Wednesday, 2015-01-14

*** bill_az has joined #openstack-cinder00:00
jgriffithanish: gotta run00:00
*** atan8 has joined #openstack-cinder00:00
jgriffithanish: let's chat tomorrow on those failing unit tests if you don't get it sorted00:00
*** dannywil_ has quit IRC00:00
*** annashen has quit IRC00:01
anishjgriffith: sure, thanks !00:01
*** annashen has joined #openstack-cinder00:01
*** jseiler__ has joined #openstack-cinder00:01
*** dannywilson has joined #openstack-cinder00:02
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** gtt116_ has joined #openstack-cinder00:02
*** tsekiyam_ has joined #openstack-cinder00:02
*** madskier has joined #openstack-cinder00:03
*** scottda_ has joined #openstack-cinder00:04
*** chlong has joined #openstack-cinder00:05
*** Mandell has joined #openstack-cinder00:07
*** changbl has quit IRC00:08
*** ho has joined #openstack-cinder00:08
*** tsekiyama has quit IRC00:09
*** jseiler_ has quit IRC00:09
*** hypnosb has quit IRC00:09
*** ebalduf has quit IRC00:09
*** tellesnobrega has quit IRC00:09
*** gtt116 has quit IRC00:09
*** annashen has quit IRC00:10
*** annashen has joined #openstack-cinder00:10
*** Longgeek has quit IRC00:10
*** hypnosb has joined #openstack-cinder00:11
*** tellesnobrega has joined #openstack-cinder00:11
*** Mandell has quit IRC00:12
*** scottda_ has quit IRC00:12
*** tsekiyama has joined #openstack-cinder00:19
*** madskier has quit IRC00:19
*** madskier has joined #openstack-cinder00:20
*** tsekiyam_ has quit IRC00:21
*** tsekiyama has quit IRC00:23
*** madskier has quit IRC00:25
*** bswartz has joined #openstack-cinder00:26
*** david-lyle has joined #openstack-cinder00:28
*** tbarron has joined #openstack-cinder00:31
*** tbarron has quit IRC00:31
*** tbarron has joined #openstack-cinder00:32
*** EmilienM is now known as EmilienM|afk00:39
*** annashen has quit IRC00:44
*** ameade has joined #openstack-cinder00:44
*** fandi has joined #openstack-cinder00:45
*** Masahiro has joined #openstack-cinder00:46
*** zhiyan has joined #openstack-cinder00:48
*** cppforlife has joined #openstack-cinder00:50
*** EmilienM|afk has quit IRC00:52
*** sandywalsh_ has joined #openstack-cinder00:54
*** sandywalsh has quit IRC00:56
*** EmilienM has joined #openstack-cinder00:56
*** david-lyle has quit IRC00:58
*** dannywilson has quit IRC00:58
*** david-lyle has joined #openstack-cinder00:58
*** dannywilson has joined #openstack-cinder00:59
*** dannywilson has quit IRC01:03
*** david-lyle has quit IRC01:05
*** EmilienM has quit IRC01:07
*** jdurgin has quit IRC01:07
*** EmilienM has joined #openstack-cinder01:08
*** david-lyle has joined #openstack-cinder01:09
*** jdurgin has joined #openstack-cinder01:11
*** shakamunyi has quit IRC01:13
*** ebalduf has joined #openstack-cinder01:21
*** rwsu has quit IRC01:21
*** david-lyle has quit IRC01:24
*** afazekas has quit IRC01:25
*** ebalduf has quit IRC01:26
*** david-lyle has joined #openstack-cinder01:29
*** _cjones_ has quit IRC01:33
*** _cjones_ has joined #openstack-cinder01:33
*** david-lyle has quit IRC01:35
*** bill_az has quit IRC01:37
*** _cjones_ has quit IRC01:40
*** lcurtis has quit IRC01:51
*** david-lyle has joined #openstack-cinder01:52
*** EmilienM has quit IRC01:52
*** Longgeek has joined #openstack-cinder01:55
*** smoriya has quit IRC01:55
*** EmilienM has joined #openstack-cinder01:56
*** rushil has joined #openstack-cinder01:57
*** david-lyle has quit IRC01:59
*** smoriya has joined #openstack-cinder02:01
*** zhiyan has quit IRC02:05
*** zhiyan has joined #openstack-cinder02:07
*** Masahiro has quit IRC02:07
*** haomaiwang has joined #openstack-cinder02:08
*** tellesnobrega has quit IRC02:09
*** Longgeek has quit IRC02:11
*** vishy has quit IRC02:11
*** vishy has joined #openstack-cinder02:14
*** thingee has quit IRC02:15
*** peluse has quit IRC02:16
*** peluse has joined #openstack-cinder02:16
*** Yogi11 has joined #openstack-cinder02:17
*** mriedem1 has joined #openstack-cinder02:20
*** tellesnobrega has joined #openstack-cinder02:20
*** mriedem has quit IRC02:22
*** ebalduf has joined #openstack-cinder02:22
*** EmilienM has quit IRC02:22
*** EmilienM has joined #openstack-cinder02:23
*** julim has quit IRC02:24
*** IanGovett has quit IRC02:25
*** mtanino has quit IRC02:25
*** ebalduf has quit IRC02:27
*** Masahiro has joined #openstack-cinder02:28
*** Masahiro has quit IRC02:30
*** Masahiro has joined #openstack-cinder02:30
*** Masahiro has quit IRC02:34
*** mberlin1 has joined #openstack-cinder02:34
*** Masahiro has joined #openstack-cinder02:34
*** mberlin has quit IRC02:35
openstackgerritChangBo Guo(gcb) proposed openstack/cinder: Use is_valid_ipv6 from oslo.utils
*** Masahiro has quit IRC02:38
*** Masahiro has joined #openstack-cinder02:40
*** david-lyle has joined #openstack-cinder02:40
*** shakamunyi has joined #openstack-cinder02:41
*** alonmarx_ has joined #openstack-cinder02:45
*** harlowja is now known as harlowja_away02:46
*** git-harry has quit IRC02:47
*** jbernard has quit IRC02:47
*** shakamunyi has quit IRC02:47
*** mgagne has quit IRC02:47
*** russellb has quit IRC02:47
*** alonmarx has quit IRC02:48
*** krtaylor has quit IRC02:48
*** shakamunyi has joined #openstack-cinder02:48
*** git-harry has joined #openstack-cinder02:49
openstackgerritLi Yingjun proposed openstack/python-cinderclient: Add missing all-tenants option to transfer-list
*** zhiyan has quit IRC02:55
*** zhiyan has joined #openstack-cinder02:57
*** patrickeast has quit IRC03:01
*** cppforlife has quit IRC03:01
*** cppforlife has joined #openstack-cinder03:02
*** krtaylor has joined #openstack-cinder03:02
*** russellb has joined #openstack-cinder03:02
*** mgagne has joined #openstack-cinder03:04
*** mgagne is now known as Guest3302503:05
*** shakamunyi has quit IRC03:05
*** tbarron has quit IRC03:06
*** jordandh has quit IRC03:06
*** jordandh has joined #openstack-cinder03:06
*** jcru has joined #openstack-cinder03:07
*** annashen has joined #openstack-cinder03:11
*** Yogi11 has quit IRC03:12
*** fandi has quit IRC03:14
*** rushil has quit IRC03:15
*** russellb has quit IRC03:15
*** junhongl has quit IRC03:18
*** junhongl has joined #openstack-cinder03:19
*** jordandh has quit IRC03:21
*** gtt116__ has joined #openstack-cinder03:22
*** ebalduf has joined #openstack-cinder03:23
*** jordandh has joined #openstack-cinder03:23
*** david-lyle has quit IRC03:25
*** russellb has joined #openstack-cinder03:25
*** gtt116_ has quit IRC03:25
*** bill_az has joined #openstack-cinder03:26
*** ebalduf has quit IRC03:28
*** russellb has quit IRC03:38
*** atan8 has quit IRC03:39
*** russellb has joined #openstack-cinder03:39
openstackgerritTrung Trinh proposed openstack/cinder: Delete volume attached to already-deleted VM
*** EmilienM has quit IRC03:40
*** jcru has quit IRC03:42
*** EmilienM has joined #openstack-cinder03:44
*** Masahiro has quit IRC03:45
*** Longgeek has joined #openstack-cinder03:51
*** anshul has joined #openstack-cinder03:53
*** patrickeast has joined #openstack-cinder03:56
*** asselin_ has joined #openstack-cinder03:58
*** dannywilson has joined #openstack-cinder03:59
*** EmilienM has quit IRC04:01
*** annashen has quit IRC04:01
*** Masahiro has joined #openstack-cinder04:02
*** mriedem1 has left #openstack-cinder04:03
*** dannywilson has quit IRC04:03
*** annashen has joined #openstack-cinder04:08
*** EmilienM has joined #openstack-cinder04:08
*** gtt116_ has joined #openstack-cinder04:10
*** ho has quit IRC04:10
*** ho has joined #openstack-cinder04:10
*** rushiagr_away is now known as rushiagr04:12
*** gtt116__ has quit IRC04:13
*** ebalduf has joined #openstack-cinder04:25
*** annashen has quit IRC04:28
*** bill_az has quit IRC04:29
*** ebalduf has quit IRC04:29
*** Guest33025 has quit IRC04:32
openstackgerritxing-yang proposed openstack/cinder-specs: Consistency Groups Kilo Update
*** mgagne has joined #openstack-cinder04:34
*** EmilienM has quit IRC04:35
*** mgagne is now known as Guest5760504:35
openstackgerritxing-yang proposed openstack/cinder-specs: Consistency Groups Kilo Update
*** EmilienM has joined #openstack-cinder04:38
*** _cjones_ has joined #openstack-cinder04:41
*** Longgeek has quit IRC04:42
*** Longgeek has joined #openstack-cinder04:44
*** _cjones_ has quit IRC04:45
*** changbl has joined #openstack-cinder04:46
*** teju has joined #openstack-cinder04:51
*** patrickeast has quit IRC04:57
*** vishy has quit IRC04:57
*** vishy has joined #openstack-cinder05:04
*** annashen has joined #openstack-cinder05:09
*** asselin_ has quit IRC05:10
*** EmilienM has quit IRC05:12
*** ho__ has joined #openstack-cinder05:13
*** ho has quit IRC05:15
*** EmilienM has joined #openstack-cinder05:16
*** mwichmann has quit IRC05:17
*** annashen has quit IRC05:19
*** zigo has quit IRC05:20
*** zigo has joined #openstack-cinder05:21
*** dannywilson has joined #openstack-cinder05:21
*** mwichmann has joined #openstack-cinder05:22
*** nkrinner has joined #openstack-cinder05:22
*** ebalduf has joined #openstack-cinder05:25
openstackgerritVincent Hou proposed openstack/cinder: Remove the useless next link for volumes, transfers and backups
openstackgerritxing-yang proposed openstack/cinder: Verify the instance's existance in the VMAX driver
*** Raj_ has joined #openstack-cinder05:28
*** ebalduf has quit IRC05:29
openstackgerritHuang Zhiteng proposed openstack/cinder: Cleanup unused DB APIs, part I
*** jbernard has joined #openstack-cinder05:38
openstackgerritxing-yang proposed openstack/cinder: Verify the instance's existance in the VMAX driver
openstackgerritVincent Hou proposed openstack/cinder: Remove the useless next link for volumes, transfers and backups
*** ebalduf has joined #openstack-cinder05:47
*** russellb has quit IRC05:49
*** russellb has joined #openstack-cinder05:49
*** ebalduf has quit IRC05:51
*** annashen has joined #openstack-cinder05:59
*** lan_ has quit IRC06:04
*** nshaikh has joined #openstack-cinder06:06
openstackgerritDunrong Huang proposed openstack/cinder: Fix eqlx endless loop when server closes the connection
*** lan_ has joined #openstack-cinder06:08
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Imported Translations from Transifex
openstackgerritxing-yang proposed openstack/cinder: Allow multi-attach to happen in VMAX driver
*** annashen has quit IRC06:13
*** annashen has joined #openstack-cinder06:14
*** annashen has quit IRC06:14
*** Longgeek has quit IRC06:19
openstackgerritVincent Hou proposed openstack/cinder: Remove the useless next link for volumes, transfers and backups
*** anshul has quit IRC06:27
*** cppforlife has quit IRC06:28
*** cppforlife has joined #openstack-cinder06:30
*** Longgeek has joined #openstack-cinder06:34
*** nlevinki has joined #openstack-cinder06:41
*** jcru has joined #openstack-cinder06:43
*** anshul has joined #openstack-cinder06:44
*** Masahiro has quit IRC06:45
*** Masahiro has joined #openstack-cinder06:47
*** jcru has quit IRC06:47
*** ebalduf has joined #openstack-cinder06:47
openstackgerritDave Chen proposed openstack/python-cinderclient: This is Cinder Client specific changes for modifying volume and snapshot's image metadata.
*** sgotliv has quit IRC06:51
*** ebalduf has quit IRC06:51
*** Longgeek has quit IRC06:53
*** Masahiro has quit IRC06:54
*** Masahiro has joined #openstack-cinder06:54
openstackgerritDave Chen proposed openstack/python-cinderclient: Add commands for modifying image metadata
*** Longgeek has joined #openstack-cinder06:59
*** dannywilson has quit IRC07:00
Raj_Hi all07:07
Raj_while migration in cinder07:07
Raj_from where it will get host[capabilities] ..?07:08
*** zhiyan has quit IRC07:08
*** zhiyan has joined #openstack-cinder07:11
*** lan_ has quit IRC07:18
*** oomichi_ has quit IRC07:37
*** russellb has quit IRC07:37
*** Miouge has joined #openstack-cinder07:38
*** jordandh has quit IRC07:39
*** russellb has joined #openstack-cinder07:40
*** TobiasE has joined #openstack-cinder07:41
*** jordandh has joined #openstack-cinder07:45
*** lan__ has joined #openstack-cinder07:45
*** chlong has quit IRC07:47
*** pwehrle has joined #openstack-cinder07:58
*** fandi has joined #openstack-cinder08:00
*** jistr has joined #openstack-cinder08:00
*** Mandell has joined #openstack-cinder08:01
*** dulek has joined #openstack-cinder08:05
openstackgerritAbhijeet Malawade proposed openstack/cinder-specs: Return request ID to caller
*** Miouge has quit IRC08:11
*** sgotliv has joined #openstack-cinder08:17
openstackgerritVincent Hou proposed openstack/cinder: Remove the useless next link for volumes, transfers and backups
*** smoriya has quit IRC08:22
*** tshefi has joined #openstack-cinder08:29
*** russellb has quit IRC08:30
*** Guest57605 has quit IRC08:31
*** mgagne has joined #openstack-cinder08:33
*** mgagne is now known as Guest553808:33
*** sgotliv has quit IRC08:34
*** theanalyst has joined #openstack-cinder08:38
*** Masahiro has quit IRC08:38
*** russellb has joined #openstack-cinder08:39
*** Masahiro has joined #openstack-cinder08:39
*** Miouge has joined #openstack-cinder08:40
*** pradipta_away is now known as pradipta08:44
*** Mandell has quit IRC08:45
*** lan__ has quit IRC08:49
*** ebalduf has joined #openstack-cinder08:49
*** lan__ has joined #openstack-cinder08:50
*** ndipanov_ is now known as ndipanov08:51
*** ndipanov has quit IRC08:51
flip214argg, my devstack won't start anymore ... because there's no /opt/stack/cinder/bin/cinder-volume anymore.08:52
openstackgerritHuang Zhiteng proposed openstack/cinder: Raise correct exception when validate_connector failed
*** ebalduf has quit IRC08:54
flip214and gate-cinder-python27 failed because of oslo.concurrency. grmpf.08:54
*** jpich has joined #openstack-cinder09:06
*** ndipanov has joined #openstack-cinder09:09
*** jordanP has joined #openstack-cinder09:10
*** Miouge has quit IRC09:10
*** Miouge has joined #openstack-cinder09:15
*** nellysmitt has joined #openstack-cinder09:25
*** jamielennox is now known as jamielennox|away09:31
*** Miouge has quit IRC09:32
*** Miouge has joined #openstack-cinder09:33
openstackgerritPh. Marek proposed openstack/cinder: DRBD: Log an error if libraries can't be loaded
*** abhijeetm has joined #openstack-cinder09:39
*** Guest5538 has quit IRC09:51
*** Raj_ has quit IRC09:52
*** mgagne has joined #openstack-cinder09:55
*** mgagne is now known as Guest8097909:55
*** vishy has quit IRC09:59
*** coolsvap|afk is now known as coolsvap09:59
*** e0ne has joined #openstack-cinder10:00
*** dannywilson has joined #openstack-cinder10:01
*** dulek has quit IRC10:04
*** dannywilson has quit IRC10:05
*** dulek has joined #openstack-cinder10:10
*** vishy has joined #openstack-cinder10:10
*** lpetrut has joined #openstack-cinder10:11
*** vishy has quit IRC10:21
*** aix has joined #openstack-cinder10:22
*** russellb has quit IRC10:22
*** tbarron has joined #openstack-cinder10:24
*** russellb has joined #openstack-cinder10:26
*** sgotliv has joined #openstack-cinder10:28
*** jistr has quit IRC10:31
*** e0ne is now known as e0ne_10:31
*** tbarron has quit IRC10:33
*** tbarron has joined #openstack-cinder10:34
*** vishy has joined #openstack-cinder10:34
*** karimb has joined #openstack-cinder10:35
*** lpetrut has quit IRC10:37
*** coolsvap is now known as coolsvap|afk10:39
*** e0ne_ has quit IRC10:42
*** Masahiro has quit IRC10:42
*** lpetrut has joined #openstack-cinder10:47
*** jistr has joined #openstack-cinder10:48
*** coolsvap|afk is now known as coolsvap10:51
flip214may I offer my opinion that the gerrit reviews are a bit fragile?10:51
openstackgerritgit-harry proposed openstack/cinder: Fix bug in tgt conf for volume
*** e0ne has joined #openstack-cinder10:57
*** russellb has quit IRC11:00
*** coolsvap is now known as coolsvap|afk11:02
*** alexpilotti has joined #openstack-cinder11:03
*** russellb has joined #openstack-cinder11:06
*** coolsvap|afk is now known as coolsvap11:09
*** haomaiwang has quit IRC11:11
*** aix has quit IRC11:14
*** takedakn has joined #openstack-cinder11:18
openstackgerritRushi Agrawal proposed openstack/cinder: WIP: Public snapshots
*** e0ne_ has joined #openstack-cinder11:23
*** takedakn has quit IRC11:24
*** e0ne has quit IRC11:26
*** mdenny has quit IRC11:27
*** asselin has quit IRC11:28
*** aix has joined #openstack-cinder11:28
*** Masahiro has joined #openstack-cinder11:43
*** Masahiro has quit IRC11:47
*** erlon has joined #openstack-cinder11:49
*** ebalduf has joined #openstack-cinder11:52
*** rushiagr is now known as rushiagr_away11:53
*** e0ne_ has quit IRC11:55
*** ganso_ has joined #openstack-cinder11:56
*** ebalduf has quit IRC11:57
*** e0ne has joined #openstack-cinder11:58
*** ho__ has quit IRC11:58
*** takedakn has joined #openstack-cinder12:05
*** chlong has joined #openstack-cinder12:06
*** teju has left #openstack-cinder12:08
*** lan__ has quit IRC12:10
*** takedakn has quit IRC12:11
*** marcusvrn1 has quit IRC12:15
*** miarmak has joined #openstack-cinder12:16
*** lpetrut has quit IRC12:18
*** Raj__ has joined #openstack-cinder12:30
*** pradipta is now known as pradipta_away12:30
*** rushiagr_away is now known as rushiagr12:31
*** Raj__ has quit IRC12:31
*** RAJ__ has joined #openstack-cinder12:32
RAJ__Regarding cinder replication support12:32
RAJ__it is necessary of12:32
RAJ__Replication relationship db table12:32
RAJ__and multi back end support ..?12:32
*** marcusvrn has joined #openstack-cinder12:34
*** IanGovett has joined #openstack-cinder12:38
*** Masahiro has joined #openstack-cinder12:46
*** sgotliv_ has joined #openstack-cinder12:47
*** sgotliv has quit IRC12:47
*** rushil has joined #openstack-cinder12:48
*** sandywalsh_ has quit IRC12:49
*** Masahiro has quit IRC12:50
*** nlevinki has quit IRC12:53
*** ryanpetrello has joined #openstack-cinder12:53
*** lpetrut has joined #openstack-cinder12:55
*** boris-42 has joined #openstack-cinder12:56
*** e0ne has quit IRC12:56
DuncanTRAJ__: Can you rephrase the question, please?12:56
*** e0ne has joined #openstack-cinder12:57
RAJ__DuncaT: replication relationship table is needed for replication support  ??12:58
DuncanTErm, yes... the driver doesn't have to care about it though, the manager handles that.12:59
RAJ__DuncanT: hey can you look @my code13:02
*** chlong has quit IRC13:02
DuncanTI can look, but I know very little about the replication code13:02
*** zhipeng has joined #openstack-cinder13:02
openstackgerritDuncan Thomas proposed openstack/cinder: Fix LVM thin pool creation race
DuncanTMaybe put up a review and mark it as work-in-progress?13:06
*** nlevinki has joined #openstack-cinder13:08
*** lpetrut has quit IRC13:10
*** takedakn has joined #openstack-cinder13:11
*** takedakn has quit IRC13:12
*** takedakn has joined #openstack-cinder13:12
RAJ__Link for code ...13:12
RAJ__take the review and let me knw about that13:13
*** e0ne is now known as e0ne_13:14
RAJ__e0ne_: can u take a look13:14
DuncanTThere are a whole bunch of things in there that make me nervous, but what is the problem? Does it work?13:17
*** takedakn has quit IRC13:19
*** takedakn has joined #openstack-cinder13:20
*** e0ne_ is now known as e0ne13:23
e0neRAJ__: what should i see in this code? i'm a bit out of context13:24
*** lpetrut has joined #openstack-cinder13:25
RAJ__DuncanT: it is working with volume type which includes :13:26
RAJ__capabilities=replication now13:26
RAJ__i m confused regarding13:26
RAJ__DuncanT: promote and reenable ..?13:26
RAJ__DuncanT: from where storage node 1 knw that which is 2nd storage node ..?13:27
rushiagrDuncanT: saw your comment on public snapshots work. I don't think sharing snapshots has anything to do with glance<->cinder interactions.13:27
RAJ__e0ne: hi canb you help me to clear doubts related to13:27
RAJ__promote and reenable replication part ..?13:27
rushiagri.e., I don't think glance is involved at all..13:27
DuncanTRAJ__: I don't know the answer to that, and I haven't time right now to go digging through the code, sorry13:29
DuncanTrushiagr: I'm saying we *could* do public snapshots, but they aren't much different to glance images, and glance already does ACLs and a bunch of other useful things, so why not use glance images instead of adding public snapshots13:30
rushiagrDuncanT: oh, so you're just saying that use glance's code to implement same functionality here?13:31
rushiagrthat's a good idea actually13:32
rushiagroh, maybe you mean just copy stuff to glance if one wants to share13:32
DuncanTYes, I mean actually copy into glance (and make that fast, which has other benefits)13:33
*** rushil has quit IRC13:33
*** takedakn has quit IRC13:34
*** RAJ__ has quit IRC13:42
*** nshaikh has quit IRC13:42
DuncanTrushiagr: If we do public snapshots, I worry we'll have to deal with all of the other things glance has to deal with (access control, lifecycle questions, attestation, metadata, etc) in future.13:44
DuncanTrushiagr: We want glance to be fast for uploading and downloading BfV images anyway, and once that is done then you get public snapshot-like behaviour for free13:45
*** nshaikh has joined #openstack-cinder13:47
*** bill_az has joined #openstack-cinder13:51
rushiagrDuncanT: I submitted a blueprint which talks about sharing snapshots (not only making it public), which is simpler than ACLs13:52
rushiagr(spec not yet up for the BP)13:52
*** nlevinki has quit IRC13:53
rushiagrso this is the blueprint where I was thinking to end the 'sharing' idea, and not go for full ACLs, and other stuff around it13:53
DuncanTrushiagr: Again, I'd say use glance, it already lets you do this13:53
DuncanTrushiagr: It is reinventing the wheel IMO13:53
rushiagrDuncanT: but I'll have to copy snapshot to glance to make this happen13:54
rushiagrwhile in case of sharing, I don't have to create any more bits. Only when the new tenant creates a volume out of that snapshot. Until that point, sharing means just an additional database entry or two13:55
*** takedakn has joined #openstack-cinder13:55
DuncanTAt the moment, yeah. Optimisations like the solidfire one means it would only get copied once (and you can pre-populate the cache on upload so that it never needs to bepulled back down), and you could make cinder a backend for glance if you really wanted13:56
DuncanTI'd rather see the macro case, i.e. the inefficiency of glance<->cinder interactions, fixed, rather than adding hacks that fix sub-cases as they come up13:56
*** nlevinki has joined #openstack-cinder13:58
rushiagrI'm not talking about bfv case, but in general. Someone might want to share a snapshot ('hey I have a DB dump in my vol snapshot, just create a vol out of it and take it!')13:58
rushiagrI am tending towards thinking that the problems are completely different. Optimizing glance-cinder interactions is a valid point, but this snapshot sharing functionality isn't anyway related to it as far as I can see..14:00
*** julim has joined #openstack-cinder14:02
rushiagrWe had a problem with sharing snapshots. Some people here wanted to do that, so I proposed a solution. I never thought it would be a subcase of another bigger issue :)14:04
*** sandywalsh has joined #openstack-cinder14:04
*** nlevinki has quit IRC14:04
*** tbarron has quit IRC14:04
*** tbarron has joined #openstack-cinder14:05
*** Yogi1 has joined #openstack-cinder14:05
jgriffithrushiagr: I still say "who cares"14:09
*** fandi has quit IRC14:09
jgriffithrushiagr: honestly all these "great features we have to have" and most of them are meh14:09
*** nlevinki has joined #openstack-cinder14:10
jgriffithrushiagr: I think ACL's are a horrible idea too FWIW14:10
rushiagrjgriffith: I am with you with all the above points you mentioned :)14:10
jgriffithrushiagr: but I also think the "sharing" thing your proposing for snapshots is "not so good" either14:10
*** mkerrin has joined #openstack-cinder14:11
rushiagrjgriffith: not so big a benefit, but little, and non-zero :)14:11
*** fandi has joined #openstack-cinder14:11
rushiagrjgriffith: do you have any specific concerns? I'm curious to know..14:12
*** dustins has joined #openstack-cinder14:12
jgriffithDuncanT: by the way what I submitted is NOT a *hack* as you call it.  It's an enhancement that take advantage of products specific architecture14:12
*** dustins has quit IRC14:12
rushiagrjgriffith: apart from 'more code to maintain' which I can totally understand..14:12
jgriffithrushiagr: actually can't say I have any real concern right now since there's no code :)14:13
jgriffithrushiagr: I've never been real big on the idea of ACL's at all.... and as far as this, YES it sounds like it would be more sane14:14
rushiagrjgriffith: :) I didn't want to waste writing code if the idea is going to be rejected straightaway :)14:14
jgriffithrushiagr: but the idea of breaking the relationship betwen snap and volume just to share it seems overkill14:14
rushiagrjgriffith: but if code give more weight to the idea, I'll write it first and then talk :)14:14
jgriffithrushiagr: one is able to share volumes14:14
jgriffithrushiagr: or more accurately "give"14:15
jgriffithrushiagr: which I suppose might be an argument (for consistency sake)...14:15
*** dalgaaf has joined #openstack-cinder14:17
DuncanTjgriffith: Your optimisation I don't consider a hack, adding snapshot sharing I consider a hack. Sorry if that wasn't clear14:18
*** takedakn has quit IRC14:19
jgriffithDuncanT: ahh... thanks, I was very confused14:19
DuncanTjgriffith: I'm saying we should do *more* optimisations like yours, as long as they don't change the behaviour and generally fall back to the old behaviour, which is exactly what your code does14:20
*** kaufer has joined #openstack-cinder14:20
*** nlevinki has quit IRC14:20
*** diegows has joined #openstack-cinder14:21
jgriffithDuncanT: well of course I'd agree with that :)  Sorry, I wasn't following the thread correctly.14:22
*** Raj2 has joined #openstack-cinder14:22
Raj2hey anone seen avishay today ../14:23
*** jungleboyj has quit IRC14:25
*** takedakn has joined #openstack-cinder14:26
*** nkrinner has quit IRC14:28
rushiagrjgriffith: I think sharing of snapshots is still an alternative way to share data. If you ask me, it is a better way than the current volume transfer. Sharing snapshot just creates a permission for other guy to create a volume out of it. The other guy then can decide if he wants to do that and when and how many times, and charged accordingly14:30
jgriffithrushiagr: sure14:30
*** Yogi11 has joined #openstack-cinder14:30
jgriffithrushiagr: but TBF, xfr volume, create a snap, do whatever you want14:30
jgriffithrushiagr: it's all the same14:30
jgriffithrushiagr: that was a sequence of events BTW14:31
jgriffithxfr-volume--->new user does whatever they want14:31
jgriffithand that's why we have all these crazy "create from snap,image,volume etc"14:32
jgriffithrushiagr: I dunno, maybe I'm not seeing it but maybe others do/will14:32
*** nlevinki has joined #openstack-cinder14:33
rushiagrif I transfer 10 volumes to another tenant, he'll be charged for those 10 right away, starting from the moment they're transferred14:33
*** Yogi1 has quit IRC14:34
rushiagrin case of snapshot, the other user isn't charged, as it's 'my' snapshot, which he can list in 'snapshot list'. He can decide when he wants to create a vol out of it14:35
*** Masahiro has joined #openstack-cinder14:35
rushiagrto be honest, I'm making up these examples inpromptu. Trying to make the idea appear different from (and valid at the same time!) vol xfer14:37
*** tbarron has quit IRC14:38
*** tbarron has joined #openstack-cinder14:39
rushiagrI have added it to today's meeting agenda... Let's see14:39
*** Masahiro has quit IRC14:39
*** mwichmann has joined #openstack-cinder14:42
openstackgerritEric Harney proposed openstack/cinder: RemoteFS: move volume/snap locking code to remotefs
openstackgerritEric Harney proposed openstack/cinder: WIP: NFS snapshots
*** akerr has joined #openstack-cinder14:42
*** dustins_ has joined #openstack-cinder14:45
*** tsekiyama has joined #openstack-cinder14:47
DuncanTrushiagr: He can already do exactly that - upload the volume image to glance, set it public or set an ACL on it14:47
*** mtanino has joined #openstack-cinder14:49
DuncanTrushiagr: Anybody who wants to can create a volume from that image. Not one line of code needed14:50
*** miarmak has quit IRC14:51
*** tbarron has quit IRC14:53
*** tbarron has joined #openstack-cinder14:53
rushiagragree, but there is a data transfer involved. If glance and cinder are on different backends.14:53
rushiagrI'm completely in agreement that things can be done currently too. I disagree that that's the best way to do.14:54
jgriffithrushiagr: I dunno, in this case I think it might be the better way to do it (the way DuncanT describes it being better)14:56
*** ebalduf has joined #openstack-cinder14:56
jgriffithrushiagr: so you could do that, you could xfr a volume, you could backup/restore....14:57
jgriffithwhat's another way14:57
jgriffithrushiagr: oh... attach to an instance and export it that way14:57
*** tsekiyam_ has joined #openstack-cinder14:58
*** abhijeetm has quit IRC14:59
*** tsekiya__ has joined #openstack-cinder15:00
*** tsekiyam_ has quit IRC15:00
*** ebalduf has quit IRC15:00
*** Mandell has joined #openstack-cinder15:01
*** tsekiyama has quit IRC15:01
*** dustins_ has quit IRC15:02
*** jungleboyj has joined #openstack-cinder15:02
*** thangp has joined #openstack-cinder15:05
rushiagrif one has a snapshot, its two steps to the end user for sharing (create a vol from that snapshot, transfer vol). I was just to make it a single-step..15:06
*** rushil has joined #openstack-cinder15:07
jgriffithrushiagr: yeah, I get ya15:07
rushiagrs/just/just trying/15:07
*** jecarey has joined #openstack-cinder15:08
*** jcru has joined #openstack-cinder15:11
openstackgerritJay Bryant proposed openstack/cinder: Add hacking check for oslo namespace usage
*** jcru has quit IRC15:12
flip214can somebody explain me the check-grenade-dsvm FAILURE from, please? link is
*** jcru has joined #openstack-cinder15:14
*** mriedem has joined #openstack-cinder15:16
*** Yogi11 has quit IRC15:16
*** tsekiya__ has quit IRC15:17
nikesh_vedamsjgriffith: good evening15:19
nikesh_vedamsjgriffith: sorry morning15:19
*** tsekiyama has joined #openstack-cinder15:20
openstackgerritIvan Kolodyazhny proposed openstack/python-cinderclient: Added formeted view for snapshots in volume details
openstackgerritIvan Kolodyazhny proposed openstack/cinder: Added snapshots list to volume details view
*** Miouge has quit IRC15:22
openstackgerritNavneet Singh proposed openstack/cinder: Implement Cinder Manage/Unmanage APIs in NFS driver.
*** tsekiyama has quit IRC15:23
*** tsekiyama has joined #openstack-cinder15:24
dulekGuys, I'm looking for a way to add periodic_task to cinder-api (is this even possible?) or hook to service startup. Any ideas where I can start?15:25
e0nedulek: look to the cinder-volume. it already uses periodic tasks15:26
dulekYeah, I've seen that, but cinder-api doesn't use manager where I would put a task. Or I'm unable to find it.15:27
*** fischerw has joined #openstack-cinder15:28
dulekSame problem with service startup. I mean cinder.wsgi.Server is rather generic and isn't a good place.15:28
guitarzandulek: you say you need a periodic task in the api... what do you really need? :)15:28
dulekOkay, so I'm looking for a way to resume running TaskFlow's workflow.15:29
dulekI can get them into the DB using taskflow.persistence module.15:30
*** dustins has joined #openstack-cinder15:30
*** Yogi11 has joined #openstack-cinder15:30
guitarzanwhy are your api tasks stopping?15:30
dulekI'm just killing it, I'm aiming to add more fault tolerance here.15:31
dulekSo resuming killed tasks in cinder-volume and cinder-scheduler is easy - periodic task or service startup.15:31
guitarzanthe general idea for restarting/continuing tasks was to start up all the tasks on startup I thought15:31
dulekIs there a good way to do it in the API?15:31
guitarzanfor example, if your api service or volume service dies, you can restart the tasks when you restart the service15:31
DuncanTJust do it on service startup, if a task dies but the service is still up then you should just fix the bug that caused that to happen15:32
*** rhe00_ has joined #openstack-cinder15:33
dulekDuncanT: That's not that easy in case of scheduler - at startup it doesn't have info about cinder-volume nodes, so it will fail the scheduling.15:33
dulekFor cinder-volume this is working.15:33
dulekAnd for API service - probably will work also, but I can't find decent place to put the restarting code.15:34
dulekAny ideas?15:35
DuncanTdulek: Ok, the scheduler shouldn't do anything except collect stats until at least the bcast interval has passed, resuming tasks included.15:35
dulekDuncanT: Correct.15:36
dulekSo my only concern is cinder-api now.15:36
*** ndipanov has quit IRC15:37
*** ebalduf has joined #openstack-cinder15:38
*** Miouge has joined #openstack-cinder15:38
*** e0ne is now known as e0ne_15:38
*** e0ne_ is now known as e0ne15:38
dulekHm, okay I think I'm looking at this from the wrong side...15:39
*** ndipanov has joined #openstack-cinder15:39
*** xyang has joined #openstack-cinder15:40
e0nedulek: actually, it's the same what DuncanT mentioned in a bug15:40
*** Mandell has quit IRC15:40
duleke0ne: Can you point me to the bug?15:40
dulekI guess the place would be cinder.volume.api.API.__init__15:41
e0nei'm going to provide patch for it later this week15:42
duleke0ne: Ah, that's cool. I'm doing unrelated work, but will hit this eventually. Your solution will be helpful, thanks!15:43
e0nedulek: ok. i'll keep you updated on it15:44
*** Yogi11 has quit IRC15:46
*** r-daneel has joined #openstack-cinder15:48
*** asselin has joined #openstack-cinder15:55
flip214  /join #openstack-meeting15:57
*** tshefi has quit IRC15:57
*** Guest80979 is now known as mgagne15:57
*** mgagne has joined #openstack-cinder15:57
*** thingee has joined #openstack-cinder15:57
*** jdurgin has quit IRC15:58
*** avishay has joined #openstack-cinder16:00
*** nshaikh has quit IRC16:00
*** AlkaD has joined #openstack-cinder16:00
*** scohen has joined #openstack-cinder16:01
*** diegows has quit IRC16:01
*** rmesta has joined #openstack-cinder16:01
*** deepakcs has joined #openstack-cinder16:01
*** nlevinki has quit IRC16:03
openstackgerritIvan Kolodyazhny proposed openstack/cinder: Added snapshots list to volume details view
*** ryanpetrello_ has joined #openstack-cinder16:04
*** patrickeast has joined #openstack-cinder16:05
*** kvidvans has quit IRC16:06
*** Yogi1 has joined #openstack-cinder16:06
*** ryanpetrello has quit IRC16:08
*** ryanpetrello_ is now known as ryanpetrello16:08
*** winston-d_zZZ is now known as winston-d16:09
openstackgerritAnton Arefiev proposed openstack/cinder: Add tenant validation to quota
*** bruff has joined #openstack-cinder16:09
*** alexpilotti_ has joined #openstack-cinder16:10
*** alexpilotti has quit IRC16:12
*** alexpilotti_ is now known as alexpilotti16:12
*** rushil has quit IRC16:13
openstackgerritAnton Arefiev proposed openstack/cinder: Add tenant validation to quota
*** jdurgin has joined #openstack-cinder16:14
*** markvoelker has joined #openstack-cinder16:15
*** rakesh___510 has joined #openstack-cinder16:17
*** hypnosb_ has joined #openstack-cinder16:20
*** Raj2 has quit IRC16:20
hypnosb_Hi, was there any progress with the iser implementation or was it tossed out?16:21
*** kkk has joined #openstack-cinder16:23
*** rakesh___510_ has joined #openstack-cinder16:23
*** krtaylor has quit IRC16:23
*** Masahiro has joined #openstack-cinder16:24
*** pwehrle has quit IRC16:25
*** rakesh___510 has quit IRC16:26
*** ndipanov has quit IRC16:27
*** Masahiro has quit IRC16:28
*** Longgeek has quit IRC16:29
*** takedakn has quit IRC16:29
nikesh_vedamsjgriffith: hi16:32
*** juzuluag has joined #openstack-cinder16:33
*** mdenny has joined #openstack-cinder16:34
openstackgerritGary W. Smith proposed openstack/cinder: Add 3PAR extra specs schema into volume stats
*** ndipanov has joined #openstack-cinder16:39
*** markvoelker has quit IRC16:40
*** krtaylor has joined #openstack-cinder16:41
openstackgerritXiaolei Hu proposed openstack/cinder: Implements: blueprint ibm-flashsystem-driver-iscsi Change-Id: I25f8ae0e1b67b1f65e2accebbd9dd978d11a5037
*** rakesh___510_ has quit IRC16:44
openstackgerritEdwin Wang proposed openstack/cinder: Add iSCSI protocol support for IBM FlashSystem.
jgriffithwinston-d: so that's a really good question16:49
jgriffithwinston-d: my suspicion is "no"16:49
jgriffithwinston-d: but when you look at the code it's easy to get confused on what that var is used for I think16:50
winston-djgriffith: i'd guess the same16:50
jgriffithwinston-d: also, I personally thought that it would be more accurate to update filter-sched info on create/delete rather than rely on periodic as well16:50
jgriffithwinston-d: the quota thing is troubling too16:51
*** ebalduf has quit IRC16:51
jgriffithwinston-d: but I *think* we have a backup for that in other places16:51
*** ebalduf has joined #openstack-cinder16:52
winston-djgriffith: fyi, scheduler keeps track of capacity in between two periodic updates in its memory, so scheduler wouldn't allow a create to be placed on a backend without capacity.16:52
jgriffithwinston-d: ahh!  I didn't see that16:53
jgriffithwinston-d: that's good to know16:53
jgriffithwinston-d: now I remember you fixed that :)16:53
winston-dbut if there are more than one scheduler, well, typical distributed computing problem, we don't get schedulers to sync for now.16:53
jgriffithwinston-d: I logged a bug a long time ago for that16:53
jgriffithwinston-d: understood (WRT multiple schedulers)16:54
*** rushil has joined #openstack-cinder16:55
winston-djgriffith: there's another bug about scheduler don't get updates fast enough when it starts, Ivan assigned himself to fix that bug, but I remember we fixed that once.16:55
winston-dI need to take another look.16:56
*** madskier has joined #openstack-cinder16:56
winston-djgriffith: sorry, I missed the conversation about taskflow/retries, not sure how it messes up with allocated_capacity.16:57
*** annashen has joined #openstack-cinder16:57
*** ebalduf has quit IRC16:57
winston-djgriffith: I thought no matter how many times taskflow retries, vol manager only increament/decreament once when create/delete succeeded.16:58
openstackgerritGary W. Smith proposed openstack/python-cinderclient: Expose cinder's scheduler pool stats API
jgriffithrushiagr: DuncanT avishay what ever happened with mgagne 's public private stuff?17:00
*** kvidvans has joined #openstack-cinder17:00
thingeeDuncanT: I hope you don't mind me taking over the CI enforcing stuff?17:01
jgriffithI think it was mgagne that was working on that17:01
DuncanTthingee: Definitely not. I was getting accused of being obsessed / too pushy with it17:01
rushiagrjgriffith: hmm..Let me check that out. I'm remembering something..but maybe that wasn't what exactly mgagne was doing..17:01
*** avishay has quit IRC17:01
DuncanTVolumes are different to snapshots because volumes are mutable BTW17:02
*** lpetrut has quit IRC17:02
winston-djgriffith: that was public/private volume type17:02
DuncanT@avishay ^^^17:02
rushiagrjgriffith: no, I don't see work from mgagne related to snapshot sharing17:02
winston-djust type17:02
jgriffithwinston-d: ahh... yes17:03
jgriffithwinston-d: well... now the precedent is set :)17:03
jgriffithI say we extend it17:03
*** patrickeast has quit IRC17:03
*** coolsvap is now known as coolsvap|afk17:03
*** dulek has quit IRC17:03
DuncanTSo if we do public snapshots, then how do you catalogue them? How do you filter them?17:03
*** AlkaD has quit IRC17:03
jgriffithDuncanT: I don't know what that means :)17:04
jgriffithDuncanT: you mean in a "snapshot list"17:04
*** AlkaD has joined #openstack-cinder17:04
jgriffithDuncanT: same way we do already with things that have the notion of public acess17:04
thingeejgriffith: I think he means if you have snapshots shared with you, what does snapshot list do to filter those in with your snapshots you own primarily17:04
*** anshul has quit IRC17:04
rushiagrDuncanT: not sure about catalog. Filter should be easy I think17:04
jgriffithDuncanT: you modify the get command to get all tenant owned and all public17:04
DuncanTjgriffith: Yeah. Take a public cloud as a degenerate case. there might be thousands of public snaps17:05
jgriffithDuncanT: better implement a quota :)17:05
DuncanTDo you *really* want them all showing up on every snapshot list17:05
jgriffithDuncanT: honestly, maybe it doesn't make sense to turn it on in a public cloud?17:05
rushiagrpass a param, 'owner=me' types, or better to make it the default..17:05
jgriffithDuncanT: certainly not, but IMO this is kinda of a niche case17:06
jgriffithDuncanT: and if I were running a public cloud I probably wouldn't even turn it on17:06
DuncanTjgriffith: Maybe. I still think glance has already dealt with this usecase, other than the performance, so can we just make it perform better?17:06
jgriffithDuncanT: sure, I'm not saying we should do it one way or other.  Just an idea17:06
jgriffithDuncanT: and also consider if consistency is an option17:07
DuncanTjgriffith: Maybe even bite the bullet and make cinder a fully first class glance backend, then you get rid of all the data copying17:07
mgagnejgriffith: only worked on public/private volume types17:07
jgriffithmgagne: yeah, winston-d corrected me :)17:07
jgriffithDuncanT: well, I looked at that again...17:08
DuncanTI'd wonder if there isn't a quota / DoS issue too... create a public snap, let somebody else clone it and write to it, delete the public snap, repeat. 2xdisk space usage for every one of those volumes on many backends17:08
hypnosb_DuncanT: i vote yes17:08
DuncanThypnosb_: ?17:09
hypnosb_cinder glance backender17:09
DuncanTjgriffith: You looked at it again... any road blocks?17:10
*** Longgeek has joined #openstack-cinder17:10
jgriffithDuncanT: yeah, it was a big fat PITA :)17:10
DuncanTjgriffith: :-(17:10
jgriffithDuncanT: seriously though.... I think there's a good way to do it17:10
jgriffithI just hadn't solved it17:11
*** asselin has quit IRC17:11
jgriffithDuncanT: we *should* be able to just leverage what nova does with snapshot bootable instances17:11
jgriffithDuncanT: just modify it a bit and be done17:11
jgriffithDuncanT: everything is apparantly *there*17:11
hypnosb_i did some silly shit yesterday and exported a n iscsi lun because images are copied from glance17:12
hemnaakerr, saw your comments about FC and doing CI17:12
jgriffithbut then I thought... "what's the win here"17:12
DuncanTYeah, feels that way to me too17:12
hemnaakerr, we are doing CI for our FC driver17:12
hypnosb_still broke17:12
hypnosb_idk why the damn thign is read only17:12
hypnosb_stupid liotarget17:12
thingeehypnosb_, jgriffith: yeah I think we need to deprecate cinder backstore in glance.17:13
thingeeno interest in fixing it17:13
thingeethere was a ML post about it17:13
thingeeI pushed back to give time for people to fix it, but that has proved that no one seems to care.17:14
DuncanTjgriffith: I think the wins are 1) Faster BfV & copy volume to image 2) Fast public snapshots 3) In theory faster ephemeral if you CoW over an iSCSI import of a cinder volume (particularly if you can R/O multi-attach that volume)17:14
DuncanTThe existing cinder backingstore appears fundamentally broken...17:14
jgriffithDuncanT: the first two I agree for sure17:15
jgriffithDuncanT: you'll have to help me with number 317:15
jgriffithDuncanT: I'll take a look again17:15
DuncanTNumber three is a bit next-gen, but:17:15
ameadeakerr: he's running back to his desk17:16
jgriffiththingee: yeah, well we just have to do *something* it's sat their incomplete and causing confusion for WAY too long17:16
thingeejgriffith, hypnosb_: I think I'll post that topic for the next cinder meeting, proposal to deprecate it.17:16
akerrhemna: hi17:16
thingeejgriffith, hypnosb_: or maybe just submit a patch to glance17:16
ameadeheh put wrong nick17:16
jgriffiththingee: but there's nothing to deprecate :)17:16
hemnaakerr, hey17:16
thingeejgriffith: :)17:16
DuncanTIf you multi-attach a boot image, as a volume, to a compute node, R/O, then throw a cow layer over the top, you avoid copying any bit of the image that isn't accessed, and can cache on the compute node, giving you faster provisioning and less network usage17:16
jgriffiththingee: just some half written code and some bp's17:16
hypnosb_its really the glance folks that have the cinder backed ...17:17
thingeejgriffith: true17:17
akerrhemna: ya, we think its certainly doable, but I wasn't sure we had the manpower to get it done right before 3-1917:17
*** Mandell has joined #openstack-cinder17:17
hemnaakerr, so doing FC in CI is only slightly more difficult than CI17:17
jgriffithDuncanT: interesting17:17
*** jpich has quit IRC17:17
jgriffithDuncanT: so you'd have Nova put a FS and a cow layer on the volume?  Maybe as part of BDM?17:17
thingeehypnosb_: right17:17
jgriffithDuncanT: or am I completely missing something17:18
DuncanTjgriffith: Yeah17:18
jgriffithDuncanT: that's pretty cool17:18
DuncanTjgriffith: I've done PoC work on it17:18
jgriffithDuncanT: how do we distinguish that property in Cinder?17:18
DuncanTjgriffith: Make all the glance images owned by a specific tenant maybe? After that it is just an attach17:19
DuncanTjgriffith: I've not yet found an open COW-over-raw device driver, but maybe I'll be able to get ours released17:20
*** vbala has joined #openstack-cinder17:20
thingeerushiagr: your item is first for next meeting
jgriffithso 1 and 2 are wins anyway17:20
jgriffith3 is gravy17:20
DuncanTjgriffith: qcow is an option but that slightly compliicates upload/download17:20
rushiagrDuncanT: I couldn't fully grasp the DoS/quota scenario you meantioned17:21
DuncanTYeah, 3 is something to shoot for, the other two are easy wins17:21
rushiagrI'll keep a spec ready by next meeting17:21
rushiagrthingee: thanks for that..17:21
jgriffithalright, I'll look at it again17:21
jgriffithbut if I don't have something by end of next week I'm punting or turning it over to someobdy else17:21
*** e0ne is now known as e0ne_17:21
DuncanTjgriffith: Cool. I'll have a read of the glance backend API myself17:21
jgriffithand if it's not in K3 it needs to get wiped from history :)17:21
DuncanTjgriffith: The existing code should be nuked anyway IMO, it is crap17:22
hypnosb_big problem17:22
*** annashen has quit IRC17:22
hypnosb_was going to say we could have an image driver.. but then cinder isnt an image solution?17:22
*** sgotliv_ has quit IRC17:23
DuncanThypnosb_: ?17:23
*** xyang has quit IRC17:23
jgriffithhypnosb_: yo... we're not exactly timid or prudish around here but you might be close to pushing some boundaries17:24
*** Longgeek has quit IRC17:24
hypnosb_my b17:25
jgriffithhypnosb_: no worries :)17:25
kkkHi all geek folks17:26
hypnosb_thats a pretty big boundry push ^17:26
kkki need help asap for cinder replication pls help me in thta17:26
kkki am very new to it pls guys17:26
*** TobiasE has quit IRC17:26
DuncanTkkk: What are you trying to do, exactly?17:27
*** juzuluag has quit IRC17:27
*** TobiasE has joined #openstack-cinder17:27
jgriffithhypnosb_: LOL17:27
hypnosb_well i was thinking if the cinder backend for glance isnt working out so well17:27
*** Raj__ has joined #openstack-cinder17:27
*** kkk has quit IRC17:27
Raj__hey its mee only due to another browser its kk nick name17:28
Raj__i want to report  cinder rep with some days only17:28
*** rwsu has joined #openstack-cinder17:29
Raj__DuncanT: i working on that ...17:29
Raj__todays issue till now17:29
hypnosb_maybe the cinder itself should pull from the catalog?17:29
hypnosb_that could be easier?17:29
*** leeantho has joined #openstack-cinder17:30
hypnosb_im so confused17:31
*** _cjones_ has joined #openstack-cinder17:31
*** TobiasE has quit IRC17:31
*** takedakn has joined #openstack-cinder17:31
*** asselin has joined #openstack-cinder17:32
*** e0ne_ has quit IRC17:32
openstackgerritRichard Hagarty proposed openstack/cinder: Set snapCPG when managing existing volumes
*** _cjones_ has quit IRC17:32
*** EmilienM is now known as EmilienM|afk17:32
*** lpetrut has joined #openstack-cinder17:32
hypnosb_from what I assume in the implementation of the services is what hardware you place them and the types of IO patters?17:33
DuncanTRaj: You need to say very clearly what your problem is, or nobody is going to be able to help17:33
*** deepakcs has quit IRC17:34
hypnosb_like for example with cinder I have a 40G 10k sas array, but for glance... i dont need that?17:34
*** Raj__ has quit IRC17:34
DuncanThypnosb_: None of the service nodes need that much in the way of local disk I/O generally17:34
hypnosb_aathe sme ith a0 gwhich is a problem wihe va computenodes ahe  l disks17:35
*** xyang has joined #openstack-cinder17:35
hypnosb_at the same time, its a problem when images are 40GB and compute nodes have lsmall disks17:35
*** _cjones_ has joined #openstack-cinder17:36
*** diegows has joined #openstack-cinder17:36
DuncanThypnosb_: If you're using large ephemeral boot disks then you need to spec the compute nodes appropriately... otherwise use ceph or a SAN17:36
marcusvrnDuncanT: thingee Hi! I was in another meeting, so I couldn't be in the cinder meeting. I have read the log of the meeting and about CI I have some questions that maybe are related winston-d questions. Sometimes, some job fail because the devstack failed to run. How can we know if the CI is reliable or not? if it fails once, is it not reliable? Looking for another scenario...once the CI is OK and after k-3 the CI stop w17:37
DuncanTmarcusvrn: If your CI disagrees with jenkins 5 patches in a row for any reason other than a genuine code bug, it can be considered broken17:38
*** annashen has joined #openstack-cinder17:38
DuncanTmarcusvrn: If that is caused by devstack failing on your CI but not on jenkins, it still counts as your CI being broken. Bugs and patches against devstack welcome.17:38
*** Raj__ has joined #openstack-cinder17:38
*** jistr has quit IRC17:39
DuncanTmarcusvrn: Running a CI takes some effort. That isn't too unexpected...17:39
*** hypnosb_ has quit IRC17:40
marcusvrnDuncanT: ok...yes! But how many time will we have to fix the CI after infra team have disabled it, before the driver be cut off?17:41
DuncanTmarcusvrn: I don't think there will be a simple answer to that. If you have somebody answering emails, trying things and generally staying in contact, there will be a high degree of flexibility. If all contact goes dead, expect less time17:43
*** takedakn has quit IRC17:43
DuncanTmarcusvrn: We don't /want/ to remove any drivers, we /do/ want to encourage participation and communication17:44
marcusvrnDuncanT: awesome! it's very nice17:44
Raj__DuncanT: u got something from net regarding that questions ...?17:45
*** e0ne has joined #openstack-cinder17:45
*** scohen has quit IRC17:46
DuncanTRaj: ? Regarding what questions?17:46
*** ganso_ has left #openstack-cinder17:47
*** jordanP has quit IRC17:48
*** Longgeek has joined #openstack-cinder17:49
*** dannywilson has joined #openstack-cinder17:50
*** scohen has joined #openstack-cinder17:50
asselinmarcusvrn, I'm interested to know what kind of devstack failures you get that jenkins isnot17:50
*** annegentle_ has quit IRC17:51
nikesh_vedamsjgriffith : now launch instance is working but since i am using trusty image which have a default username "ubuntu" to give in ssh login17:52
nikesh_vedamsjgriffith: so in sos-ci do we have provision to login like this ssh -i /home/vedams/cloudkey.pem ubuntu@
*** ndipanov has quit IRC17:55
*** kbyrne has quit IRC17:56
nikesh_vedamsjgriffith: how can we give absolute path of cloudkey.pem and ubuntu@instance-ip17:56
Raj__DuncanT: we can discussed regarding17:58
*** Mandell has quit IRC18:00
*** dannywilson has quit IRC18:00
*** Raj__ has quit IRC18:01
jgriffithnikesh_vedams: your questions have nothing to do with CI, whether it be sos-ci or anything else18:01
jgriffithnikesh_vedams: this is openstack user 10118:01
jgriffithnikesh_vedams: ssh -i keyname ubuntu@xxxxx is how cloud images work!18:01
jgriffiththere's no customization there or magic18:02
jgriffithnikesh_vedams: you can set things in your .ssh/config to make key and login usage automatic18:02
jgriffithagain, standard ssh stuff, not sos-ci or anything else18:02
*** Mandell has joined #openstack-cinder18:03
marcusvrnasselin: yesterday I was running the CI and an unauthorized error while devstack was trying to upload an image was raised, but running again, it was OK18:03
*** e0ne has quit IRC18:04
*** annashen has quit IRC18:05
*** ndipanov has joined #openstack-cinder18:08
*** annashen has joined #openstack-cinder18:08
*** chlong has joined #openstack-cinder18:10
*** harlowja_away is now known as harlowja18:10
harlowjalol, jgriffith u are GREAT!18:12
harlowja*form meeting*18:12
thingeeharlowja: heh18:12
thingeejgriffith is great18:12
thingeedon't forget to vote!18:12
*** Masahiro has joined #openstack-cinder18:12
harlowjaah, voting18:13
harlowjaguess i should do that18:13
harlowjabtw; i'm not the only one that understands taskflow i hope, others in #openstack-state-management i hope do :-P18:13
harlowja is afaik built on taskflow ;)18:13
harlowjaso if that means anything...18:14
harlowjaand ML2 blah blah and some other projects are adopting it18:14
thingeeharlowja: where's vilobh?18:14
thingeeshould be back from vacation now?18:14
harlowjashould be back soon i thinks18:14
harlowjalet me poke someone18:14
thingeeharlowja: I have retargeted
jungleboyjthingee: or DuncanT any issues with getting this in:  Would like to start building the other namespace changes on it today if we can get it in.18:15
harlowjathingee kk18:15
harlowjai'll find out that date; gotta be soon, marriages don't take that long, lol18:15
harlowjabut they last for a lifetime i hear, ha18:15
*** Guest97290 has joined #openstack-cinder18:16
*** crose has joined #openstack-cinder18:16
*** Masahiro has quit IRC18:18
openstackgerritAbel Lopez proposed openstack/cinder: Purge deleted rows
harlowjajgriffith u guys do know that sqlalchemy is done by like one guy ;)18:19
openstackgerritThang Pham proposed openstack/cinder: WIP: Snapshot object
openstackgerritThang Pham proposed openstack/cinder: WIP: Cinder objects base
harlowjaanyways, here to help18:20
harlowjathingee next week (after mLK) vilobh be back18:20
harlowjaor thats what i'm told18:20
thingeejungleboyj: lgtm18:21
jungleboyjthingee: Thanks!  Going to spend the afternoon on namespace changes then.18:21
*** crose has quit IRC18:29
*** EmilienM|afk is now known as EmilienM18:29
*** patrickeast has joined #openstack-cinder18:30
jungleboyjjgriffith: I would like to get this fixed and pushed back up:  Any concerns with me adding the use of the fixture and pushing it back up?18:32
*** rushiagr is now known as rushiagr_away18:37
*** Mandell has quit IRC18:37
*** Mandell has joined #openstack-cinder18:39
nikesh_vedamsjgriffith :ok thanks,18:40
nikesh_vedamsjgriffith: in boot_openstack_instance.yml,name: wait for ssh to be ready single nic,timeout=120 is giving ssh timeout for me so increased to 300,that also timeout so now given 600 :)18:42
*** e0ne has joined #openstack-cinder18:42
*** e0ne is now known as e0ne_18:48
openstackgerritCorey Bryant proposed openstack/cinder: Deal with PEP-0476 certificate chaining checking
*** jungleboyj has quit IRC18:49
*** e0ne_ has quit IRC18:53
*** Guest97290 is now known as annegentle18:53
*** ndipanov has quit IRC18:55
*** karimb has quit IRC18:56
*** _cjones_ has quit IRC18:57
*** _cjones_ has joined #openstack-cinder19:01
*** bruff has left #openstack-cinder19:01
*** vbala has left #openstack-cinder19:04
*** mudassirlatif has joined #openstack-cinder19:05
*** juzuluag has joined #openstack-cinder19:06
*** e0ne has joined #openstack-cinder19:06
*** _cjones_ has quit IRC19:07
*** _cjones_ has joined #openstack-cinder19:08
*** dannywilson has joined #openstack-cinder19:09
*** alexpilotti has quit IRC19:10
*** csky has joined #openstack-cinder19:10
*** annashen has quit IRC19:18
*** jungleboyj has joined #openstack-cinder19:21
*** aix has quit IRC19:22
*** Mandell has quit IRC19:22
*** Mandell has joined #openstack-cinder19:24
*** esker has joined #openstack-cinder19:28
*** xyang has quit IRC19:39
*** Mandell has quit IRC19:40
hemnaanyone run lately?19:47
hemnait's failing locally from master on oslo_concurrency lockutils19:47
thangphemna: i think jungleboyj fixed that19:47
thangphemna: like last night?19:48
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: cinder-manage should use API instead of direct DB communication
thangphemna: yup, saw it19:48
jungleboyjhemna: Yes, I fixed that yesterday.19:48
hemnajungleboyj, but hasn't merged yet?19:49
hemnaor has it ?19:49
jungleboyjIt has.19:49
hemnaI'll pull and see19:49
jungleboyjYou oslo man was on top of that one.19:49
hemnanice job Captain Oslo!19:50
hemnayou'll be in the next Avengers movie.19:50
*** zhipeng has quit IRC19:54
*** jecarey has quit IRC19:54
kvidvansthingee: can enhancements be submited for new drivers added in kilo?19:55
anishhemna: is that what you are seeing ?19:56
*** AlkaD has quit IRC19:56
hemnaanish, no19:56
anishno bueno19:57
hemnaI did a pull master and it fixed the problems I was seeing.19:57
hemnathanks to captain oslo19:57
* anish tries again19:57
tsekiyamahemna: please give me a comment on nova-specs of multipath enhancement - about multipathd detection and fallback to single path setup.19:58
*** Mandell has joined #openstack-cinder19:59
anishtsekiyama: I'm doing something only mildly related, but I kinda see the same issue that you might20:01
anishyou are saying login failure on a subset of portal when multipathing is enabled is acceptable, yes ?20:01
*** Masahiro has joined #openstack-cinder20:01
*** e0ne is now known as e0ne_20:01
*** AlkaD has joined #openstack-cinder20:02
tsekiyamaanish: yeah, that is one of the use cases of my proposal.20:02
anishtsekiyama: that is great, I was wondering if that would be a side effect of using multipath in conjuction with
*** krtaylor has quit IRC20:05
tsekiyamaanish: reading it .. just a moment plz20:05
anishsure, take your time. i'm afk for 10 min20:06
*** Masahiro has quit IRC20:06
*** e0ne_ has quit IRC20:07
mudassirlatifI have a question about provider_location field in the volume ….. is it ok for a driver to use this field to store information about volume location?20:08
*** david-lyle has joined #openstack-cinder20:10
mudassirlatifor in other words … is there a restriction by cinder on using this field to store driver specific data?20:11
*** _cjones_ has quit IRC20:11
*** e0ne has joined #openstack-cinder20:12
thingeekvidvans: yes20:14
*** krtaylor has joined #openstack-cinder20:18
anishtsekiyama: back. any comments ?20:22
tsekiyamaHow the device file looks like with offloading? something like /dev/disk/by-path/pci-0000:00.00.0-ip-<ip-address>-iscsi-<target>-<lun>?20:23
tsekiyamaanish: hmm, I need to more closely on the patch, but conceptually this should work with multipath.20:24
anishthe main issue there is that on older versions of open-iscsi, discovery does not work over an offloaded transport, though 874 does support it20:24
anishand you're supposed to manually add the interface parameter when you login20:24
tsekiyamaanish: ah, then my patch will be useful to omit discovery to tell nova the all the paths to be login20:25
anishwhich is doable when doing a regular --op new, but transports are bound to interfaces, and multipath may go over multiple interfaces and cause login failures if you use transports20:26
anishyeah,your patch does solve any potential issues for me20:26
tsekiyamaanish: although I will have to modify the code to merge with the patch for specifying transport, as these are modifying the same place around volume attach/detach20:28
*** crose has joined #openstack-cinder20:28
openstackgerritIvan Kolodyazhny proposed openstack/cinder-specs: Cinder DB Archiving
*** harlowja is now known as harlowja_away20:30
*** _cjones_ has joined #openstack-cinder20:30
anishtsekiyama: yes. it might be a bit complicated for you. something like try transport, if it fails then try with tcp, if that also fails, bail20:30
anish(unless you can come up with a better solution)20:31
anishthis is partly mitigated when transport is iser, the target config explicitly asks for iser (someone else is looking at the code for this already)20:31
anishbut hardware offload transports don't require any changes to remote target unlike iser and hence do not require anything target side20:32
erlonwinston-d: hi Huang, can you you give a look in this change?
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder: Add volume multi attach support
erlonwinston-d: the unit tests are in a separete patch:
anishtsekiyama: the only changes that really clash are the ones in _get_host_device(s)20:35
*** elazro has quit IRC20:38
*** sgotliv has joined #openstack-cinder20:38
*** harlowja_away is now known as harlowja20:38
tsekiyamaanish: right, it looks not so difficult to merge with multipath stuff20:38
*** Mandell has quit IRC20:38
*** elazro has joined #openstack-cinder20:38
anishtsekiyama: yep. just that whoever gets merged later needs to make changes20:39
openstackgerritAbel Lopez proposed openstack/cinder: Purge deleted rows
*** lcurtis has joined #openstack-cinder20:54
*** kmartin has quit IRC21:00
*** david-lyle has quit IRC21:01
*** Mandell has joined #openstack-cinder21:03
jungleboyjI have a new version of the _lazy fix for stable/juno pushed up if anyone can take a look:
anishjgriffith: failure no longer seen after a git pull today21:07
*** _cjones_ has quit IRC21:10
*** kmartin has joined #openstack-cinder21:10
*** AlkaD has quit IRC21:11
*** rhe00_ has quit IRC21:11
*** _cjones_ has joined #openstack-cinder21:13
*** rhe00_ has joined #openstack-cinder21:13
*** AlkaD has joined #openstack-cinder21:14
*** e0ne is now known as e0ne_21:18
anishquestion, my changeset changes the way the iscsiadm is used in nova.21:19
anishA lot of this code seems to have an exact copy in cinder/brick/initiator/connector.py21:19
anishdo I need to change that as well ?21:19
jgriffithanish: yeah, hemna 's plan is to have that be shared code between the two21:24
openstackgerritErlon R. Cruz proposed openstack/cinder: Fix HNAS driver confusing error message (iSCSI driver)
anishjgriffith: but changes aren't automatically transplanted yet ?21:24
*** Mandell has quit IRC21:25
hemnaanish, for now, yes.  if changes are made in nova's iscsi/libvirt code, then those changes should get mapped into cinder brick as well.21:25
hemnaanish, correct.21:25
hemnaanish, I'm actually in the process of extracting cinder's brick initiator code into a separate cinder subproject called 'brick'21:25
hemnaafter that is done, cinder will get a patch to use that external lib.21:25
hemnaand then I'll submit a patch in nova at some point to also use brick21:26
anishso what is the correct workflow ? If my changes get merged to nova, should I open a review for brick as well ?21:26
hemnaanish, yes21:26
hemnaplease :)21:26
*** markstur has quit IRC21:26
*** tbarron has quit IRC21:26
anisheasey peasey. will do as soon as nova goes through21:26
*** tbarron has joined #openstack-cinder21:27
*** julim has quit IRC21:27
*** tbarron has quit IRC21:28
*** e0ne_ is now known as e0ne21:28
*** tbarron has joined #openstack-cinder21:28
*** Mandell has joined #openstack-cinder21:33
anishhemna: btw, is there any plan to get things like iscsi_ip_address integrated into neutron ?21:35
anishI ask because of some potential issues with automation of open-iscsi transports. When using the tcp transport, routing is done automatically to figure out correct interface to use. For hardware transports on the other hand, routing information must be known beforehand, and kinda implies that user is aware of what is the correct interface to use and provides this in the iscsiadm command; something that would not be too easy to do in brick as of21:35
anish(sorry bout the long text)21:35
jgriffithanish: no, there's not21:36
jgriffithanish: the design has been that the data network is independent and part of infrastructure21:36
jgriffithanish: keeping in mind that Instances have no real knowledge of iscsi or networking21:37
jgriffithanish: WRT to data path at any rate21:37
jgriffithanish: in other words, don't forget that the iscsi initiator is ONLY on the compute node21:37
jgriffithanish: attach, and pass through via virt-io21:37
anishoh, so some manual sanity for things like iscsi_ip_address is okay then ?21:39
jgriffithanish: honestly I don't really know what you're referring to21:40
jgriffithas the initiator code needs and should handle all of that21:40
jgriffithno, there should not be manual "work" in there21:40
*** e0ne is now known as e0ne_21:40
*** diegows has quit IRC21:42
*** e0ne_ is now known as e0ne21:42
*** david-lyle has joined #openstack-cinder21:42
anishjgriffith: so a one sentence explanation. if you have 4 interfaces, with only one of them actually connected to a target, the libiscsi kernel modules figure out the correct interface to use for you21:42
*** e0ne has quit IRC21:43
hemnaisn't that a routing problem on the host ?21:44
anishnow if you use a hardware transport with open-iscsi (via the -I param), then libiscsi will only try the interface you tell it to try21:44
hemnaanish, how do you determine which interface to use ?21:44
anishyou sorta have to know before hand. As the hardware transports are as the name implies retricted to certain hardware,you give it a 3-tuple of mac address, ip & a name to use for transport21:45
hemnaanish, other than keep trying until you find one that successfully creates an iSCSI session ?21:45
anishfor the default transport 'tcp' this is the case, yes21:46
anishfor an offloaded hardware transport, this is not possible as it is bound to the 3-tuple you provided21:46
hemnabut don't you have to tell the target side what 3-tuple you export to as well ?21:47
*** chlong has quit IRC21:47
anishno, that is only needed for iser21:47
hemnaalmost like how FC works21:47
anishis is transparent to target side when used21:48
hemnawho determines which h/w interface to use ?21:48
anishthis gets muddy only because iser is also classified as a transport, but it is not transparent21:48
anishhemna: left to user21:48
hemnawhat 'user' in this case?21:48
hemnauser is a loaded term here.21:48
anishsorry, admin21:49
hemnaso that would be a cinder.conf entry ?21:49
*** david-lyle has quit IRC21:49
anishwell it's a nova.conf entry currently21:49
anishsince it's only needed on initiator side21:49
hemnasomething has to pass to nova/brick the h/w transport to use instead of the default of TCP today21:49
*** thangp has quit IRC21:50
*** Masahiro has joined #openstack-cinder21:50
anishright now, nova.conf does the job21:50
hemnastill trying to wrap my head around this21:51
hemnaso, you could have N h/w transports over different interfaces21:51
*** sgotliv has quit IRC21:51
anish(my current code only allows for one)21:51
hemnaand one of those is the right one connected to the cinder array ?21:51
*** r-daneel has quit IRC21:53
hemnasounds like the initiator side needs to know which h/w transport is associated with which target host.21:54
anishas tp N h/w transports, the correct way to see this would be every mac-ip combo can have one transport. Most vendors only allow one mac-ip combo per mac21:54
hemnaor, you have to try each one until it works.21:54
*** Masahiro has quit IRC21:54
hemnaeach cinder iSCSI driver returns a target_portal21:55
hemnawhich contains ip:port #21:55
hemnathis sounds like a major pain for admins to maintain21:56
anishyou essentially need to know if the ip is reacheable via the transport, yes21:56
hemnaif you have to manually know which initiator-target pairs are associated21:56
anishwell, you could just put them all on the same subnet21:56
hemnaFC gets this for free, by the zone manager lookup service21:56
anishat least that is how I have seen it done21:57
hemnawhich basically returns a list of initiator-target maps that see each other.21:57
anishI don't think there's an iscsi equivalent, it's always same subnet + ACL as far as I have seen21:58
anishbut I'm on the dev side, not deployment21:58
*** nellysmitt has quit IRC21:59
hemnais there a way to inquire the h/w transport which hosts it sees ?21:59
*** annashen has joined #openstack-cinder21:59
hemnayah with TCP, you just let the routing table and ARP do the work21:59
anishping maybe ?21:59
hemnayah possibly21:59
hemnaI don't know enough about the h/w transports22:00
hemnabut I'm assuming they would encapsulate iSCSI ( and hence TCP) traffic over the wire22:00
hemnaand hope they support ICMP22:00
anishthink of it as ifconfig | grep for mac id22:00
anishthey do, it's jsut an abstraction over regular nics22:01
anish'regular' as in 'regular' nics that understand an iscsi packet in hardware22:01
hemnaok then a ping on each interface first prior to calling login22:01
hemnajust cycle through each interface until ping with a timeout.22:02
hemnaif you get a reply, use that one.  if not, next.22:02
hemnano replies = impossible to attach.22:02
hemnasome network admins disable ICMP22:03
anishdoes not seem like a graceful way either22:03
anishyeah, target arrays may not support ICMP either22:03
hemnabut at least this way, the admin doesn't have to pre configure/manage an association dictionary... FOR EACH compute host.22:03
hemnawhich just seems like an epic fail22:04
anishsadly yes22:04
anishhowever, even then22:04
anishsomeone has to make a h/w transport configuration to try22:04
hemnathey do?22:04
anishotherwise it would be more like this :22:05
openstackgerritAnthony Lee proposed openstack/cinder: Remove locks from LeftHand driver
anishfind pinging interface22:05
anishlookup pci id of pinging interface to see if it's one of the 7 known transport vendors22:05
anishfigure out transport name based on pci id, generate hw transport22:05
anishthen proceed22:05
anishI feel like that is instant gerrit rejection22:06
*** rajinir has joined #openstack-cinder22:06
hemnawell you can always try22:07
*** rajinir is now known as rajini22:07
*** markstur has joined #openstack-cinder22:07
*** rajini is now known as Guest5852422:07
*** Guest58524 has quit IRC22:08
boris-42jgriffith: hi22:09
*** Guest58524 has joined #openstack-cinder22:09
boris-42jgriffith:  how are you?22:09
*** sgotliv has joined #openstack-cinder22:09
jgriffithboris-42: hey22:10
*** _cjones_ has quit IRC22:11
boris-42jgriffith: so is there any progress on attach issue?)22:11
jgriffithboris-42: you tell me22:12
jgriffithboris-42: what's rally saying now?22:12
boris-42jgriffith: so today heat/neutron break rally gates22:12
boris-42jgriffith: and we are trying to disable that scenario22:12
jgriffithboris-42: well that's not helpful22:12
boris-42jgriffith: and we can't cause we are facing cinder attach=)22:12
jgriffithboris-42: so I haven't looked since I put the extra debug messages in22:12
jgriffithboris-42: let me do a Kibana search....22:13
boris-42jgriffith: this is job that failed22:13
*** _cjones_ has joined #openstack-cinder22:13
boris-42jgriffith: and another one
*** akerr has quit IRC22:15
*** Yogi1 has quit IRC22:15
*** juzuluag has quit IRC22:16
anishhemna: wait no, that is not 100% correct. you can do it without user intervention. iscsiadm is capabale of making h/w transport configurations on its own. You'll just end up with every possible configuration that the present hardware to test/ping etc22:16
anish*present hardware is capable of to test/ping and find the relevant one22:17
jgriffithso we sent a create target call... but clearly the target is already there22:17
jgriffithboris-42: hmmm22:19
jgriffithboris-42: ok, so interesting22:19
*** esker has quit IRC22:19
jgriffithboris-42: the persistence file is there22:19
jgriffithbut the target itself is never created22:20
boris-42jgriffith: magic*22:20
jgriffithboris-42: well... not really, just means tgt get's "hosed"22:20
jgriffithboris-42: notice also that the ONLY Lun is the controller lun22:21
jgriffithboris-42: in the "targets after udpate" display22:21
jgriffithboris-42: BUT22:21
jgriffithfurther down you see the tgt created for 35d9ce....22:22
jgriffithboris-42: so I *think* we have a race here22:22
jgriffithin other both of these calls are coming at the same time and tgt pukes22:22
jgriffiths/in other/in other words/22:23
*** diegows has joined #openstack-cinder22:23
boris-42jgriffith: but why we have few for one tgt?22:23
jgriffithboris-42: not sure what you mean?22:24
boris-42why there are few calls22:24
boris-42at all22:24
boris-42we are attaching different volumes22:24
jgriffithsure, but we're running concurrent tests right?22:25
boris-42we have simultaneouly N attaches22:25
jgriffithboris-42: so we can actually have two create target calls come in at the same time22:25
jgriffithboris-42: which would confirm it's a race and why we don't see 100% of the time22:26
jgriffithboris-42: of course I could be completely wrong, but this is sure what it looks like22:26
jgriffithboris-42: so the good thing is there are easy hacks to try and fix this now that we know what is going on22:26
jgriffithboris-42: It should be simple enough to do a retry, maybe even put a sleep in there22:27
jgriffithboris-42: some might say "put in a lock"22:27
jgriffithso we only do one at a time22:27
jgriffithmight not be a bad idea either (as much as I hate locks)22:27
jgriffithboris-42: know what I'm sayin?22:28
*** rushil has quit IRC22:29
boris-42jgriffith: I would try to avoid locks (cause as far as I understand they should be distributed in case of few drivers instaces?)22:30
jgriffithboris-42: I agree22:30
boris-42jgriffith: so retries seems a bit better idea22:30
*** EmilienM is now known as EmilienM|afk22:32
jgriffithboris-42: something like this maybe:
jgriffithboris-42: probably a sleep inbetween wouldn't be a terrible thing22:32
*** david-lyle has joined #openstack-cinder22:33
boris-42jgriffith: btw22:34
boris-42jgriffith:  we can test this solution easily22:34
boris-42jgriffith: just make a very very long task for rally22:34
boris-42in the depending patch22:34
jgriffithOh... good idea22:35
boris-42and recheck few times22:35
boris-42if it pass => good one22:35
jgriffithso let me look at some ideas22:35
jgriffithyou be around for a little bit?22:35
boris-42yep  for few hours22:35
jgriffithI'll get a patch ready but I want to look at a couple other things first22:35
jgriffiththere might be a better way to do these updates22:35
jgriffithlike an "update ALL"22:35
*** dustins has quit IRC22:40
*** kaufer has quit IRC22:45
openstackgerritGary W. Smith proposed openstack/python-cinderclient: Expose cinder's scheduler pool API
*** olc has quit IRC22:45
openstackgerritTomoki Sekiyama proposed openstack/cinder: TgtAdm: Don't change CHAP username/password on live migration
*** crose has quit IRC22:46
*** Guest58524 has quit IRC22:53
*** jamielennox|away is now known as jamielennox22:56
*** lpetrut has quit IRC22:59
*** ryanpetrello has quit IRC23:02
*** marcusvrn has quit IRC23:02
*** ryanpetrello has joined #openstack-cinder23:03
*** marcusvrn has joined #openstack-cinder23:03
jgriffithboris-42: let's try and experiment :)23:03
jgriffithboris-42: rather than do updates on specific targets, why not do an update on all targets in the persist file23:04
jgriffiththis may have a bad side effect that I'm not thinking of, but I'm not sure what23:04
jgriffithboris-42: it should be *safe* for active targets as long as I don't use forece23:04
boris-42jgriffith: let's try23:05
boris-42jgriffith: i was playing with logstach23:05
boris-42jgriffith: interesting that there were no messages 5 Jan23:05
*** tbarron has quit IRC23:06
boris-42jgriffith: and seems like adding more benchmarks to rally task in cinder start causing more failures of rally job23:06
boris-42jgriffith: in cinder23:06
jgriffithboris-42: so actually that kinda makes sense (I think)23:06
jgriffithboris-42: since I believe it's an issue of overlapped tgt update calls, increasing the load increases the likliehood of hitting overlaps23:07
jgriffithboris-42: now...23:07
jgriffithboris-42: what would be interesting is if there's a pattern23:07
jgriffithboris-42: like "it's always the n'th target" etc23:07
boris-42jgriffith: hm that will be hard to check23:08
jgriffithboris-42: sadly it's purely manual :(23:09
jgriffithboris-42: likely not worth thinking about yet23:09
boris-42jgriffith: we can try to run rally N times and collect info about what itereations failed actually23:09
*** _cjones_ has quit IRC23:09
boris-42jgriffith: iteration is more or less ID23:09
*** _cjones_ has joined #openstack-cinder23:11
boris-42jgriffith: going to burn my laptop23:11
boris-42jgriffith: btw23:12
boris-42jgriffith: I finished really cool stuff23:12
boris-42jgriffith: for such kind fo testing23:12
boris-42jgriffith: so you don't need to copy-paste by hands in task, you can write loop=)23:12
boris-42jgriffith: like goal you would like to test that all images works with all flaovrs23:14
boris-42jgriffith: you can write such stuff
jungleboyjAnyone else seeing that it looks like unit test is broken again?23:16
jgriffithboris-42: let's see what that does23:16
jgriffithoops... I beat it23:16
openstackgerritJohn Griffith proposed openstack/cinder: Use tgt-admin --update ALL
jgriffithI beat the internet... I'm faster than the internet!!!!23:16
*** david-lyle has quit IRC23:16
*** lcurtis has quit IRC23:17
jgriffithjungleboyj: haven't tried a fresh setup today... doing so now23:17
jgriffithjungleboyj: anything in Jenkins?23:17
jgriffithjungleboyj: looks okie-dokie to me23:17
jungleboyjHaven't looked that far yet.  Trying a fresh tree pull.23:18
boris-42jgriffith: lol23:18
*** _cjones_ has quit IRC23:18
jungleboyjjgriffith: Getting 'no such file or directory' looking for cinder-rtstool in a number of test cases.23:19
*** erlon has quit IRC23:20
jgriffithjungleboyj: yup, just hit it23:21
jungleboyjOk.  Good.  Couldn't figure out how my namespace changes could cause that.23:22
*** AlkaD has quit IRC23:24
jgriffithjungleboyj: thingee It appears at first glance that rtstool updated23:24
jgriffithand be broken23:24
*** _cjones_ has joined #openstack-cinder23:25
jungleboyjjgriffith: I think, though, that it is this commit as all the test cases I see failing are in here:
jgriffithjungleboyj: thingee 2.1.5123:26
thingeejgriffith: didn't you have a patch recently that resolved this by adding an additional path of where rtstool could reside23:26
jungleboyjIs rtstool set up as a test requirement?23:26
thingeejgriffith: at least for debian based23:27
thingeejungleboyj: it's in cinder/cmd/cinder-rtstool23:27
jgriffithjungleboyj: it's in requirements23:27
thingeertslib-fb is part of requirements23:27
jgriffithlooking... probably right that it's related to that commit somehow23:27
jungleboyjjgriffith: In each case it is just doing a 'cinder-rtstool verify' and not finding cinder-rtstool .23:28
thingeejungleboyj: link to whatever is failing?23:28
jgriffiththingee: jungleboyj so that commit added lio driver tests23:30
*** lcurtis has joined #openstack-cinder23:30
jgriffithin cinder.tests.targets.test_lio_driver23:30
jgriffithbut doesn't fake the rtstool calls23:30
jgriffithso "boom"23:30
jungleboyjjgriffith: Ah.23:30
jgriffithjungleboyj: thingee verifying now, will push a short term fix for now23:31
*** sgotliv has quit IRC23:31
jgriffithassuming this works :)23:31
thingeeha that's why I couldn't find the commit23:31
jgriffithjust removing the test_lio file for now23:31
thingee"Remove unused variables from ensure_export() "23:31
thingeethat's a really bad summary that this commit introduces tests for targets as well23:32
thingeemtanino: ^23:32
jgriffiththingee: partially my fault, he pointed out there were no LIO tests and I said.. yeah you should add some :)23:32
jgriffiththe tests that are in there are actually to test his change23:32
mtaninothingee: hi23:32
thingeemtanino: read scrollback please.23:33
mtaninothingee: I see.23:33
*** _cjones_ has quit IRC23:33
openstackgerritRichard Hagarty proposed openstack/cinder: Set snapCPG when managing existing volumes
jungleboyjPhew, I didn't merge it.  :-)23:33
jgriffithmtanino: thingee it's not really a problem, and since we "require" unit tests well... what would you expect :)23:33
jgriffithjungleboyj: well... if you run it on my ubuntu system it works fine :)23:34
*** dannywil_ has joined #openstack-cinder23:34
thingeejgriffith: :)23:34
jungleboyjjgriffith: ?  How is that?23:34
thingeejgriffith: I'll add a skip for now on the test class?23:34
jgriffithjungleboyj: just as if you run in gate it's fine23:34
thingeejungleboyj: he has it installed23:34
*** dannywil_ has quit IRC23:34
jungleboyjthingee: jgriffith Ah, yes, then it would work.23:35
jgriffiththingee: mtanino so that's what I had, and yes that solves any issues23:35
jgriffiththingee: I'll let you push that up23:35
*** dannywil_ has joined #openstack-cinder23:35
thingeejgriffith: k23:35
mtaninojgriffith: Thank you.23:35
jgriffithmtanino: you want to work on a fake for that at some point?  Maybe open a bug on it?23:35
jgriffithmtanino: if you have time23:35
jgriffithmtanino: if not I'm sure one of us can pick it up as well23:35
*** patrickeast_ has joined #openstack-cinder23:36
mtaninojgriffith: Could you push it?23:36
*** patrickeast has quit IRC23:36
*** patrickeast_ is now known as patrickeast23:36
jgriffithfor sure23:36
jgriffithwell... I think thingee has it already qued up23:36
*** mudassirlatif has quit IRC23:36
jgriffithI'll file a bug and mock out the rts calls23:37
*** mudassirlatif has joined #openstack-cinder23:37
jgriffithno big deal.... not a crisis like the crazy stuff last week23:37
jgriffiththis one is ez/pz23:37
*** dannywilson has quit IRC23:37
thingeejgriffith: don't worry about writing something to fake it, it's my target afterall...well forked target thing23:37
*** madskier has quit IRC23:38
thingeeI'd rather eharney write the fakes :P23:38
*** ebalduf has joined #openstack-cinder23:38
*** Masahiro has joined #openstack-cinder23:39
thingeeis anyone already filing a bug?23:39
*** _cjones_ has joined #openstack-cinder23:40
jungleboyjthingee: I can write one.23:40
thingeejungleboyj: thanks23:41
jgriffiththingee: ?23:41
jgriffithall you need to do is mock out a few of the calls in the test23:41
jgriffithshouldn't be a big deal23:41
jgriffithdon't know why eharney has to do it but... ok23:41
thingeejgriffith: it was a joke. I was talking about it's my target thing...well the forked version and I was poking fun since it's a redhat fork23:42
jgriffiththingee: lol... ahh, you funny guy23:42
jgriffithbb in a bit23:42
*** ebalduf has quit IRC23:42
*** Masahiro has quit IRC23:43
thingeejungleboyj: let me know when you got a bug #23:44
jungleboyjthingee: ^^23:45
openstackgerritWalter A. Boring IV (hemna) proposed openstack/cinder: Fix LOG formatting in api initialize_connection
*** ryanpetrello has quit IRC23:46
hemnawe need a ninja for ^^^23:46
*** kvidvans has quit IRC23:48
openstackgerritMike Perez proposed openstack/cinder: Skip LIO target drivers until properly mocked
thingeehemna, jungleboyj ^^23:49
*** lcurtis has quit IRC23:49
openstackgerritMike Perez proposed openstack/cinder: Skip LIO target unit tests until mocked
hemnajgriffith, thanks man23:51
thingeehemna: sorry one more time...fixed summary of commit23:51
thingeehemna: thanks23:51
*** mriedem has quit IRC23:52
boris-42jgriffith: lol23:53
*** smoriya has joined #openstack-cinder23:53
boris-42jgriffith: today we have happy day23:53
jungleboyjWas about to +2/+A ... You Ninja.23:53
mtaninothingee: Thank you for the patch.23:54
boris-42jgriffith: 3 times attach race for one patch=)23:54
hemnajungleboyj, that was my first official ninja23:56
*** dalgaaf has quit IRC23:57

Generated by 2.14.0 by Marius Gedminas - find it at!