
jgriffithpatrickeast: well, trying to get it to work :(19:56
jgriffithpatrickeast: which I was unsuccesful in doing sadly19:56
patrickeastjgriffith: if you want to take another stab at it i would recommend pulling the latest patchset on the review… it fixed a ton of little things19:57
patrickeastjgriffith: but i am interested in your thoughts ‘super simple’ approach though19:57
patrickeasts/thoughts/thoughts on/19:59
jgriffithpatrickeast: so what I thought was leave it in the clone_image for now20:03
jgriffithpatrickeast: but have it return and populate the database with the appropriate info20:03
jgriffithpatrickeast: that way all the crazy lookups and translations go away20:03
jgriffithpatrickeast: and it all goes in manager, nothing in flow or API20:04
*** DericHorn-HP has quit IRC20:04
patrickeastjgriffith: maybe i’m missing something, i thought clone_image was a volume driver method, how would it go to the manager20:04
jgriffithpatrickeast: so the manager calls it, and we could give it a response20:05
jgriffithpatrickeast: then subsequent calls just check the DB20:05
patrickeastjgriffith: should we put it in the BaseVD impl of it?20:05
jgriffithpatrickeast: so everything is there already.20:06
jgriffithpatrickeast: so here's the problem I had with the generic and I think we still have20:06
patrickeastjgriffith: clone_image is called from the flow and calls the driver directly20:06
jgriffithpatrickeast: some devices will need to use snaps, some clones.. some maybe something else20:06
patrickeastjgriffith: right, thats why it just does a create volume from image, and create from source20:06
patrickeastjgriffith: and lets the drivers do whatever they need20:06
patrickeastjgriffith: all drivers can do those actions, right?20:07
* patrickeast likes to make grand assumptions20:07
jgriffithpatrickeast: hehe20:07
jgriffithpatrickeast: well that's a 'good' assumption probably20:07
*** DericHorn-HP has joined #openstack-cinder20:08
patrickeastjgriffith: so i like the idea of cutting out as much as we can, imo more simple = more better20:08
jgriffithpatrickeast: and you're right.. sorry nothing goes through manager currently20:09
patrickeastjgriffith: and i think we can do some more database magic in the managers flow to avoid calling back down through the api20:09
jgriffithpatrickeast: yeah, that's what I'm wondering as well20:09
patrickeastjgriffith: i guess the trade off is duplicating any of the create_volume api flow that already does the db/quota stuff in the manager create_volume flow20:09
jgriffithpatrickeast: so really if I get the thing to work maybe I'll think what you have is the best thing since sliced bread :)20:09
patrickeastjgriffith: haha20:10
patrickeastjgriffith: i do need to figure out the whole private image thing though…20:10
patrickeastjgriffith: i’ve been banging on it a little bit on a local vm and that seems to be the only big issues i found20:11
*** belmoreira has joined #openstack-cinder20:11
patrickeastjgriffith: right now its trying to create the volume from image using the internal tenant context, which doesn’t work if the internal tenant can’t access the image from glance20:11
*** daneyon has joined #openstack-cinder20:11
patrickeastjgriffith: i didn’t notice it because tempest was happy and i was only using public images on my devstack : /20:12
*** belmoreira has quit IRC20:15
*** rhagarty has joined #openstack-cinder20:17
*** rhagarty_ has joined #openstack-cinder20:17
openstackgerritSean McGinnis proposed openstack/cinder: Remove useless logging from unit tests
*** e0ne has joined #openstack-cinder20:18
*** garthb has quit IRC20:21
*** cburgess_ has quit IRC20:23
*** cburgess has joined #openstack-cinder20:23
*** marcusvrn1 has quit IRC20:28
*** Yogi1 has joined #openstack-cinder20:28
tbarronmriedem: thanks for all the good stuff on the bugs I raised!  I've been idle this afternoon as my family is under the impression that I'm on vacation this week and had some plans ...20:34
mriedemtbarron: np20:35
mriedemjust lock yourself in the bathroom20:35
tbarronmriedem: :-)20:35
*** earlephilhower has quit IRC20:35
*** earlephilhower has joined #openstack-cinder20:36
*** openstackgerrit has quit IRC20:37
tbarronmriedem: I think the 3 retries may work.  There isn't really anything different up to the point of the 'device  busy' excetption between our iSCSI logs and, say, LVM.20:37
tbarronI downloaded the cryptsetup code and didn't find anything enlightening.20:37
*** openstackgerrit has joined #openstack-cinder20:37
tbarron/dev/sdh is not a valid LUKS20:38
tbarrondevice; formatting device for first use20:38
tbarronThe successfull iscsi CI logs that I looked at also have the 'not a valid LUKS device;20:39
tbarronformatting device for first use' msg.20:39
tbarronI think that is expected.20:39
mriedemyeah, it's info level20:39
*** VManiac16 has joined #openstack-cinder20:39
tbarroncinder doesn't really do much here, nova (with providers) does all the formatting, encrypting, etc.20:39
tbarronWell, someone is knocking on the bathroom door :-)  Let's get the retry fix merged as it is harmless and run CI again on your main fix.20:41
tbarronmriedem: thanks again!20:41
*** nkrinner has quit IRC20:44
*** DericHorn-HP has quit IRC20:45
*** annashen has joined #openstack-cinder20:46
*** daneyon has quit IRC20:47
*** mriedem has quit IRC20:49
*** kjnelson has left #openstack-cinder20:53
*** laudo has joined #openstack-cinder20:53
*** Lee1092 has quit IRC20:55
VManiac16hemna: Are you around?20:56
hemnahere and there20:56
VManiac16Just curious if the HP XP7 CI was down.20:56
hemnano idea, I don't watch the XP CI20:57
hemnaasselin, where is nestor ?20:57
tsekiyamahemna: Hmm, your nova patch for os-brick switching seems hitting requirements issue... It looks like Jenkins gate-nova-requirements hate test-requirements-py3.txt20:58
VManiac16hemna: No need to bother asselin20:59
*** marcusvrn has joined #openstack-cinder21:00
*** jasondotstar has joined #openstack-cinder21:01
hemnatsekiyama, did it hit that again?21:02
hemnalast time I looked at the log and it didn't really say what the problem was21:02
hemnaVManiac16, well, I don't watch the CI systems, asselin does for us.21:02
VManiac16hemna: Ok ... thanks.21:03
tsekiyamahemna: yeah,, it is hidden until you hit the "Toggle CI" button on the gerrit review page..21:04
openstackgerritDiem Tran proposed openstack/cinder: Fix PEP 476 and debug format messages of Oracle ZFSSA Cinder drivers
tsekiyamahemna: My glance patch is hitting the same problem. I'm trying to look into it21:06
*** e0ne has quit IRC21:08
asselinVManiac16, hemna he's on vacatino21:10
VManiac16asselin: Right ... lounging in So. Cal.21:11
laudowhat is the default size of an IOP when creating  read_iops_sec qos for a cinder backend?21:14
*** IlyaG has joined #openstack-cinder21:21
*** marcusvrn has quit IRC21:23
*** edmondsw has quit IRC21:28
jgriffithlaudo: IOP is an Io Per Second21:29
*** jasondotstar has quit IRC21:30
*** Yogi1 has quit IRC21:37
*** IlyaG has quit IRC21:44
*** Yogi1 has joined #openstack-cinder21:45
*** Yogi1 has quit IRC21:52
*** dannywilson has quit IRC21:55
*** dannywilson has joined #openstack-cinder21:56
*** annegentle has joined #openstack-cinder22:02
*** MentalRay has quit IRC22:09
*** chlong has joined #openstack-cinder22:10
*** thingee has quit IRC22:11
*** thingee has joined #openstack-cinder22:11
*** thingee has quit IRC22:11
*** thingee has joined #openstack-cinder22:11
openstackgerritPeter Penchev proposed openstack/cinder: Implement the update_migrated_volume for the drivers
*** dims_ has joined #openstack-cinder22:17
*** annashen has quit IRC22:20
*** dims has quit IRC22:21
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** annashen has joined #openstack-cinder22:21
laudojgriffith: So there is no size in kb for iops? its per sec only?22:23
*** annashen has quit IRC22:25
*** chlong has quit IRC22:30
*** chlong has joined #openstack-cinder22:30
*** lcurtis has quit IRC22:40
*** VManiac16 has quit IRC22:48
*** hodos has quit IRC22:57
*** dannywilson has quit IRC23:02
*** dannywilson has joined #openstack-cinder23:03
*** jasondotstar has joined #openstack-cinder23:09
*** annegentle has quit IRC23:10
*** DericHorn-HP has joined #openstack-cinder23:13
*** zhenguo has joined #openstack-cinder23:14
*** vokt has joined #openstack-cinder23:15
*** mutoulbj has joined #openstack-cinder23:22
*** ctina_ has joined #openstack-cinder23:22
*** IanGovett has quit IRC23:30
*** annegentle has joined #openstack-cinder23:34
*** lixiaoy1 has joined #openstack-cinder23:37
*** ctina_ has quit IRC23:41
*** mutoulbj has quit IRC23:46
*** cebruns has quit IRC23:49
*** DericHorn-HP has quit IRC23:50
*** cebruns has joined #openstack-cinder23:51
*** annashen has joined #openstack-cinder23:51
*** DericHorn-HP has joined #openstack-cinder23:52
*** jasondotstar has quit IRC23:52
*** DericHorn-HP has quit IRC23:53
*** annashen has quit IRC23:56
*** rmesta1 has quit IRC23:58
*** Yogi1 has joined #openstack-cinder23:59
*** laughterwym has joined #openstack-cinder23:59

Generated by 2.14.0 by Marius Gedminas - find it at!