Thursday, 2016-02-18

*** rhagarty has joined #openstack-cinder00:03
*** sdague has quit IRC00:06
*** rhagarty_ has joined #openstack-cinder00:06
jgriffithpatrickeast: so the fat finger was "cctxt" vs "ctxt" sigh00:07
*** rhagarty has quit IRC00:08
*** sgotliv has quit IRC00:11
*** sgotliv has joined #openstack-cinder00:11
openstackgerritHelen Walsh proposed openstack/cinder: EMC VMAX - get iscsi ip from port in existing MV
*** baojg has quit IRC00:17
*** zhonghua has quit IRC00:21
*** huanan_M has quit IRC00:22
*** huanan_M has joined #openstack-cinder00:22
*** huyang has quit IRC00:22
jgriffithSweet!... finally got that mess working00:22
*** huyang has joined #openstack-cinder00:22
*** zhonghua-lee has joined #openstack-cinder00:24
*** mdenny has quit IRC00:24
mc_nairjgriffith: on to the next mess then?00:27
*** garthb has quit IRC00:31
*** laughter_ has joined #openstack-cinder00:31
*** laughter_ has joined #openstack-cinder00:32
*** laughter_ has quit IRC00:32
*** ntpttr has quit IRC00:33
openstackgerritOpenStack Proposal Bot proposed openstack/cinder: Updated from global requirements
*** laughterwym has joined #openstack-cinder00:37
*** RA__ has joined #openstack-cinder00:38
patrickeastjgriffith: are those api additions you mentioned earlier changing the like top level api's? i've got my setup to the point where i'm ready to start to failover stuff and haz questions about how the http requests are supposed to be formed00:44
*** cknight has joined #openstack-cinder00:45
*** rcernin has quit IRC00:46
patrickeastjgriffith: im going to take a break for a bit this evening, lemmie know if you have them working or newer code, I don't wanna spend time duplicating any triage/fixes you've already got ;)00:46
*** cknight1 has joined #openstack-cinder00:48
*** cknight has quit IRC00:50
*** yangyapeng has joined #openstack-cinder00:51
*** arch-nemesis has quit IRC00:53
*** Suyash has quit IRC01:06
*** yangyapeng has quit IRC01:06
jgriffithpatrickeast: yeah, I just completely trashed my branch with a rebase01:07
jgriffithI give up01:07
jgriffithwill fix it later tonight after I stop sobbing01:07
jgriffithor hell, let's see what happens01:08
patrickeastjgriffith: oh no, can you just go back with the reflog?01:08
jgriffithpatrickeast: so, not sure how to fix01:09
jgriffithpatrickeast: I accidentally did a "git pull"01:09
jgriffiththe git commit --amend -a much later after finishing up my work01:09
jgriffithand kaboom01:09
jgriffithI don't know how to back it out without loosing my changes01:10
patrickeastOh, so you ammended to the wrong commit?01:10
patrickeastMm you can probably get then back out if you do a diff of the one you modified compared to the original01:12
patrickeastSave it and apply it as a patch XD01:12
jgriffithpatrickeast: yeah, trying that now, but can't get friking git to ignore binary files in the diff GRRRR01:12
jgriffithcould check google :)01:12
patrickeastHaha yea, a consultation with Dr Google might be helpful01:13
*** annasort has joined #openstack-cinder01:14
*** alonma has joined #openstack-cinder01:16
*** EinstCrazy has joined #openstack-cinder01:17
*** mtanino has quit IRC01:18
*** rhagarty_ has quit IRC01:20
*** rhagarty has joined #openstack-cinder01:20
*** alonma has quit IRC01:21
*** alonma has joined #openstack-cinder01:23
*** Suyash has joined #openstack-cinder01:25
openstackgerritJohn Griffith proposed openstack/python-cinderclient: Add replication v2.1 calls
*** alonma has quit IRC01:27
openstackgerritJohn Griffith proposed openstack/cinder: Replication v2.1
*** bill_az has quit IRC01:28
*** Suyash has quit IRC01:29
jgriffithOk, well that got most of it back... I'll finish it up later.01:31
*** yangyapeng has joined #openstack-cinder01:31
*** sheel has joined #openstack-cinder01:32
*** ebalduf has joined #openstack-cinder01:33
*** yangyapeng has quit IRC01:38
*** ebalduf has quit IRC01:38
*** yangyapeng has joined #openstack-cinder01:38
*** davechen has joined #openstack-cinder01:39
*** diogogmt has joined #openstack-cinder01:42
*** diogogmt has quit IRC01:42
*** diogogmt has joined #openstack-cinder01:44
*** baojg has joined #openstack-cinder01:45
*** leeantho has quit IRC01:45
*** esker has joined #openstack-cinder01:46
*** davechen1 has joined #openstack-cinder01:48
*** diogogmt has quit IRC01:49
*** davechen has quit IRC01:50
*** esker has quit IRC01:51
*** Lee1092 has joined #openstack-cinder01:56
*** rhagarty has quit IRC01:58
*** alonma has joined #openstack-cinder02:04
*** davechen has joined #openstack-cinder02:08
*** alonma has quit IRC02:08
*** davechen1 has quit IRC02:10
*** annasort has quit IRC02:11
*** haomaiwa_ has joined #openstack-cinder02:12
*** mylu has joined #openstack-cinder02:14
*** mriedem has left #openstack-cinder02:14
*** gouthamr has joined #openstack-cinder02:15
*** gouthamr_ has joined #openstack-cinder02:16
*** EinstCrazy has quit IRC02:17
*** EinstCrazy has joined #openstack-cinder02:18
*** gouthamr has quit IRC02:20
*** diogogmt has joined #openstack-cinder02:26
*** mudassirlatif has quit IRC02:29
*** houming has joined #openstack-cinder02:35
*** lixiaoy1 has joined #openstack-cinder02:38
lixiaoy1Hi, any cores help to review the patch to fix high importance bug of quota: which has got a +202:41
*** baojg has quit IRC02:43
*** annasort has joined #openstack-cinder02:45
*** baojg has joined #openstack-cinder02:47
*** apoorvad has quit IRC02:54
*** esker has joined #openstack-cinder02:55
*** esker has quit IRC03:00
*** haomaiwa_ has quit IRC03:01
*** haomaiwa_ has joined #openstack-cinder03:01
*** salv-orl_ has quit IRC03:04
*** IlyaG has joined #openstack-cinder03:14
*** IlyaG has quit IRC03:14
*** IlyaG has joined #openstack-cinder03:15
*** bardia has quit IRC03:16
*** mylu has quit IRC03:24
*** dims_ has quit IRC03:30
*** dims has joined #openstack-cinder03:32
*** john5223 has joined #openstack-cinder03:33
*** mylu has joined #openstack-cinder03:35
lixiaoy1DuncanT:dulek: Thank you to consider sheepdog things. To run gate-tempest-dsvm-full-sheepdog-nv test in jeckins, currently the problem is: devstack git clone the new os-brick patch, but it doesn't set LIBS_FROM_GIT=os-brick03:42
*** baojg has quit IRC03:42
lixiaoy1DuncanT:deule: As a result, it still uses non-updated os-brick from pip install. The infra patch is
lixiaoy1DuncanT: dulek: after it is merged, we can test the sheepdog tempest case.03:44
*** baojg has joined #openstack-cinder03:45
lixiaoy1DuncanT:dulek: above infra patch is to set up sheepdog to test with os-brick from source in experimental.03:46
*** EinstCra_ has joined #openstack-cinder03:49
*** alonma has joined #openstack-cinder03:51
*** diablo_rojo has joined #openstack-cinder03:52
*** EinstCrazy has quit IRC03:52
*** baojg has quit IRC03:55
*** alonma has quit IRC03:56
*** alonma has joined #openstack-cinder03:59
*** haomaiwa_ has quit IRC04:01
*** EinstCrazy has joined #openstack-cinder04:01
*** haomaiwang has joined #openstack-cinder04:01
*** Poornima has joined #openstack-cinder04:02
*** EinstCra_ has quit IRC04:04
*** alonma has quit IRC04:05
*** alonma has joined #openstack-cinder04:08
*** diablo_rojo has quit IRC04:09
*** davechen1 has joined #openstack-cinder04:12
*** davechen has quit IRC04:14
*** alonma has quit IRC04:14
*** baojg has joined #openstack-cinder04:25
*** RA__ has quit IRC04:37
openstackgerritPatrick East proposed openstack/cinder: Allow for Pure drivers to verify HTTPS requests
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Don't show chap password in log
*** shyama has quit IRC04:48
*** shyama has joined #openstack-cinder04:48
*** smoriya_ has joined #openstack-cinder04:48
*** bill_az has joined #openstack-cinder04:49
*** xyang1 has joined #openstack-cinder04:50
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Don't fail when port group does not exist
*** dims_ has joined #openstack-cinder04:53
*** dims has quit IRC04:55
openstackgerritWilson Liu proposed openstack/cinder: Match the ip more accurately in Huawei driver
*** xyang1 has quit IRC04:58
*** dims_ has quit IRC05:00
*** haomaiwang has quit IRC05:01
*** esker has joined #openstack-cinder05:01
*** haomaiwa_ has joined #openstack-cinder05:01
*** dims has joined #openstack-cinder05:02
*** dims has quit IRC05:02
*** alonma has joined #openstack-cinder05:03
*** esker has quit IRC05:05
*** ccesario has quit IRC05:05
*** alonma has quit IRC05:07
*** alonma has joined #openstack-cinder05:09
*** alonma has quit IRC05:14
*** laughter_ has joined #openstack-cinder05:14
*** NightKhaos has quit IRC05:15
*** NightKhaos has joined #openstack-cinder05:16
*** laughterwym has quit IRC05:18
*** ccesario has joined #openstack-cinder05:18
*** shyama has quit IRC05:19
*** shyama has joined #openstack-cinder05:20
*** esker has joined #openstack-cinder05:21
*** [1]Thelo has joined #openstack-cinder05:23
*** Thelo has quit IRC05:25
*** [1]Thelo is now known as Thelo05:25
*** ccesario has quit IRC05:26
*** esker has quit IRC05:26
*** dave-mccowan has quit IRC05:38
*** ccesario has joined #openstack-cinder05:39
*** bill_az has quit IRC05:40
*** liverpoo1er has quit IRC05:41
*** liverpooler has quit IRC05:42
*** mylu has quit IRC05:49
*** IlyaG has quit IRC05:51
*** laughter_ has quit IRC05:54
*** laughterwym has joined #openstack-cinder05:54
*** haomaiwa_ has quit IRC06:01
*** haomaiwang has joined #openstack-cinder06:01
*** abhishekk has joined #openstack-cinder06:05
*** jidar has quit IRC06:06
*** jwcroppe_ has joined #openstack-cinder06:07
*** jidar has joined #openstack-cinder06:07
*** alonma has joined #openstack-cinder06:10
*** jwcroppe has quit IRC06:10
*** jwcroppe_ has quit IRC06:11
*** alonma has quit IRC06:15
*** jwcroppe has joined #openstack-cinder06:18
*** alonma has joined #openstack-cinder06:19
*** ChubYann has quit IRC06:19
*** alonma has quit IRC06:24
*** esker has joined #openstack-cinder06:24
*** alonma has joined #openstack-cinder06:27
*** esker has quit IRC06:28
*** alonma has quit IRC06:32
*** bkumar has joined #openstack-cinder06:39
*** alonma has joined #openstack-cinder06:39
openstackgerritliuke proposed openstack/cinder: Huawei: Judgement failure when creating hypermetro
*** alonma has quit IRC06:45
*** alonma has joined #openstack-cinder06:46
*** baojg has quit IRC06:47
*** mvk has joined #openstack-cinder06:49
*** baojg has joined #openstack-cinder06:51
*** alonma has quit IRC06:51
*** alonma has joined #openstack-cinder06:52
openstackgerritliuke proposed openstack/cinder: Huawei: Judgement failure when creating hypermetro
*** alonma has quit IRC06:56
*** sheel has quit IRC06:57
*** takedakn has joined #openstack-cinder06:57
*** nkrinner has joined #openstack-cinder06:58
*** lpetrut has joined #openstack-cinder06:59
*** alonma has joined #openstack-cinder07:00
*** haomaiwang has quit IRC07:01
*** haomaiwa_ has joined #openstack-cinder07:01
*** jwcroppe has quit IRC07:02
*** alonma has quit IRC07:04
*** esker has joined #openstack-cinder07:06
*** liverpoo1er has joined #openstack-cinder07:06
*** liverpooler has joined #openstack-cinder07:06
openstackgerritLisaLi proposed openstack/cinder: Fix message of quota exceed exception
*** alonma has joined #openstack-cinder07:07
*** esker has quit IRC07:11
*** alonma has quit IRC07:11
*** alonma has joined #openstack-cinder07:13
openstackgerritliuke proposed openstack/cinder: Huawei: Create snapshot have a log error
*** baojg has quit IRC07:13
*** baojg has joined #openstack-cinder07:16
*** alonma has quit IRC07:18
*** alonma has joined #openstack-cinder07:19
*** alonma has quit IRC07:23
*** alonma has joined #openstack-cinder07:25
*** esker has joined #openstack-cinder07:27
*** alonma has quit IRC07:30
*** esker has quit IRC07:31
*** alonma has joined #openstack-cinder07:31
*** alonma has quit IRC07:36
*** ponychou has joined #openstack-cinder07:36
openstackgerritChaozhe Chen(ccz) proposed openstack/cinder: Test: use assert_has_calls() instead
*** alonma has joined #openstack-cinder07:38
*** e0ne has joined #openstack-cinder07:41
*** alonma has quit IRC07:42
*** alonma has joined #openstack-cinder07:44
*** salv-orlando has joined #openstack-cinder07:44
*** davechen1 is now known as davechen07:46
*** alonma has quit IRC07:48
*** alonma has joined #openstack-cinder07:50
*** rcernin has joined #openstack-cinder07:51
lixiaoy1DuncanT:dulek: The infra patch is merged, and DON'T MERGE patch is running with the new jeckins.07:53
*** yangyape_ has joined #openstack-cinder07:54
*** alonma has quit IRC07:54
*** EinstCra_ has joined #openstack-cinder07:55
*** alonma has joined #openstack-cinder07:55
*** yangyap__ has joined #openstack-cinder07:56
*** yangyapeng has quit IRC07:56
*** e0ne has quit IRC07:58
*** yangyape_ has quit IRC07:58
*** EinstCrazy has quit IRC07:59
*** ildikov has quit IRC08:00
*** takedakn has quit IRC08:00
*** haomaiwa_ has quit IRC08:01
*** haomaiwang has joined #openstack-cinder08:01
DuncanTlixiaoy1: Excellent stuff!08:02
*** lpetrut has quit IRC08:02
*** bkumar has quit IRC08:05
lixiaoy1DuncanT: The test finishes successfully:
*** esker has joined #openstack-cinder08:08
DuncanTGreat, expect two +2s within a few mintues08:08
DuncanTThat's all I've been waiting for08:08
lixiaoy1DuncanT: thank you!08:09
*** sb has quit IRC08:10
*** strictlyb has joined #openstack-cinder08:10
*** lpetrut has joined #openstack-cinder08:10
*** takedakn has joined #openstack-cinder08:11
*** ildikov has joined #openstack-cinder08:12
*** esker has quit IRC08:13
*** takedakn has quit IRC08:15
*** lpetrut has quit IRC08:19
*** ygl has joined #openstack-cinder08:30
yglHi All08:30
ygli have an issue with creating volumes from images08:30
yglcan anyone help me please08:30
yglis anyone here ?08:31
DuncanTygl: what exactly is the problem?08:32
*** abhishekk_ has joined #openstack-cinder08:32
yglits throwing the error saying this "ImageCopyFailure: Failed to copy image to volume: Image c72f1545-d3aa-4cbe-be0e-6ca765ee344e could not be found."08:32
yglbut i can see the image listed by glance and also in the database08:32
lixiaoy1ygl: which version are you using?08:33
yglactually i have changed the glance backend from swift to   to /var/lib/glance/images08:33
ygllixiaoy1: kilo08:34
*** abhishekk has quit IRC08:34
yglits failing at this code ""08:34
*** belmoreira has joined #openstack-cinder08:34
yglDuncanT: def _copy_image_to_volume(08:34
yglDuncanT: lixiaoy1 can u tell me how cinder downloads from the backend ?08:35
DuncanTygl: It just calls the glance client to do it08:35
yglDuncanT: lixiaoy1  i can also list the images through glance api calls08:36
yglDuncanT: ok08:36
DuncanTCan you download it though?08:36
DuncanTIt is possible to have things in the glance index but not the datastore08:36
yglDuncanT: how to test it from command line ?08:36
yglDuncanT: I mean how to download ?08:37
*** abhishekk has joined #openstack-cinder08:37
DuncanTGive me a minute, I'll see if I can find out08:37
yglDuncanT: sure thanks08:37
*** abhishekk has quit IRC08:37
*** abhishekk_ has quit IRC08:37
DuncanTglance image-download > outfile08:38
DuncanTYou'll need to put the image id in there too08:39
*** jordanP has joined #openstack-cinder08:39
openstackgerritChaozhe Chen(ccz) proposed openstack/cinder: Test: use assert_has_calls() instead
*** alonma has quit IRC08:40
yglDuncanT:  i have a three node cluster for glance services.08:42
*** laughterwym has quit IRC08:42
DuncanTygl: Does the image-download work? If not, your glance setup is bust, and I don't know much about glance08:43
*** laughterwym has joined #openstack-cinder08:43
*** lpetrut has joined #openstack-cinder08:43
yglDuncanT: it worked on one node. now trying on other nodes08:44
yglDuncanT: this is the other error I got on pther node: 404 Not Found: Image /var/lib/glance/images/c72f1545-d3aa-4cbe-be0e-6ca765ee344e not found (HTTP 404)08:44
DuncanTOk. You need to fix that, then cinder will probably work08:45
yglDuncanT: But I see the image on this other node08:46
yglDuncanT: its in /var/lib/glance/images08:46
DuncanTygl: I know very little about how glance works, sorry08:46
*** laughterwym has quit IRC08:47
yglDuncanT: ok Thanks08:47
*** ygl has left #openstack-cinder08:48
lixiaoy1ygl: do you enable debug trace?08:48
*** esker has joined #openstack-cinder08:50
*** tpatzig_ has quit IRC08:52
*** tpatzig has joined #openstack-cinder08:53
*** tpatzig has left #openstack-cinder08:53
*** esker has quit IRC08:54
openstackgerritRui Chen proposed openstack/python-cinderclient: Fix Resource.__eq__ mismatch semantics of object equal
*** haomaiwang has quit IRC09:01
*** haomaiwang has joined #openstack-cinder09:01
*** lpetrut has quit IRC09:02
*** houming has quit IRC09:07
*** esker has joined #openstack-cinder09:11
*** arecknag has joined #openstack-cinder09:14
*** esker has quit IRC09:15
*** ndipanov has joined #openstack-cinder09:18
*** EinstCrazy has joined #openstack-cinder09:19
*** yangyapeng has joined #openstack-cinder09:19
*** ildikov has quit IRC09:20
*** yangyap__ has quit IRC09:21
*** EinstCra_ has quit IRC09:21
*** ndipanov has quit IRC09:22
*** yangyape_ has joined #openstack-cinder09:22
*** yangyapeng has quit IRC09:25
*** ociuhandu has joined #openstack-cinder09:29
*** yuriy_n17 has joined #openstack-cinder09:30
*** e0ne has joined #openstack-cinder09:35
*** ildikov has joined #openstack-cinder09:36
*** ndipanov has joined #openstack-cinder09:41
*** yangyape_ has quit IRC09:45
lixiaoy1DuncanT: Please also help to review the patch about scaling backup:
*** houming has joined #openstack-cinder09:50
*** laughterwym has joined #openstack-cinder09:50
*** gouthamr_ has quit IRC09:51
*** laughterwym has quit IRC09:55
*** davechen has left #openstack-cinder09:55
DuncanTlixiaoy1: Done. Not sure what the situation is with these patches and rolling upgrade of db, might get Dulek to explain it to me again. Hopefully I'll understand this time09:57
*** mvk has quit IRC09:57
*** ankit_ag has joined #openstack-cinder09:58
*** haomaiwang has quit IRC10:01
*** haomaiwang has joined #openstack-cinder10:01
*** openstackgerrit has quit IRC10:02
*** salv-orl_ has joined #openstack-cinder10:03
*** EinstCrazy has quit IRC10:03
*** openstackgerrit has joined #openstack-cinder10:03
dulekDuncanT: Ah, there's no problems with DB part I think.10:04
DuncanTdulek: I know you've said that before. I can't remember / don't understand how adding new columns works safely though10:04
DuncanTdulek: When does the migrate get run? Before updating the code?10:05
dulekDuncanT: Before.10:05
dulekDuncanT: This is what dansmith and other Nova guys use to judge if something will block the DB or not.10:05
*** salv-orlando has quit IRC10:06
dulekDuncanT: Notice "Allows Concurrent DML" - this is probably most important column.10:06
*** mvk has joined #openstack-cinder10:06
DuncanTOk, so there'll potentially be some db records without that field filled in with anything. I guess we just check the code handles that case ok?10:06
dulekDuncanT: If we have "No" there, this means that during the migration table will be locked, so some requests during the migration may fail in a strange way.10:07
dulekDuncanT: Let me double check, but I think I've looked into it from this perspective.10:07
DuncanTI'm slowly getting my head around it :-)10:07
dulekDuncanT: "Additive Schema-only Migrations" is the section explaining this a little.10:08
*** lpetrut has joined #openstack-cinder10:10
*** lpetrut has quit IRC10:10
dulekDuncanT: Okay, so 064 migration ( is only adding a column with NULL in there by default. Which is fine, as scalable backups code will just assume that nothing is restoring that backup.10:10
dulekDuncanT: And if backup service will be shut down gracefully to be upgraded we're 100% sure that no restore was in progress before the upgrade.10:11
DuncanTThe API will still be running, so things will still go into status restoring... but that field is set in the backup manager so it will get set when things come back10:12
dulekDuncanT: Exactly.10:13
DuncanTThe only issue will be with hard kills of c-bak to do the upgrade, which might leave some thing in state restoring/backingup, which will need to be reset by hand since the new cleanup code won't see them10:13
DuncanT(the old cleanup code would have reset the states on startup)10:13
*** esker has joined #openstack-cinder10:13
dulekDuncanT: You're right, but that simply isn't the way to correctly do the upgrade.10:14
DuncanTWe can release note that case I guess.10:14
dulekDuncanT: I believe we have a lot of other places where hard-killing a service produces orphaned resources.10:14
DuncanTBackups are the hardest to drain during upgrade, since they can take hours10:14
DuncanTSo well worth a release note, since it is a change in behaviour10:15
dulekDuncanT: That's a fair remark. I'll set myself a reminder to write a note.10:15
dulekDuncanT, lixiaoy1: So now we just need to take care of correct scheduling in the backup.api if not all c-bak services are yet upgraded.10:17
*** esker has quit IRC10:18
duleklixiaoy1: Can I try to rework this patch to include all the stuff related to bp rpc-object-compatibility we've merged? It should be really simple then.10:18
*** baojg has quit IRC10:28
*** tpsilva has joined #openstack-cinder10:36
*** aarefiev has quit IRC10:37
openstackgerritJordan Pittier proposed openstack/cinder: Fix race condition in RemoteFS create_snapshot_online
openstackgerritMerged openstack/cinder: Add restore_volume_id in backup
*** meomik has joined #openstack-cinder10:42
*** Poornima has quit IRC10:43
meomikhi there, i have to cinder volumes (let say A, B) and i need to merge all data into one (A -> B or B -> A), how can i do that? do i need to make a small 'hack' in database?10:43
dulekmeomik: Hacking the DB won't help you, Cinder does not do anything with data plane.10:47
dulekmeomik: You can just attach both volumes to the VM and use dd command to do anything you want.10:47
*** sdague has joined #openstack-cinder10:49
*** shyama_ has joined #openstack-cinder10:53
*** shyama has quit IRC10:55
*** shyama_ is now known as shyama10:55
*** sheel has joined #openstack-cinder10:58
*** dims has joined #openstack-cinder10:59
*** esker has joined #openstack-cinder11:00
*** haomaiwang has quit IRC11:01
*** marcusvrn_ has joined #openstack-cinder11:01
*** haomaiwang has joined #openstack-cinder11:01
*** lprice has joined #openstack-cinder11:03
*** lprice1 has quit IRC11:03
*** jwcroppe has joined #openstack-cinder11:03
*** yhayashi has quit IRC11:03
*** esker has quit IRC11:05
*** jwcroppe has quit IRC11:07
*** smoriya_ has quit IRC11:08
*** tobe has joined #openstack-cinder11:11
*** aarefiev has joined #openstack-cinder11:11
*** esker has joined #openstack-cinder11:15
*** shyama_ has joined #openstack-cinder11:17
*** jkraj has joined #openstack-cinder11:18
*** esker has quit IRC11:20
*** shyama has quit IRC11:20
*** shyama has joined #openstack-cinder11:21
*** ociuhandu has quit IRC11:22
*** shyama_ has quit IRC11:22
*** boichev2 has quit IRC11:23
ankit_agDuncanT: Hi11:27
ankit_agDuncanT: could you please take a look on when you get time11:27
*** meomik has quit IRC11:27
*** boichev has joined #openstack-cinder11:30
*** dims_ has joined #openstack-cinder11:30
*** dims has quit IRC11:31
openstackgerritwanghao proposed openstack/cinder: Clean up image tmp file if c-vol gets restarted
DuncanTankit_ag: Will do11:42
ankit_agDuncanT: Thank you11:42
e0neDuncanT: hi! I hate to ask somebody to review my patches, but deadline is too close for me. If you'll have a time, please take a look on - attach/detach features for brickclient-extension11:44
*** lpetrut has joined #openstack-cinder11:45
*** houming has quit IRC11:45
DuncanTe0ne: You're up against the library deadline, right? I'll get to that one soon11:46
e0neDuncanT: TBH, I'm talking about clients deadline and would get as much feedback, as possible before vacation11:47
e0neI was wating for privsep in os-brick, but it's not ready and I start use rootwrap for now:(11:47
DuncanTI'm kind of glad that privsep stuff isn't going to make it TBH, I don't like the code11:48
DuncanTNeed to spend some more time on it, figure out what it is I don't like11:48
e0neit's a choise between bad and bad11:49
DuncanTYeah, but it's old bad .v. introducing a new bad... we might as well get privsep right rather than having to redo it again later11:49
e0neI don't like rootwrap implemetation - it's too implicit11:49
e0neI don't know it there is a good English proverb for phrase: good known old one better than not known a new one11:51
DuncanTIt's full of security holes, which makes it kind of pointless... it's no better than sudo rights anymore11:51
DuncanTThe proverb for that would be 'Better the devil you know than the one you don't'11:52
DuncanTUsually shortened to 'Better the devil you know' in common usage11:53
DuncanTOrigin is apparently 16th centiry Ireland11:54
e0neoh.. I've just realized that it's a first proverb which I know the time of appearance/birth11:57
*** EinstCrazy has joined #openstack-cinder11:59
*** haomaiwang has quit IRC12:01
*** haomaiwa_ has joined #openstack-cinder12:01
*** salv-orl_ has quit IRC12:03
*** raildo-afk is now known as raildo12:09
openstackgerritVictor Stinner proposed openstack/cinder: Test middleware test_faults to Python 3
*** erlon has joined #openstack-cinder12:11
*** laughterwym has joined #openstack-cinder12:22
*** laughterwym has quit IRC12:22
*** laughterwym has joined #openstack-cinder12:22
*** ociuhandu has joined #openstack-cinder12:23
*** laughterwym has quit IRC12:23
*** laughterwym has joined #openstack-cinder12:23
*** salv-orlando has joined #openstack-cinder12:26
*** dave-mccowan has joined #openstack-cinder12:26
*** haypo has joined #openstack-cinder12:27
haypohey, i have some issues with Brocade CI, see my latest comment at
haypo"Brocade Fibre Channel Zoning HTTP error: Error while authenticating with switch: Brocade Fibre Channel Zoning HTTP error: Authentication failed, verify the switch credentials, error code -13"12:28
*** JoseMello has joined #openstack-cinder12:28
haypodo you know who own this CI and can investigate?12:29
dulekhaypo: I've found "" at
dulekBut I don't know if this is the page we're supposed to look at.12:34
haypodulek: ok, i will send an email ;)12:36
*** ankit_ag has quit IRC12:39
*** ankit_ag has joined #openstack-cinder12:40
*** yangyapeng has joined #openstack-cinder12:46
*** jseiler has quit IRC12:47
openstackgerritMichal Dulko proposed openstack/cinder: Add devref on rolling upgrades
openstackgerritMichal Dulko proposed openstack/cinder: Add devref on rolling upgrades
dulekRolling upgrades docs for developers are up for review - we've talked about that on the mid-cycle.12:50
*** haomaiwa_ has quit IRC12:52
*** timcl has joined #openstack-cinder12:52
*** julim has joined #openstack-cinder12:57
openstackgerritGorka Eguileor proposed openstack/cinder: Rtstool sets wrong exception message on save
openstackgerritScott DAngelo proposed openstack/cinder: cinder-api-microversions code
*** avishay has joined #openstack-cinder13:05
avishaysmcginnis: Hey! When do newton specs open?13:06
dulekavishay: I think you can simply add newton folder with your spec.13:08
*** EinstCrazy has quit IRC13:09
smcginnisavishay: The folder is added, so submit away. :)13:09
*** EinstCrazy has joined #openstack-cinder13:09
scottdasmcginnis: Is there a good way to get visibility for this Grenade review?
openstackgerritMichal Dulko proposed openstack/cinder: Add ConsistencyGroupVolumeTypeMapping table
scottdaOr is it just to nag the right people? (#openstack-qa I think)13:11
dulekscottda: #openstack-dev is a better place from my experiences.13:12
scottdadulek: cool, thanks13:12
dulekscottda: You may try to ping dansmith or sdague.13:12
scottdaok, they're online a bit later, I think.13:12
*** e0ne has quit IRC13:13
scottdadulek: You could review that ^^, if you've time. Just a grenade change for micorversions.13:13
sdaguescottda: typically #openstack-qa, though I saw the ping now13:13
dulekscottda: I want to review whole microversions stuff today, so sure.13:14
*** e0ne has joined #openstack-cinder13:14
sdaguealso, I don't understand why you are adding a new endpoint for it13:14
scottdacool, thanks13:14
duleksmcginnis: I think you should also add newton's specs here:
smcginnisscottda: Yeah, dev is probably a good starting place.13:15
scottdasdague: This was advice from manila. So a user didn't hit the old (/v2) endpoint with a microversion header, and an older server that did not understand microversions would fall back to default behaviour, causing a silent failure.13:15
smcginnisscottda: I think the comment was how it gets updated.13:15
scottdaMeaning why grenade needs this?13:16
smcginnisdulek: Yeah, I was going to wait to link it in to the index.13:16
smcginnisI suppose I don't have to.13:17
smcginnisJust worried someone looking during the Mitaka release might not understand and think anything in Newton already made it.13:17
sdaguescottda: it is expected the user actually hits the version list up front before that do that13:17
duleksmcginnis: Well, maybe it's easier to read while Mitaka's still in development.13:17
sdaguehonestly, I think sticking on the same endpoint will provide a much smoother transition13:17
smcginnisdulek: True. I can add it.13:17
sdaguegiven that it has taken 3 years to go from 1 -> 213:18
smcginnissdague: That was our original direction.13:18
smcginnissdague: But it was pointed out by manila that that is dangerous.13:18
scottdasdague: Perhaps. But we discussed this at the Cinder mid-cycle and that's what people wanted.13:18
sdagueok, I think manilla is wrong in that regard13:18
sdaguebut so be it13:18
smcginnissdague: Hah, maybe. But we talked through some scenarios that changed my mind at the time at least.13:19
scottdasdague: I originally had your understand and implemented it that way, so would've been great with me to stick with /v2. But here we are.13:19
smcginnisscottda: Wasn't the issue of new parameters silently getting ignored if it was a pre-microversion v2 endpoint?13:20
sdagueright, it's just going to be 3 years to get things sorted as you'll need another endpoint13:20
smcginnisSo you could end up with a successful response, but not get what you requested.13:20
scottdasmcginnis: Yes, that was brought up as a/the problem use case13:20
sdaguesmcginnis: sure, but that's not really different than today right?13:21
sdaguedo you have strict param validation13:21
smcginnissdague: I'm not sure if we do everywhere.13:21
scottdasdague: the issue today is that we can't/don't add new parameters...13:21
dulekscottda: But isn't the microversion header returned, so user can validate the response by himself?13:21
*** haomaiwang has joined #openstack-cinder13:22
scottdawe tried that with passing host+instance_id during nova volume_attach and it blew up when used with old (liberty) cinder..13:22
sdaguescottda: sure, but if someone throws ?foo=bar onto a thing, do you explode?13:22
scottdaand we had to back out that change for multi-attach because of this.13:22
scottdadulek: Yes, the user could/should validate the header, and if there is not one user would know that the server did not understand micorversions....13:22
sdagueor silently drop 'foo'13:23
smcginnisdulek: Could be too late at that point.13:23
scottdabut there are those amongst us who take up the arguement that we must protect against the users who do things improperly.13:23
dulekSure, I understand both concerns.13:24
duleksdague: I think we silently drop it.13:24
smcginnissdague: I think right now we silently drop foo. Which would be a problem then.13:24
scottdasdague: I don't remember the details on the multi-attach issue, but I think we explode.13:24
sdaguedulek: right, so the API currently already does silent ignore13:24
scottdaat least for adding the host+instance_id. and I believe it was an optional parameter. This was nova-> cinder via the cinderclient.13:25
smcginnissdague: So if we silently ignore the new params then (assuming the response isn't checked for the header) we wouldn't do what was being requested.13:25
avishaysmcginnis: cool thanks!13:25
sdaguehere is the trade off. Do you assume people that know enough to integrate microversions and not actually check the server does it.13:25
smcginnisBut still return success with a different operation actually executed than what they wanted.13:25
sdagueat the cost of having to keep the /v2 endpoint forever13:26
smcginnissdague: Take the caveat emptor approach?13:26
sdaguebecause if /v2 supports microversions13:26
sdaguethen old scripts keep working on /v213:26
sdagueand they work forever13:26
*** andymaier has joined #openstack-cinder13:26
sdaguea new stuff works forever as well13:26
scottdasmcginnis: sdague Personally, I say caveat emptor. But that's not how opensource works....if someone(s) disagrees, we get to the point we are at now, with me having just implemented /v3 endpoint for this, and late in the release at that.13:27
sdaguebut the cost of changing out that url part is huge13:27
*** mylu has joined #openstack-cinder13:27
sdaguescottda: I realize you all are down this path. I'm saying make sure you *really* *really* *want* to do it this way13:27
sdaguebecause the downstream cost of this is huge13:28
*** nkrinner has quit IRC13:28
smcginnisThe thought was, the switch to /v3 would be the last. Ever. /v2 clients just keep on working. It would take a while to transition for sure.13:28
*** nkrinner has joined #openstack-cinder13:28
smcginnisAnd it would make it very explicit when to switch to microversions.13:28
sdaguesmcginnis: by a while, we're talking 3 - 5 years13:28
smcginnissdague: Or double that. ;)13:28
sdaguethe whole point of microversions was to not have to be explicit to switching :)13:29
*** mylu has quit IRC13:29
sdaguenot in the url at least13:29
smcginnisbswartz: You around?13:29
smcginnissdague: Yeah, fair point.13:29
sdagueok, anyway, I'll leave you to it. I still think it's worth making sure this is how you really want to do this.13:30
smcginnissdague: Thanks, good to hear a voice arguing the other side.13:30
scottdasdague:Your advice is much appreciated13:30
*** avishay has quit IRC13:30
smcginnissdague: That was our original approach until the midcycle.13:30
scottdaThat was how the code looked when it was first posted (before Tokyo summit )13:31
smcginnisscottda: Wow, it has been that long, hasn't it?13:31
scottdaso, don't take this as whining, it's really a philosophical question: How can we (openstack community) get to solutions and answers more quickly?13:32
scottdaMaybe this is as good as it gets, I don't know.13:32
scottdaBut for microversions in mitaka, it's either merge soon or punt to Newton. And waiting might be best...13:33
sdaguescottda: well, part of the issue is not realizing something was up for debate. As someone that had a bit part in carving out this whole concept and getting it into Nova, I didn't realize there was something up for debate on the cinder side for implementation.13:33
sdaguewas there an ML thread on this?13:34
duleksdague: Just the mid-cycle summary I think.13:34
scottdasdague: no, there was no ML thread. This came up in Jan at our mid-cycle.13:34
scottdaso, maybe that's my answer to how to get better communication.13:34
sdagueyeh, that's the blessing and curse of midcycles13:35
smcginnisSeems like microversions maybe should have been a cross project spec to make sure we were all in agreement on how to implement it.13:35
scottdaBut within Cinder, there wasn't much debate. The new endpoint was brought up, and we all kind of agreed.13:35
sdagueyou get a lot of people in the room to discuss things fast13:35
sdaguebut it assumes everyone that has information that's relevant is in the room13:35
sdaguesmcginnis: we're working through api-wg guidelines on that right now13:36
johnthetubaguysmcginnis: you took the words from the keyboard, might still want one so we all use a similar pattern, actually do we have an API wg doc on this? we probably want to create one so we agree a general pattern13:36
johnthetubaguyhmm, I can't type, but the message is in there13:36
smcginnisjohnthetubaguy: I got ya. ;)13:37
smcginnissdague: Do you have a link?13:37
sdaguethe thing is, up until this point, the model was copy nova, which is what ironic and manilla largely did. But there were some varients done in the process13:37
smcginnisI would be hesitant to merge something this late in the cycle knowing there's questions on the implmentation.13:37
smcginnisEspecially if it ends up counter to what the api-wg states.13:37
scottdasmcginnis: I agree13:37
sdaguesmcginnis: it's probably not going to answer the questions you have13:37
johnthetubaguysdague: true13:37
sdaguebecause this idea of forcing a new endpoint is not something that I'd ever heard of before13:38
sdaguein Nova we kind of had to because the code stacks were different13:38
*** abhishekk has joined #openstack-cinder13:38
sdaguebut it's a regret13:38
smcginnissdague: Maybe I can raise a question on it. Or point bswartz at it to explain why they did that in manila.13:39
johnthetubaguysdague: ah, good point, that was lots in translation13:39
scottdasmcginnis: I'm all for slowing down to think on this. Which means I'm about to WIP the microversions patch.....13:39
*** edmondsw has joined #openstack-cinder13:39
smcginnisscottda: I'm kinda thinking that might be better at this point.13:40
smcginnisscottda: At least for a bit while we stew on this.13:40
sdagueI think that's also the thing that we should realize is that this is not a completely solved space. So there is no "this is exactly how to do this" yet. It requires discussion with folks that have done it.13:40
abhishekkDuncanT: hi you around?13:40
smcginnissdague: True. I just hope that however we all implement it, we are _mostly_ consistent in the usage.13:40
johnthetubaguysdague: its similar with upgrade, in some ways13:41
*** edmondsw has quit IRC13:41
*** edmondsw has joined #openstack-cinder13:42
scottdaok, patch is WIP'ed . Thanks for the input, gentlemen.13:42
smcginnisscottda: Thanks man.13:42
scottdaI fell like the kids just went off to camp. What do I do now?13:42
* dulek feels a little guilty for triggering this…13:43
scottdadulek: No way! Much better to figure this out now.13:43
smcginnisdulek: It's better we have the discussion now than in a few weeks.13:43
scottdaThis is not some simple patch that we can just revert if we find a bug.13:43
* smcginnis goes off to find our midcycle video13:43
dulekOkay. Anyway I can go back to scalable backups then…13:44
smcginnisDang we have too much video. Should edit these down to useful bits.13:44
scottdaI never liked the idea of adding a new endpoint. And I'm firmly in the Caveat Emptor camp for dealing with users. We don't need to bullet-proof everything we do .13:44
scottdaThat's kinda the point of microversions. If you want them, ask for them and deal with them properly. If you don't want them (or know about them), don't do anything and get the same, original behaviour.13:45
smcginnisscottda: It would certainly simplify some things.13:46
smcginnisAnd if that's the approach nova has taken, it wouldn't be inconsistent with one of the other core projects.13:47
*** ankit_ag has quit IRC13:47
*** jgregor has joined #openstack-cinder13:47
*** xyang1 has joined #openstack-cinder13:47
*** krtaylor has joined #openstack-cinder13:47
*** xyang has joined #openstack-cinder13:48
scottdaYeah, I'll take the blame for not being a better communicator about this. I think up to the mid-cycle people still didn't understand the guts of how this worked, so the was a lot of FUD about it.13:50
dulekWhat are the next steps?13:50
dulekWe're back to the whiteboard to get it right in Newton?13:51
dulekOr we're still targeting Mitaka?13:51
scottdadulek: I have a hard time imagining we can get to consensus in the next 12 days.13:52
smcginnisI would say it's not completely off the table for Mitaka.13:52
smcginnisBut the risk is high at this point.13:52
scottdadulek: I'm not sure how much whiteboard is needed, I'll start with an ML post on this issue, and we can see....13:52
*** mylu has joined #openstack-cinder13:53
smcginnisscottda: That's probably a good idea. It would be good to get varied input on this.13:53
DuncanTabhishekk: Hi13:53
scottdasmcginnis: dulek It's be nice to gather ideas on the implications of pushing this out to Newton. What items are dependent or blocked by microversions?13:53
*** jwcroppe has joined #openstack-cinder13:54
DuncanTI remember bswartz having a good point at the midcycle, but I forget what it was. I'll read the notes.13:55
smcginnisSeems like half the things we discussed at the midcycle were dependent on microversions.13:55
smcginnisDuncanT: Yeah, I was hoping to find the youtube video to refresh on that.13:55
*** martyturner has joined #openstack-cinder13:56
smcginnisDuncanT: But pretty sure the main point being a microversion aware client talking to a pre-microversion service.13:56
smcginnisDuncanT: And that older service ignoring new things and returning success even though the user didn't get what they asked for.13:56
scottdasmcginnis: DuncanT That was the argument. But I think it defeats the point of microversions if you have a user who says "Yes, I want microversions, but I don't want to be responsible for checking that they are supported"13:57
smcginnisscottda: Yeah. I hate to flip flop - because I was behind /v3 at the midcycle - but I think you're right.13:58
abhishekkDuncanT: sorry to bother, I was unaware ankt_ag has already requested you to review request-id patches13:59
scottdasmcginnis: Things we discussed for Newton were blocked by microversions, but with my (eternal) optimism I believe that microversions will get in for early Newton, ahead of anything else. The question is: Is there anything in Mitaka that would be blocked by this?13:59
smcginnisscottda: Definitely early N.14:00
DuncanTabhishekk: I've got the in my queue, but lots of other things too14:00
dulekCheescake's fine without microversions, right?14:00
smcginnisI remember in Tokyo we didn't have an immediate need for it, but then it seemed like at the midcycle it was the panacea for all issues.14:00
*** haomaiwang has quit IRC14:01
scottdadulek: I think so. It's a new api, right? Not breaking anything or backwards-incompatible?14:01
*** mylu has quit IRC14:01
DuncanTsmcginnis: We've a bunch of stuff queued up that changes APIs in various ways14:01
*** haomaiwang has joined #openstack-cinder14:01
smcginnisDuncanT: For M? Do you know which things?14:01
abhishekkDuncanT: yes. thank you and sorry again14:01
DuncanTdulek: I think that's ok since the old replication API never worked14:01
smcginnisDuncanT: Some may argue that, but I think we're past that point. ;)14:02
DuncanTsmcginnis: There are some queued for M I'm sure. I'll see what I can find14:02
*** rlrossit has joined #openstack-cinder14:03
smcginnisDuncanT: That's the feeling I have too, I just can't think of a specific case.14:03
scottdaAnything we would have changed previously in the API does not need microversions. It is mainly for things that could break an older deployment.14:03
duleksmcginnis, scottda: Maybe we should start with an ML thread so anyone would be able to chime in. We'll also signal developers who were dependent on microversions.14:04
scottdaIf you need a good primer on microversions, some smart guy wrote a blog about it:
duleks/We'll/This will14:04
scottdadulek: Yes, i'll work that up.14:05
xyang1scottda: replication api is in contrib, even if microversion is merged, it cannot be used on contrib until we move everything to core api14:05
xyang1scottda: we had that discussion during spec review14:06
xyang1scottda: smcginnis the decision back then was not to block contrib api changes for microversion14:07
scottdaxyang1: you mean replication+microversions cannot work until everything is moved to core, right? features on contrib will still work as they have in the past...14:07
xyang1scottda: yes14:07
*** dave-mccowan has quit IRC14:09
bswartzDuncanT: the issue with microversions and the URL is when a new client calls an microversioned API on an old server, you want it to fail immediately instead of having undefined behavior. Because old servers will ignore everything in the header related to microversions, the only way to ensure failure is to use a different URL path.14:09
DuncanTbswartz: That was it, thanks14:10
*** dave-mccowan has joined #openstack-cinder14:10
DuncanTSo we really do want a V3 endpoint, otherwise every client needs an 'are microversions supported' round trip14:10
bswartzand because clients are often stateless, that round trip could be needed very frequently14:11
scottdaDuncanT: But you stated the possible alternative, every client needs a 'are microversions supported' round trip14:11
scottdaThat is the point of micorversions14:11
scottdathe client needs to know that they are supported, else you cannot use them14:11
dulekDuncanT: sdague argumented that it's no different that silently ignoring URL params.14:11
dulekDuncanT: Which we're doing now.14:12
scottdaAnd we can do some smarts in python-cinderclient, like caching , to prevent round-trips after the first one.14:12
bswartzscottda: the point of microversions is to allow clients and servers coded for different versions to find an API definition they agree on exactly14:12
smcginnisbswartz: Don't they need that round trip to find that agreed on version?14:13
duleksmcginnis: +114:13
bswartzmcginnis: not in the general case14:13
scottdabswartz: Right. And that means the client will have to understand the reply from the server. IF the server does not respond with a supported micorversion, the client can understand that the server does not support microversions.14:13
bswartzif the server supports the version the client wants to use, then it succeeds autoamtically14:14
bswartzif the server doesn't support, then the client should receive an error with details about what it does support, and the client can adjust accordingly14:14
scottdabut the client can not just assume that success means the server supported the version that is asked for. The client must verify , if the client cares enough.14:14
scottdabswartz: I disagree. the client that understands microversions must be aware that some (older) servers do not understand microversions.14:15
DuncanTIf we always include a header in the microversion response, then the client knows it is talking to an old server if that header is there14:16
bswartzscottda: I disagree -- the client should assume that a positive response means that the API call was successful -- microversions gives us an efficient way to handle this negotiation14:16
DuncanTs/is there/ is not there/14:16
bswartzscottda: yes obviously older server don't support microversions, therefore you detect that by sending microversioned requests to a different URL so that old servers are guaranteed to return 404 not found14:17
*** akerr has joined #openstack-cinder14:17
bswartzit's more efficient than always asking ahead of time what the server supports14:17
scottdabswartz: If the client wants to be efficient , they can go with old behaviour: no microversions, no new API changes.14:18
scottdaIf the client wants the latest-and-greatest api behaviour, and the client is not sure what the other end of the deployment supports, it must check the return.14:19
bswartzeventually that will be untenable though14:19
bswartzImagine if every web browser in the world started with HTTP 0.9 and negotiated upwards to HTTP 1.114:20
bswartzall the extra round trips would be hugely wasteful, and nobody uses HTTP 0.9 since the mid 90s14:20
bswartzsomeday the existing cinder v2 API will be obsolete14:21
smcginnisYou would negotiate down, wouldn't you?14:21
scottdaWhat is the difference in efficiency with 1) hitting new /v3 endpoint, which is not supported, and getting 404 and then 2) having to fall back to old /v3 endpoint vs. 2) asking for microversion and getting a response that shows ignorance of microversions and then 2) having to fall back to old, non-micorversionsed behafiour ?14:21
bswartzit's better to start with the highest version you support and to have a way to negotiate down14:21
bswartzit's dumb to start with the lowest version and negotiate up -- it guarantees bad performance in the common case14:22
smcginnisbswartz: So your web browser example isn't valid? :)14:22
bswartzsmcginnis: web browsers do start from the highest version they support14:22
scottdaThe server either does not understand microversions or returns the highest version it does understand.14:22
*** archers has joined #openstack-cinder14:22
openstackgerritMerged openstack/cinder: Delete unuseful code in Huawei driver
scottdaThe client also starts with the highest version supported/desired. The server responds with either 1) no microversion (existing/Liberty server) or 2) compute fault 406:14:24
scottda{"computeFault": {"message": "Version 3.8 is not supported by the API. Minimum is 3.0 and maximum is 3.3.", "code": 406}}14:24
scottdathe client can re-send from there.14:25
bswartzscottda: and if the server is liberty?14:25
scottdaIf the client does not wish to go through this, they can just not use microversions.14:25
scottdaThe liberty server won't have a microversion response in the reply14:25
* DuncanT can understand bswartz's point... sending a microversioned request to an old server might cause something unwanted to be done by the old server, unless we're really careful with our API changes (which, realistically, we're not going to manage)14:26
scottdaSo the client will know that it is pre-microversions and deal with it.14:26
*** diogogmt has quit IRC14:26
*** mriedem has joined #openstack-cinder14:26
DuncanTBy the time you get the response, it is too late14:26
bswartzduncant: yes14:26
DuncanTSo you'd have to probe to see if microversions are supported every time14:26
scottdamicroversions has a new API to determine supported versions, including microversions14:26
DuncanTWhich is a performance killer14:26
scottdaBut how often is your server deployment going to change?14:27
DuncanTscottda: Most clients don't cache stuff14:27
DuncanTscottda: So there'll have to be lots of (slow) probes if we don't go V314:27
scottdaI think it's ironic that has implemented some caching in their client, so this does not have to be discovered for each call.14:27
bswartzscottda: that approach works from a correctness standpoint it's just very slow compared to simply sending the request you want to send and hoping it succeeds (if the server supports it) or fails (if the server is too old) and then retrying with a lower version14:27
DuncanTbswartz: ++14:28
*** diogogmt has joined #openstack-cinder14:28
bswartzthe common case by far will be that it will succeed14:28
bswartzjust like my web browser example14:28
scottdaDuncanT: I've planned on implementing caching for cinderclient. It uses a file, so it will work with stateless clients. Expires whenever you want (2 minutes, 24 hours, 1 week...)14:28
bswartzevery web server in the world understands HTTP/1.1 today -- but back in 1998 that wasn't true14:29
openstackgerritWilson Liu proposed openstack/cinder: Match the ip more accurately in Huawei driver
*** mylu has joined #openstack-cinder14:30
bswartzcinder today is where the web was in 199814:30
*** gouthamr has joined #openstack-cinder14:30
bswartzlet's not build something we'll regret in 10 years14:30
scottdaWell every openstack cinder server in the world will understand microversions some day. Until then, you deal with the possibility that your server may not.14:30
DuncanTscottda: pyton-cinderclient is far from the only client out there, in fact servey shows it is in the minority IIRC14:30
*** baumann has joined #openstack-cinder14:30
*** baumann has left #openstack-cinder14:30
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Don't show chap password in log
scottdaWhat might we regret in 10 years? All servers will understand microversions by then.14:31
dulekDuncanT: Actually it's 85% for OpenStack clients.14:31
smcginnisWe'll need caching either way. Fail, retry. Either fail a microversion and downgrade. Or fail a /v3 call and downgrade to /v2.14:32
dulekDuncanT: Python's libcloud is used by 24%.14:32
scottdaDuncanT: OK, well, older clients who want microversions will need to change to use them. They will need to understand what they are doing, else they shouldn't be doing it. Caveat Emptor. If the other clients don't want to, or don't explicitly change to use microversions, everything is the same as today.14:32
dulekDuncanT: But then - it's 21% of "Wrote our own"14:32
smcginnisscottda: Right. libcloud will just keep working.14:32
DuncanTscottda: But you're requiring them to have caching to be efficient if you don't go the V3 route14:33
scottdaRead sdague post:
bswartzscottda: when do you decide it's safe to stop calling the query microversion API?14:33
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Don't fail when port group does not exist
DuncanTThey're going to require microversions to access new features14:33
bswartzafter 1 year? after 3 years?14:33
*** eharney has quit IRC14:33
DuncanTSo you're changing the API requirements in a way no other service has actually done14:33
bswartzyou never know for sure that all the non-microversioned servers are gone14:33
scottdaDuncanT: They will be no more efficient with v3 route. They will still 1) try v3, which fails 2) fall back to v2. What's the difference between that and figuring out the same thing when they hit a v2 enpoint?14:34
bswartzthere are crazy people running openstack Essex still today14:34
DuncanTThey mostly won't fail - v3 will be in most places within a year14:34
*** knikolla has joined #openstack-cinder14:34
smcginnisDuncanT: That actually breaks libcloud then.14:35
*** mtanino has joined #openstack-cinder14:35
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Check the QoS status before we use
scottdawell, if the server has v3 they would also have v2+microversions14:35
DuncanTThe difference is, in the V3 fast path case, which is the future, they just call the API and version they need for the operation, which will be there14:35
smcginnisIf we use /v2 then not asking for new versions will just work.14:35
DuncanTIn your case, they have to query *every time*14:35
smcginnisDuncanT: You still need to query.14:35
DuncanTsmcginnis: Why?14:35
*** cknight1 has quit IRC14:35
openstackgerritWilson Liu proposed openstack/cinder: Add description when rename LUN in Huawei driver
smcginnisDuncanT: Say I want version 3.5. Even if microversions are supported by the server, I have to check that it doesn't only support 3.4.14:36
scottdabecause the server might not support v3. If it does not, you have to fall back to v2. Same as querying if server supports v2+microversions, and then falling back to v2-microversions14:36
*** jkraj has quit IRC14:36
*** jkraj has joined #openstack-cinder14:37
DuncanTsmcginnis: Just call 3.5 - the chances are, if you're calling an API in anger, you're fairly sure it is supported. The fallback case therefore happens less often14:37
scottdaDuncanT: You can add a new feature if you re-use v2 endpoint, and not microversion it. Microversion is only required if it is a breaking API change.14:37
smcginnisbswartz: If I remember right, the main thing you were saying is we could do a create request for v3.5 and if 3.5 is supported, no worries.14:37
openstackgerritWilson Liu proposed openstack/cinder: Huawei: Ensure the port is online
DuncanTscottda: Catching 'breaking' API changes is damn difficult. Go look at our history.14:38
smcginnisbswartz: But if we call the 3.5 create and it isn't supported, and it isn't microversion aware, it could do a create differently than what we expect.14:38
*** mvk has quit IRC14:38
smcginnisbswartz: That was the crux of the argument for /v3, right?14:38
DuncanTscottda: Ignoring a parameter in the request can have unwanted side effects - we've already had bugs around that14:38
bswartzsmcginnis: yes I'm not in favor of calling a what-do-you-support API before every API request, even with caching14:38
scottdaDuncanT: IF you add a parameter, bump the microversion14:38
bswartzcaching can't be used everywhere and it introduces a new class of bad behavior when the server version changes between 2 requests14:39
DuncanTscottda: Which goes back to the making everybody slow arguement14:39
*** mvk has joined #openstack-cinder14:39
*** porrua has joined #openstack-cinder14:39
DuncanTscottda: You're going round in circles14:39
smcginnissdague: If you're still around - how does nova client handle caching. Or not.14:39
*** esp has joined #openstack-cinder14:39
bswartzI believe the client should just call the API it wants and the server should always succeed if it can, and always fail if it can't handle the version14:39
openstackgerritWilson Liu proposed openstack/cinder: Add LUN wwn check in Huawei driver
*** kvidvans has joined #openstack-cinder14:42
*** diablo_rojo has joined #openstack-cinder14:42
scottdabswartz: I don't think that's controversial. Either way you have that14:42
scottdaok, I guess I see you're point...14:43
scottdawe're back to the 'silent failure' on v214:43
bswartzscottda: if we don't query the server version before sending, for example, a create-snapshot API, then it's possible that you send a super-spiffy-new v3 snapshot creation option and the server silently ignores it if it's plain v214:43
smcginnisscottda: I think we need a ML post for sure. If Nova is not going this route, I'd like to know how they are handling this.14:43
DuncanTIt's not just a silent failure, it could well be that something happened that you didn't mean14:43
bswartzI'd rather that the server returns 404 not found so the client knows that super-spiffy snapshot creation options cannot be used with that server14:44
scottdasmcginnis: Yeah, I was going to look for the video reference, but I think bswartz has nicely re-iterated his arguement. I'll post something this morning.14:44
scottdabswartz: I get you. Which is what you said at the mid-cycle. Which is why we agreed. Which is why I added v3. and now here we are.....14:45
smcginnisscottda: Just watched the video again. I think it's covered here.14:45
scottdasmcginnis: can you post the link and timestamp?14:45
*** esp has quit IRC14:46
*** mylu has quit IRC14:46
smcginnisTrying to find when it actually starts.14:46
bswartzbtw the nova guys did get this right14:47
*** dustins has joined #openstack-cinder14:47
*** abhishekk has left #openstack-cinder14:47
smcginnisbswartz: What do you mean?14:47
bswartzthey changed their endpoint from /v2 to /v2.1, which while I think the choice of strings was poor, the fact that they changed it to something else guarantees that new clients talking to old servers won't get unwanted behavior14:48
*** bill_az has joined #openstack-cinder14:48
*** mylu has joined #openstack-cinder14:48
*** dave-mccowan has quit IRC14:48
smcginnisbswartz: Hmm, sdague was here earlier stating they didn't do a different endpoint.14:48
smcginnisscottda: Around the 1:20 mark.14:48
duleksmcginnis: No. He stated that different endpoint was a mistake.14:48
smcginnisbswartz: Which is what led to all of this getting rehashed again.14:49
dulek<sdague> because this idea of forcing a new endpoint is not something that I'd ever heard of before14:49
dulek<sdague> in Nova we kind of had to because the code stacks were different14:49
dulek<sdague> but it's a regret14:49
bswartzwell I disagree with sdague about that14:49
DuncanTMe too. Pretty sure we need a new endpoint.14:49
*** rajinir has joined #openstack-cinder14:50
*** eharney has joined #openstack-cinder14:50
*** huanan_L has joined #openstack-cinder14:50
*** zhonghua has joined #openstack-cinder14:50
openstackgerritPaul Belanger proposed openstack/cinder: [DNF] Testing experimental-bindep-jobs
scottdaOK, I'll craft a ML post with what I believe are the arguments and we can go from there.14:52
bswartzI think the nova guys have a strange stance on API backwards compatibility14:52
bswartzthey (some of them at least) seem to believe that even if you have microversions, you still can't change anything in the APIs14:53
*** zhonghua-lee has quit IRC14:53
bswartzI believe that microversions are the mechanism that guarantees backwards compatibility, and it's safe to add/remove things from APIs in new microversions14:53
*** huanan_M has quit IRC14:53
bswartzI worry that nova is trying to obtain the worst of both worlds when it comes to API versioning14:54
*** arecknag has quit IRC14:55
DuncanTbswartz: I think it is safe if and only if you can continue to support all of the semantics of the old API... which is tricky and limiting14:55
johnthetubaguybswartz: there are new arguments hotting up, but the sdague wrote up the principals of the idea here:
*** martyturner has quit IRC14:57
bswartzwell cinder has an unusually hard problem because not only do we want to version APIs, we want to completely rip out the attachment API14:57
bswartzjohnthetubaguy: thx14:57
*** tobe has quit IRC14:57
bswartzDuncanT: I haven't thought hard enough about whether it's possible to emulate the "old" attachment API semantics on top of a new attachment design14:57
*** cknight has joined #openstack-cinder14:58
*** diogogmt has quit IRC14:58
*** martyturner has joined #openstack-cinder14:59
*** belmoreira has quit IRC14:59
*** belmorei_ has joined #openstack-cinder14:59
DuncanTbswartz: It's really, really difficult. The only clean way I can see to do it is to set a flag on the volume record whenever it is out of state 'available', and only let either the old or the new api affect the volume at a time, until it goes back to available14:59
sheelscottda: DuncanT: smcginnis: bswartz: some of the explainations about handling different microversions are kept here:
sheelI think it has some details about what is being discussed above..15:00
smcginnissheel: Thanks!15:00
*** mvk has quit IRC15:00
*** haomaiwang has quit IRC15:01
sheelsmcginnis: welcome..15:01
*** haomaiwang has joined #openstack-cinder15:01
*** markstur_ has joined #openstack-cinder15:02
*** dave-mccowan has joined #openstack-cinder15:02
*** rhagarty has joined #openstack-cinder15:03
*** markstur_ has left #openstack-cinder15:03
sheelspecifically Use Case 7B: New Client/New Ironic: Negotiated version (user-specified)15:03
*** merooney has joined #openstack-cinder15:03
sheel        The user specifies a particular microversion (e.g. 1.15) that the client should use15:04
sheel        The client makes a connection to Ironic, supplying 1.15 as the microversion15:04
sheel        Ironic responds with a 406 Not Acceptable, along with the -Min- and -Max- headers that it can support (in this case 1.1 and 1.10)15:04
sheel        The client reports this to the user and exits15:04
DuncanTThat ironic page talks about defaulting high, which makes sense15:04
DuncanTYou can only do that with a new endpoint though15:04
scottdaThat doc also has a use case :15:05
scottdaThis assumes the "Caveat Emptor" approach. If the client attempts to use microversion and the server does not support it, the client must figure this out and fail for the user.15:06
sheelyes 3B and 7B are somehow related to what was being discussed...15:06
sheelsimply 406 if version is not supported  by server..15:07
*** esker has joined #openstack-cinder15:07
*** esker has quit IRC15:07
*** esker has joined #openstack-cinder15:07
*** asselin__ has joined #openstack-cinder15:08
*** asselin_ has quit IRC15:10
scottdaYes, but that is the debate: Put the logic in the server to protect against bad clients, or put the burden on the client to take responsibility when using microversions.15:11
*** mylu has quit IRC15:12
sheelscottda: yes, these use cases provide answer partially to same point i think,15:14
sheelactually it states shared responsibility by simply returning 406 in case client requested version which is not supported by server(using min max versions)..15:14
*** archers has quit IRC15:14
*** lcurtis has joined #openstack-cinder15:14
sheelIn case 406 is returned, server will also send back the -Min- and -Max- headers that it can support so that client can decide how to proceed15:15
sheelplease refer use case  Use Case 5: New Client/New Ironic: Unsupported Ironic version for more details15:16
scottdasheel: I don't think 7b is the controversy, it is 3b15:16
sheelscottda: hmm..15:17
*** mylu has joined #openstack-cinder15:19
*** baumann has joined #openstack-cinder15:20
*** [1]Thelo has joined #openstack-cinder15:23
*** haomaiwang has quit IRC15:25
*** Thelo has quit IRC15:26
*** [1]Thelo is now known as Thelo15:26
*** lprice1 has joined #openstack-cinder15:29
*** bill_az has quit IRC15:30
*** lprice has quit IRC15:32
*** martyturner has quit IRC15:34
openstackgerritBéla Vancsics proposed openstack/cinder: Reduced the complexity of the send_cmd method (in ProphetStor)
scottdaML posting on microversions has been sent. Have at it.15:39
smcginnisscottda: Thanks!15:39
scottdasure. I reckon if we come to some resolution we can still get this in Mitaka, either with new /v3 or old /v2. The code is written either way.15:40
*** tobe has joined #openstack-cinder15:40
smcginnisscottda: Thanks for taking this on. I'm sure it's more than it seemed at first.15:42
smcginnis"Manila already did this. I can just copy their code." :)15:42
scottdahaha...mc_nair pointed out I should just get the 'api-microversion' tattoo, since I'll be living with it forever.15:43
smcginnisscottda: You can get that when mc_nair gets his nested quota tattoo.15:44
scottdaIf you look at the liberty mid-cycle video, you can see hemnafk saying "someone needs to implement microversions" and then giving me the I blame him.15:44
*** haomaiwang has joined #openstack-cinder15:46
mc_nairnah I'm just gonna get some Russian nesting dolls that I carry around with different quotas written on them15:47
*** yangyapeng has quit IRC15:47
mc_nairlike all the kool kids are doing15:47
smcginnismc_nair: Hah, nice!15:47
jordanPI guys. I've come up with that interesting patch for an intesting bug : Fix race condition in RemoteFS create_snapshot_online  I would love some feedback. Yeah I know I submited that today, but I can"t wait :)15:50
*** j_king has joined #openstack-cinder15:50
*** haomaiwang has quit IRC15:51
*** EinstCrazy has quit IRC15:51
*** bill_az has joined #openstack-cinder15:51
duleksmcginnis, mc_nair: I don't like tattoos, but a car sticker saying "They see me rolling (upgrade)…" would probably cool. :D15:51
*** zul has quit IRC15:53
*** zul has joined #openstack-cinder15:53
mc_nairdulek: haha I like it. Not sure it'll have a big hit rate for people understanding it, but for that single person it lands with it'll be pretty epic...15:53
scottdaI'm not sure what a good API tattoo would look like, but I've a feeling thing we had one.15:53
mc_nairA good API tattoo still looks like bad decisions15:55
*** diogogmt has joined #openstack-cinder15:57
*** yangyapeng has joined #openstack-cinder15:59
*** krotscheck_dcm is now known as krotscheck16:00
*** ntpttr has joined #openstack-cinder16:00
scottdafricken' phone. I meant to say that I suspected thingee had an API tattoo16:00
*** timcl has quit IRC16:01
smcginnisI wouldn't be surprised.16:01
scottdaand also, we could use thingee 's opinion on the new endpoint for microversion debate. On the ML. Please.16:01
*** raildo is now known as raildo-afk16:02
*** belmorei_ has quit IRC16:03
*** kvidvans has quit IRC16:03
*** salv-orl_ has joined #openstack-cinder16:03
*** belmoreira has joined #openstack-cinder16:03
*** zigo has quit IRC16:03
*** zigo has joined #openstack-cinder16:05
*** salv-orlando has quit IRC16:06
*** coolsvap|away has quit IRC16:06
mriedemdiablo_rojo: mc_nair:
mriedempatrickeast: ^ per yesterday16:06
mc_nairmriedem: nice, should be useful for debugging that multipath issue16:09
*** liverpooler has quit IRC16:11
*** liverpoo1er has quit IRC16:11
*** martyturner has joined #openstack-cinder16:11
*** jungleboyj has joined #openstack-cinder16:12
*** mylu has quit IRC16:14
*** garthb has joined #openstack-cinder16:15
*** nkrinner has quit IRC16:16
*** yangyapeng has quit IRC16:17
*** raildo-afk is now known as raildo16:17
*** garthb has quit IRC16:17
*** rcernin has quit IRC16:17
mc_nairjordanP: is there a reason an oslo looping call couldn't be used instead of the while True there?16:18
*** mylu has joined #openstack-cinder16:18
jordanPmc_nair, I am not aware of any 'oslo looping call' ? What's that a new oslo-looping module ?16:19
jordanPthat one has not been annonced on the ML :p16:19
*** coolsvap|away has joined #openstack-cinder16:20
jordanPmc_nair, that is  not a master piece in term of programming... It could benefit from some refactoring anyway16:20
*** mylu has quit IRC16:20
mc_nairjordanP: haha - it's
*** mylu has joined #openstack-cinder16:22
mc_nairjordanP: not that I think it would really change the behavior, was just my initial thought looking at the code16:22
*** bill_az has quit IRC16:23
jordanPmc_nair, we want to sleep only if s['status'] == 'creating' so I am not sure we can use that oslo looping call that easilly16:23
*** aorourke_ has joined #openstack-cinder16:23
mc_nairjordanP: think you should be able to use the RetryDecorator... just throw some error that you're expecting to retry on for the creating case and some other errors that you wont retry on for the non-creating case16:25
mc_nairjordanP: if that still doesn't make sense I can take a closer look and try so I make sure I'm not just making stuff up :)16:26
*** ildikov has quit IRC16:26
jordanPmc_nair, I agree that we could do better here, but I'd like to avoid mixing refactoring and fixing a bug. (note: I am not being lazy, just that I don't have a lot of time left to spend on that issue)16:27
*** ildikov has joined #openstack-cinder16:27
eharneythat fix is nice and backportable as is, it wouldn't be if it involved a big refactor too16:28
eharneyfwiw i think that code predated use of loopingcalls in Cinder16:29
*** daneyon has joined #openstack-cinder16:30
mc_nairjordanP: certainly, I'm with ya. Was more trying to see if there was any particular reason for it (which eharney has cleared up).16:30
*** daneyon__ has joined #openstack-cinder16:31
mc_nairthanks for clearing it up for me16:31
*** timcl has joined #openstack-cinder16:31
*** jdurgin1 has joined #openstack-cinder16:31
*** haomaiwang has joined #openstack-cinder16:32
*** laughter_ has joined #openstack-cinder16:32
*** laughte__ has joined #openstack-cinder16:34
*** daneyon_ has quit IRC16:34
*** belmoreira has quit IRC16:34
*** daneyon has quit IRC16:34
*** laughterwym has quit IRC16:35
*** dramakri has joined #openstack-cinder16:37
*** laughter_ has quit IRC16:37
mriedemjungleboyj: where is this code? out of tree driver?16:37
*** laughte__ has quit IRC16:38
*** laughterwym has joined #openstack-cinder16:38
mriedemoh awesome, you have to get it through ibm's fix central site?16:39
*** jungleboyj has quit IRC16:39
*** laughterwym has quit IRC16:43
mc_nairmriedem: yea, something like that. Had trouble getting that before (and just got .pyc code)16:46
mriedemb/c you know, pyc means you can't read the source16:47
mriedemanywho, bunch of emails into the thread before it gets to me, there is basically a NoneType exception in their closed source driver - thanks for thinking of me to resolve your issue guys!16:47
*** hemnafk is now known as hemna16:48
*** xyang has quit IRC16:48
*** dims has joined #openstack-cinder16:50
*** dims_ has quit IRC16:51
*** jungleboyj has joined #openstack-cinder16:52
jgriffithmriedem: maybe you can help me convince people why that approach is a BAD thing16:54
mriedemseems pretty awesome to me16:55
mc_nairmriedem: that sounds eerily similar to something that came to me a few days ago... got to the point of getting to the compiled Python code for that and saying it would be helpful to get the original code. I was thinking decompiling wouldn't get back method names and such but reading now it looks like it would for .pyc?16:56
jgriffithmriedem: ok, guess I will seek enlistees elsewhere :)16:56
mriedemmc_nair: sounds like you spent entirely too much time on someone else's problem16:56
mc_nairmriedem: yea, should've just not said anything and made it your problem then ;)16:57
openstackgerritAlon Marx proposed openstack/cinder: Remove a vol in error state from a CG
openstackgerritIvan Kolodyazhny proposed openstack/python-brick-cinderclient-ext: Attach/detach features implementation
mriedemjgriffith: i was being sarcastic16:58
jgriffithmriedem: oh... in that case16:58
jgriffithmriedem: it's a great way to not have your code reviewed :)16:59
*** e0ne has quit IRC16:59
*** haypo has left #openstack-cinder16:59
mriedemclassic bit17:00
*** haomaiwang has quit IRC17:01
*** haomaiwa_ has joined #openstack-cinder17:01
*** esker has quit IRC17:03
*** esker has joined #openstack-cinder17:03
*** apoorvad has joined #openstack-cinder17:07
*** rhagarty_ has joined #openstack-cinder17:08
*** rhagarty has quit IRC17:10
*** ericksonsantos has joined #openstack-cinder17:11
*** esker has quit IRC17:11
jgriffithmriedem: haha..Nice17:12
*** tobe has quit IRC17:13
*** mudassirlatif has joined #openstack-cinder17:14
*** daneyon has joined #openstack-cinder17:16
*** daneyon_ has joined #openstack-cinder17:17
*** daneyon__ has quit IRC17:19
ildikovhemna: hey17:19
hemnaildikov, hey17:19
ildikovJust wanted to check in as the ML thread is getting slow and I'm not sure we got a consensus there17:20
hemnathat's what I figured as well17:20
*** daneyon has quit IRC17:21
ildikovI still like the idea of having the connector_info stored with the attachments in Cinder ;)17:21
*** mylu has quit IRC17:21
openstackgerritSheel Rana proposed openstack/python-cinderclient: Extra 'u' in output of cinder cli commands
ildikovok, back to serious, I strongly believe that as Cinder handles attachments at the moment we cannot put all the administration to Nova17:22
hemnayes, it makes sense17:22
hemnabut it's for force detach and split brain scenarios between nova and cinder17:22
ildikov*on Nova17:22
hemnait doesn't cover the case for determining if it's correct to call brick's disconnect_volume.17:22
ildikovwhy not?17:23
hemna2 completely separate issues IMHO and people are confusing the two, by interchangeably discussing them both.17:23
hemnathe connector info is the initiator information of the nova compute host17:23
hemnait has no target information at all17:23
ildikovok, I mixed it up the last time then I think17:24
hemnathere is a confusion of names and information17:24
hemnait's easy to get confused17:24
ildikovI was wondering whether we have any chance to store the target somehow17:25
hemnaconnector is the initiator side, nova compute initiator information17:25
hemnaconnection_info is what comes back from cinder's initialize_connection, which is the target information.17:25
hemna2 different things, and 2 different purposes.17:25
jgriffithhaha.. "Chris didn't actually show up"17:25
hemnaI don't think it makes a lot of sense to store the connection_info17:25
hemnawe can always ask Cinder (initialize_connection) for it.17:26
ildikovhemna: you mean connector=initiator and connector_info=target?17:26
hemnaildikov, correct17:26
ildikovuh, you really don't make the outsiders' life easy in any sense and/or extent... :)17:27
*** leeantho has joined #openstack-cinder17:27
hemnaI didn't pick those names! :)17:27
hemnaconnection_info = target17:27
hemnajust to be specific17:27
ildikovyeah, I know, it's always someone else's fault ;)17:27
openstackgerritSheel Rana proposed openstack/python-cinderclient: Extra 'u' in output of cinder cli commands
ildikovhemna: almost the same :)17:27
hemnait is!17:27
*** krotscheck is now known as krotscheck_dr17:27
ildikovis that correct that if the connection_info is different for two attachments than the connector will be different too?17:28
jgriffithmriedem: GREAT.. now I'll be watching Kids In The Hall clips all day!17:29
hemnaildikov, yes I believe so17:29
ildikovand also can the connector be different if the connection_info is the same?17:29
hemnathe connection_info has the iqn/wwn and lun id (if any) of the target.17:29
hemnaconnector is always the same on the same n-cpu host.17:30
ildikovhow do you mean always?17:32
hemnaunless the compute host changes itself, it will always be the same information17:33
hemnacat /etc/iscsi/initiatorname.iscsi17:33
hemnathat's in there17:33
hemnaalso any FC HBAs17:33
hemnaits all of the initiator side information on the host itself17:33
hemnait 'never' changes17:34
hemnaif it does, then there are other problems17:34
ildikovbut if there are multiple targets there's still one initiator?17:34
hemnatargets are irrelevant to the initiator information17:34
hemnacompletely separate information17:34
ildikovsorry for the dumb questions, I'm not really an expert of the area still :S :(17:34
hemnait's ok :)17:35
openstackgerritApoorva Deshpande proposed openstack/cinder: Tintri image cache cleanup
ildikovhemna: but when you attach a new volume to a host then a new initiator will be created for that new volume right?17:36
*** rcernin has joined #openstack-cinder17:37
hemnainitiators aren't created17:37
hemnathat information is collected17:37
hemna(connector) and passed to cinder17:37
ildikovprobably I'm missing the terms here :)17:37
hemnacinder's driver uses that initiator information to tell the backend array how/where to export a volume to.17:37
hemnathen the driver collects the target information about the volume on the array, and puts that into the connection_info, and passes that back to Nova17:38
hemnanova uses the connection_info to connect to that target (iSCSI Session logins, etc)17:38
hemnathen the volume shows up17:38
hemnaand it's discovered and passed into the vm17:38
ildikovthis is the Nova side log after disconnect_volume:
ildikovwhich is I assume the cleanup in case of a regular detach17:42
ildikovso my question is/was that when we have multiple targets exported for the same volume in multiattach case does that mean multiple rounds of cleanup as well if the attachments are on the same host?17:43
mriedemjgriffith: it's close enough to the weekend that i think you can justify slacking off for the rest of the week17:46
ildikovmriedem: that youtube video made my day :)17:47
dulekGuys, I would love to get two more rolling upgrades related commits in Mitaka. Pretty simple ones:17:47
dulek - SIGHUP handling to reload RPC version pins cache.17:48
dulek - bugfix for a RPC version we've got in.17:48
smcginnismriedem: No, he has repl to finish up. Then he can slack off.17:49
dulek(okay, there will be third to follow probably - as scalable backups are getting merged right now)17:49
*** SergeyLukjanov has left #openstack-cinder17:50
duleksmcginnis: Whoa, that was quick, thanks. ;)17:51
smcginnisdulek: Good timing before I got distracted with other things. ;)17:51
*** cknight has quit IRC17:53
*** timcl has quit IRC17:53
*** jdurgin1 has quit IRC17:58
ildikovhemna: I assume I asked a too stupid question that you disappeared :D17:58
*** mylu has joined #openstack-cinder17:59
*** haomaiwa_ has quit IRC18:01
*** haomaiwang has joined #openstack-cinder18:01
*** haomaiwang is now known as 16WAAAVR418:01
*** timcl has joined #openstack-cinder18:01
*** jordanP has quit IRC18:02
*** mriedem has quit IRC18:02
*** lpetrut has quit IRC18:02
*** mriedem has joined #openstack-cinder18:04
*** bardia has joined #openstack-cinder18:05
*** arch-nemesis has joined #openstack-cinder18:06
*** e0ne has joined #openstack-cinder18:07
*** cburgess has quit IRC18:07
*** cburgess has joined #openstack-cinder18:08
*** raildo is now known as raildo-afk18:10
*** e0ne has quit IRC18:16
*** mylu has quit IRC18:18
*** angela-s has joined #openstack-cinder18:18
*** raildo-afk is now known as raildo18:22
*** notmorgan has quit IRC18:22
*** fthiagogv_ has joined #openstack-cinder18:24
*** fthiagogv__ has joined #openstack-cinder18:26
*** porrua has quit IRC18:26
*** fthiagogv has quit IRC18:27
*** fthiagogv_ has quit IRC18:29
*** esker has joined #openstack-cinder18:29
*** 16WAAAVR4 has quit IRC18:30
*** e0ne has joined #openstack-cinder18:30
*** esker has quit IRC18:30
*** esker has joined #openstack-cinder18:31
*** vivekd has joined #openstack-cinder18:31
*** lcurtis has quit IRC18:34
*** eharney has quit IRC18:34
*** eharney has joined #openstack-cinder18:37
*** mvk has joined #openstack-cinder18:38
openstackgerritVivek Dhayaal proposed openstack/cinder: Support HA in cinder backup service
*** rlrossit has quit IRC18:41
*** bill_az has joined #openstack-cinder18:41
*** ChubYann has joined #openstack-cinder18:44
scottdasmcginnis: Is there a deadline for releasing the cinderclient18:46
hemnaildikov, in the case of N targets for N attachments to the same compute host, that means N calls to brick's disconnect_volume.18:47
ildikovhemna: in this case the connection_info could be used to identify whether to call disconnect_volume or not right?18:49
hemnathat's my assertion18:49
ildikovhemna: and by storing it in Cinder that would reduce the roundtrips18:49
hemnaif it's the same target exported on the array, the connection_info would be the same.18:49
hemnano need to store it on cinder18:49
hemnajust call cinder's initialize_connection18:49
hemnait will return the connection_info18:49
ildikovbut it's too many calls18:50
hemnaafaik, each driver will either export a new target for every export or all will the a shared target18:50
*** rlrossit_ has joined #openstack-cinder18:50
hemnayou won't have a mixed mode18:50
hemnaso comparing 2 connection_info should be sufficient to determine shared or not shared target.18:50
hemna2 calls to cinder is too many ?18:50
ildikovand also a bit confusing why someone tries to initialize the connection at detach time18:50
hemnait's only needed in the case of multiattach18:51
ildikovwe already call Cinder once to get the attachments and it's already too much18:51
hemnaand if the volume is multiattach enabled18:51
hemnaand only if that volume is attached > 1 time on that same host18:51
scottdae0ne: thanks18:51
hemnawell, if you don't want to call cinder, then all bets are off18:51
hemnaand I can't help18:51
ildikovwe have to call out to Cinder to determine whether multiattach is enable or not... :)18:51
hemnaI don't understand that logic at all18:52
hemnayou have that information in the volume object already18:52
hemnathat volume object that's already there in nova has all of the attachments18:52
ildikovI would like to be able to get all the info that's needed from the volume attachment info18:52
*** dramakri has quit IRC18:52
hemnait makes no sense for the attachments to store the connection_info18:52
hemnabecause it's 1) it will most likely get out of date (live migration) and 2) you can readily get the correct information from initialize_connection at any time.18:53
hemnathere will always be some case that causes that information to get out of date18:53
hemnacalling initialize_connection is the source of truth18:54
ildikovcan we store any flag in the volume info to check which type of back end it's on?18:54
hemnamaking 2 calls to cinder is a very small price to pay in this very specific case18:54
openstackgerritMerged openstack/cinder: Scaling backup service
hemnait won't happen on every non multiattach case18:54
hemnaso I don't see what the problem is.18:54
hemnawho is going to store that flag18:54
hemnaand make sure it's correct ?18:54
ildikovyou can check back the review comments on the Nova patches, I don't make it up18:54
hemnaok, well this is my $0.0218:55
ildikovthe volume has one back end it has to set once18:55
hemnainitialize_connection will always be correct.18:55
hemnadoing anything else is going to be a mess IMHO18:55
ildikovlook, I'm 100% with you, I try to find solution to the already existing concerns I faced with during the past 2,5 months18:56
ildikovif it's the only solution we need to have the answers to all variants why they are bad18:57
ildikovso regarding the volume back end it can be added to the volume info when the volume is created18:57
ildikovlike how the multiattach support depends on the back end today18:58
ildikovas you said we will not have mixed cases one back end has one behavior18:58
openstackgerritVivek Dhayaal proposed openstack/cinder: Support ZeroMQ messaging in cinder multibackend
ildikovregarding the connection_info if there is an API in Cinder to update it that can work also, but I accept that it's not the silver bullet either18:59
hemnaif we go the flag route18:59
hemnaeach cinder backend would have to be updated to set that flag19:00
hemnaor whatever19:00
*** ndipanov has quit IRC19:00
hemnabut Nova would still have to fetch that information from calling initialize_connection19:00
hemnaand either store it in the bdm at attach time19:00
hemnaor a detach time19:00
ildikovcannot that be added to the volume info?19:01
hemnaI'm not sure that makes sense19:01
hemnaas not every volume is multiattachable19:01
ildikovwhat's against it?19:01
hemnamost of the time that flag is unused19:01
hemnato me, that's more overhead19:01
ildikovcan a flag be optional?19:01
hemnathan calling initialize_connection twice in the case of a multiattachable volume that has more than 1 volume attachment on a host, at detach time19:02
hemnavs. every call to get any volume, has now another flag19:02
hemnawe don't have optional volume columns19:02
ildikovmultiattach itself is unused most of the time as well19:02
hemnaunless it's stored in metadata19:02
hemnawhich sux as well19:03
hemnaildikov, exactly my point19:03
ildikovbut that's just a flag still :)19:03
ildikovI mean if someone does nto need it they will not use it19:03
hemnaok, well that's my $0.02 on it all19:04
hemnaI'm kinda over it at this point19:04
hemnaeveryone wants to make this the most impossible as possible.19:04
ildikovI'm just trying to check all possibilities we might have19:05
hemnait's this kind of stuff that makes me stay away from Nova entirely.19:05
ildikovotherwise we need to have this discussion we have now in review comments during N, which will hurt even more IMHO19:06
hemnaildikov, do you have the review url handy ?19:06
openstackgerritPatrick East proposed openstack/cinder: Allow for Pure drivers to verify HTTPS requests
ildikovhemna: the Nova patches you mean?19:07
ildikova few patches are abandoned as I had to reorganize some stuff19:07
hemnawhich review specifically talks about this19:08
ildikovI also have to write up a spec for N...19:08
ildikovI don't remember exactly now which oone had the roundtrip comments :S19:08
ildikovI think it's better to have a cross-project session on the Summit to be on the same page as much as possible19:09
*** andymaier has quit IRC19:11
jungleboyjAll ... Please be aware that our account for holding 3rd Party CI logs on SoftLayer has had issues.  So, IBM's 3rd Party CI Logs are not currently accessible.  The issue is being actively worked.19:12
ildikovhemna: probably this one has some as I worked on reducing the number of roundtrips here:
*** lpetrut has joined #openstack-cinder19:12
ildikovhemna: and please don't hate me, I just want to have this done finally like you19:14
*** akerr has quit IRC19:14
*** akerr has joined #openstack-cinder19:16
*** dmz has joined #openstack-cinder19:16
*** eharney has quit IRC19:16
hemnaildikov, lol, it's ok I don't hate you19:17
dmzhowdy y'all, i'm trying to create a volume from an image and keep getting the following error: Failed to copy image to volume No space left on device; no matter what size I set the volume to (even 2x the size of the image); any suggestoins?19:17
hemnaildikov, we need to see how to make this happen19:17
hemnathere are just too many damn cooks in the kitchen that don't understand how to cook IMHO.19:17
ildikovhemna: yeah, this is why I'm annoying you now :)19:18
hemnanot that I'm a good cook or anything...19:18
bswartzdmz: did you check "image_conversion_dir" in cinder.conf?19:18
hemnaI can boil water19:18
patrickeastdmz: do you have a stack trace for it in the cinder-volume log? there are a few places it could be coming from19:18
ildikovhemna: haha :)19:18
ildikovhemna: I can do more, but you still kill all my ideas here :)19:19
dmzi'm checking conversion dir now19:20
bswartzdmz: you need to have enough free space (2x the uncompresed image size) in the filesystem that that directory points to19:21
dmzthe image isn't compressed, it's a raw image taken (dd) from a disk19:21
dmzif the image_conversion_dir is commented out where would it default to?19:21
bswartzmight be /tmp19:22
*** jwcroppe has quit IRC19:22
bswartzmight be /var/lib/cinder/conversions19:22
dmzthat might explain it; tmp is only 5 gig19:22
*** jwcroppe has joined #openstack-cinder19:22
bswartzin any case, you should set it, so you know where the temp file are going19:22
dmzand var is also limited; let me try changing that to somewhere else; thanks!19:22
*** bkumar has joined #openstack-cinder19:25
*** jwcroppe has quit IRC19:26
*** flip214 has quit IRC19:28
*** aorourke_ has quit IRC19:28
*** aorourke_ has joined #openstack-cinder19:29
*** cknight has joined #openstack-cinder19:31
*** Lee1092 has quit IRC19:32
*** haomaiwang has joined #openstack-cinder19:32
*** eharney has joined #openstack-cinder19:33
*** jwcroppe has joined #openstack-cinder19:34
*** flip214 has joined #openstack-cinder19:35
*** haomaiwang has quit IRC19:37
*** jkraj has quit IRC19:38
*** lpetrut has quit IRC19:42
*** lpetrut has joined #openstack-cinder19:46
*** angela-s has quit IRC19:55
*** rajinir has quit IRC19:55
*** rcernin has quit IRC19:55
*** hemna is now known as hemnafk19:55
*** ociuhandu has quit IRC19:58
*** krtaylor has quit IRC19:59
*** rlrossit_ has quit IRC20:00
*** zhonghua-lee has joined #openstack-cinder20:07
*** annasort_ has joined #openstack-cinder20:07
*** markvoelker_ has joined #openstack-cinder20:07
*** flip214_ has joined #openstack-cinder20:09
*** krtaylor has joined #openstack-cinder20:11
*** guitarza1 has joined #openstack-cinder20:11
*** zmd has joined #openstack-cinder20:11
*** annasort has quit IRC20:12
*** john5223 has quit IRC20:12
*** flip214 has quit IRC20:12
*** markvoelker has quit IRC20:12
*** bill_az has quit IRC20:12
*** zhonghua has quit IRC20:12
*** guitarzan has quit IRC20:12
*** ircuser-1 has quit IRC20:12
*** dmz has quit IRC20:12
*** stevemar has quit IRC20:12
*** annasort_ is now known as annasort20:12
*** vivekd has quit IRC20:12
*** stevemar has joined #openstack-cinder20:13
*** ircuser-1_ has joined #openstack-cinder20:13
*** bkumar has quit IRC20:14
*** boris-42 has quit IRC20:16
openstackgerritChuck Fouts proposed openstack/cinder: Support for iSCSI CHAP Uni-directional Auth
*** boris-42 has joined #openstack-cinder20:18
*** boris-42 has quit IRC20:24
*** ociuhandu has joined #openstack-cinder20:31
*** angela-s has joined #openstack-cinder20:38
*** e0ne has quit IRC20:45
*** e0ne has joined #openstack-cinder20:45
*** e0ne has quit IRC20:46
*** porrua has joined #openstack-cinder20:57
*** esp has joined #openstack-cinder20:59
*** apoorvad has quit IRC20:59
*** rlrossit has joined #openstack-cinder20:59
*** cdelatte has quit IRC21:05
*** esp has quit IRC21:05
*** raildo is now known as raildo-afk21:06
*** sheel has quit IRC21:07
*** jwcroppe has quit IRC21:09
*** esker has quit IRC21:13
*** merooney has quit IRC21:15
*** aorourke__ has joined #openstack-cinder21:18
*** aorourke_ has quit IRC21:18
*** stupidnic has joined #openstack-cinder21:22
*** jwcroppe has joined #openstack-cinder21:25
stupidnicI seem to be having an issue with Cinder using the Upload to Image function. I have some metadata set on a volume and when it goes transfer the volume to the image, I get an error about "is not of type u'string'21:26
*** martyturner has quit IRC21:29
*** martyturner has joined #openstack-cinder21:30
*** diablo_rojo has quit IRC21:31
*** cdelatte has joined #openstack-cinder21:31
scottdaDoes the os-brick release that just went out mean no more os-brick release for mitaka?21:32
scottdaAnd am I right that change from rootwrap to privsep is off the table for Mitaka?21:33
scottdahemnafk: seems AFK21:33
patrickeastscottda: that was my interpretation, but not 100% sure21:33
*** jungleboyj has quit IRC21:34
scottdapatrickeast: Thanks. We're talking about this in nova meeting ATM so I'm wanting to be sure.21:34
*** jgregor has quit IRC21:37
*** jgregor has joined #openstack-cinder21:37
*** timcl has quit IRC21:37
*** jgregor has quit IRC21:38
*** akerr has quit IRC21:38
*** krtaylor has quit IRC21:38
*** e0ne has joined #openstack-cinder21:38
openstackgerritRyan McNair proposed openstack/cinder: WIP Re-enable -1 child limits
*** ociuhandu has quit IRC21:41
scottdaSo, Nova folks seem to think we are doing another release of os-brick, and it must have the privsep change.21:41
scottdaWhich may be true.21:41
scottdaSo I'll stop spreading rumours and wait for something definitive from some definitive-type person.21:42
smcginnisscottda: Nope, no privsep for Mitaka.21:45
scottdaplease state that now in #openstack-meeting21:45
scottdasmcginnis: They are of a different opinion.21:45
*** krtaylor has joined #openstack-cinder21:51
*** jgregor has joined #openstack-cinder21:51
*** melwitt has joined #openstack-cinder21:54
*** merooney has joined #openstack-cinder21:54
*** mikal has joined #openstack-cinder21:54
*** chlong has quit IRC21:57
*** chlong_ has joined #openstack-cinder21:58
*** martyturner has quit IRC21:58
*** martyturner has joined #openstack-cinder21:59
*** apoorvad has joined #openstack-cinder21:59
*** cknight has quit IRC21:59
*** krtaylor has quit IRC22:02
*** mudassirlatif_ has joined #openstack-cinder22:02
*** salv-orlando has joined #openstack-cinder22:03
*** martyturner has quit IRC22:03
*** mudassirlatif has quit IRC22:04
*** mudassirlatif_ is now known as mudassirlatif22:04
*** apoorvad has quit IRC22:04
*** hemnafk is now known as hemna22:05
hemnasorry, lots of things being juggled22:05
*** salv-orl_ has quit IRC22:05
*** jwcroppe has quit IRC22:05
smcginnishemna: No change in privsep, right?22:06
hemnasmcginnis, correct22:06
hemnalast time I tried it still hung nova22:06
smcginnisJust want to make sure I'm not taking a hard stance on something I'm going to have to back pedal on. ;)22:06
smcginnisOK, that was my understanding. We should be fine I think.22:06
hemnayah you are good22:06
*** martyturner has joined #openstack-cinder22:07
smcginnisI think there were only two rootwrap changes so far that don't have corresponding changes in nova, so I don't think this should be an issue.22:07
smcginnisOther than not being able to get rid of rootwrap as soon as some would like.22:07
*** knikolla has quit IRC22:09
scottdaThanks guys.22:09
*** dustins has quit IRC22:09
openstackgerritEric Harney proposed openstack/python-cinderclient: Add extra_specs_list test
*** merooeny_ has joined #openstack-cinder22:11
openstackgerritJohn Griffith proposed openstack/cinder: Replication v2.1
openstackgerritYucong Feng proposed openstack/cinder: Remove extra get_active_zoneset query from Brocade CLI
smcginnisjgriffith: Happy happy joy joy22:13
jgriffithsmcginnis: haha... the rebase made a real mess.  Still not completely fixed, but wanted to get it pushed back up before I trash it again like last night :)22:14
*** merooney has quit IRC22:14
smcginnisjgriffith: Hah, play it safe. ;)22:15
*** krtaylor has joined #openstack-cinder22:15
bswartzdid the cheesecake moniker get dropped? :-(22:16
*** eharney has quit IRC22:16
smcginnisbswartz: The cool kids still use it. ;)22:16
smcginnisjgriffith: L70, should that be root.set() or are you missing adding elem to root?22:17
*** merooney has joined #openstack-cinder22:20
*** mylu has joined #openstack-cinder22:21
*** apoorvad has joined #openstack-cinder22:21
*** porrua has quit IRC22:22
*** merooeny_ has quit IRC22:22
*** mriedem has quit IRC22:24
*** arch-nemesis has quit IRC22:25
*** dave-mccowan has quit IRC22:26
*** JoseMello has quit IRC22:28
*** jungleboyj has joined #openstack-cinder22:30
jgriffithsmcginnis: I don't even know any more :)22:30
openstackgerritYucong Feng proposed openstack/cinder: Remove extra get_active_zoneset query from Brocade CLI
jgriffithsmcginnis: hehe.. yup22:32
Swansonjgriffith, your patch has been in the queue for 13 minutes without a pep8 failure.  This might be a record.22:32
jgriffithhaha... just give it a moment22:32
jgriffithSwanson: smcginnis just pointed them out22:33
*** mylu has quit IRC22:33
openstackgerritJohn Griffith proposed openstack/cinder: Replication v2.1
jgriffithok.. I *think* I'm back to where I was prior to rebase.  Please don't merge any other db and rpc migrations :)22:34
*** cknight has joined #openstack-cinder22:34
*** jgregor has quit IRC22:35
SwansonFully featured at this point? Just need to make your way past the smcginnis and patrickeast types and we're all mergelicious?22:36
jgriffithSwanson: I think we're pretty close.  Couple more API calls and the silly periodic22:36
*** tpsilva has quit IRC22:36
jgriffithSwanson: my goal is merge by end of day tomorrow :)22:37
patrickeastthe only thing that would block me from pushing it through is that the failover rest api call seems borked22:37
Swansonjgriffith, excellent!22:37
jgriffithpatrickeast: yeah... I haven't touched that yet :)22:37
patrickeastjgriffith: saving the best part for last!22:37
jgriffithpatrickeast: sorry, I should've explained the only thing I messed with was freeze/thaw22:37
jgriffithpatrickeast: *easiest* for last :)22:37
patrickeastjgriffith: no worries, i realized that after looking at it to figure out how to test it22:37
*** mylu has joined #openstack-cinder22:39
openstackgerritAngus Lees proposed openstack/os-brick: Trivial rootwrap -> privsep replacement
*** ociuhandu has joined #openstack-cinder22:41
*** cknight1 has joined #openstack-cinder22:42
*** mylu has quit IRC22:44
*** cknight has quit IRC22:44
*** mylu has joined #openstack-cinder22:45
*** ChubYann has quit IRC22:47
*** lpetrut has quit IRC22:48
*** edmondsw has quit IRC22:49
*** thurloat has joined #openstack-cinder22:50
mc_nairwoot woot - thanks for the workflow smcginnis!22:50
jungleboyjmc_nair: First quota patch?22:50
mc_nairjungleboyj: yea, the one to split out the NestedQuota support22:51
mc_nairbest news about it is it can be turned off now :)22:51
jungleboyjmc_nair: Awesome!22:51
*** aorourke__ has quit IRC22:52
*** aorourke_ has joined #openstack-cinder22:52
mc_nairSo now I put up a preview of re-enabling the -1 limits in child quotas here ( But I'm worried about the race-conditions that are definitely present in that patch22:52
mc_nairtrying to figure out how to fix those in an efficient way where we don't slow volume creates down or things like that. I'm a bit stumped on best way to do this, so I'm actively soliciting ideas if anyone has ways to do this ^ better22:54
jungleboyjmc_nair: I am assuming that the existing locking code for volume creation doesn't help?22:54
mc_nairsmcginnis, jgriffith - if you have any thoughts on the general approach ^^ lemme know, I'm hungry for feedback and improvements on it.22:55
*** martyturner has quit IRC22:55
mc_nairor DuncanT - not sure if you're around, but you always seems to be full of feedback :)22:55
mc_nairjungleboyj: lemme take a lock at what locks we have there currently. I am just worried if I start adding locks I'm going to end up grinding things to a halt (but I don't know if that's valid)22:56
jungleboyjmc_nair: Nice Freudian slip.  :-)22:57
*** rlrossit has quit IRC22:57
jungleboyjThat would be a place to start.  Agree that adding locks could cause performance concerns.22:58
*** ChubYann has joined #openstack-cinder22:58
*** cdelatte has quit IRC22:59
*** baumann has quit IRC23:00
*** ChubYann has quit IRC23:01
*** mudassirlatif has quit IRC23:01
*** cknight1 has quit IRC23:01
*** xyang1 has quit IRC23:01
*** e0ne has quit IRC23:06
*** vmorris has joined #openstack-cinder23:06
vmorrisanyone here have experience with setting up SMI-S provider in linux for connecting cinder to EMC VMAX Symmetrix Array?23:09
jgriffithSMI-S PIIM23:09
vmorriswhen i run the TestSmiProvider and display version (dv), I get Error: The target namespace does not exist. (Invalid namespace root/emc)23:10
*** cknight has joined #openstack-cinder23:11
mc_nairjungleboyj:  So currently, we do the quota reserve and then will roll things back if we're unable to complete the request.  At a minimum I think we'll want to add the same quota-reserve for allocated (we currently only do it for in_use), that way we can rollback stuff on failure23:11
vmorrisi have the 6 gatekeeper LUNS from the storage array visible to the linux SMI-S host23:12
mc_nairjungleboyj: it's possible pushing the logic further down into the db's quota_reserve, and do the reserving up the chain it could be alright.  Still unsure though :/23:12
mc_nairseems like it'd be moving in the right direction23:12
*** gouthamr has quit IRC23:14
*** ChubYann has joined #openstack-cinder23:15
jungleboyjmc_nair: If there is one spot where we can hit it for all cases that would be better.  I need to drop off here.  Will try to look at this a little more later and we can chat tomorrow.23:15
*** jungleboyj has quit IRC23:20
*** merooney has quit IRC23:20
*** jamielennox is now known as jamielennox|away23:20
*** mudassirlatif has joined #openstack-cinder23:21
*** cknight has quit IRC23:22
*** jwcroppe has joined #openstack-cinder23:23
*** ChubYann has quit IRC23:25
*** tobe has joined #openstack-cinder23:28
*** Suyash has joined #openstack-cinder23:29
*** cheneydc has joined #openstack-cinder23:33
*** tobe has quit IRC23:35
*** dims_ has joined #openstack-cinder23:36
*** dims has quit IRC23:38
*** cheneydc has quit IRC23:38
*** dave-mccowan has joined #openstack-cinder23:41
openstackgerritHelen Walsh proposed openstack/cinder: EMC VMAX - Operations and timeout issues
openstackgerritHelen Walsh proposed openstack/cinder: EMC VMAX - SnapVX and other snapshot improvements
*** openstackgerrit has quit IRC23:47
*** openstackgerrit_ is now known as openstackgerrit23:47
*** openstackgerrit_ has joined #openstack-cinder23:48
*** openstackgerrit_ is now known as openstackgerrit23:48
*** tobe has joined #openstack-cinder23:49
*** openstackgerrit_ has joined #openstack-cinder23:49
*** yhayashi has joined #openstack-cinder23:52
*** chlong_ has quit IRC23:52
*** mylu has quit IRC23:55
*** openstackgerrit_ has quit IRC23:55
*** openstackgerrit_ has joined #openstack-cinder23:57
*** sgotliv has quit IRC23:57
*** smoriya_ has joined #openstack-cinder23:57

Generated by 2.14.0 by Marius Gedminas - find it at!