Thursday, 2016-03-17

« Wednesday, 2016-03-16 Index Friday, 2016-03-18 »
*** akshai has joined #openstack-manila00:00
markstur_bswartz, bug/1525010 was 'falling thru the cracks' but it was simple enough that I pushed up a patch now.00:01
*** yangyapeng has quit IRC00:03
*** darrenc_ has joined #openstack-manila00:06
*** darrenc has quit IRC00:07
*** mtanino has quit IRC00:13
*** number80 has quit IRC00:14
*** darrenc_ is now known as darrenc00:25
*** jasonsb has joined #openstack-manila00:46
*** chlong has joined #openstack-manila00:55
*** akshai_ has joined #openstack-manila01:01
*** akshai has quit IRC01:05
*** yangyapeng has joined #openstack-manila01:05
*** yangyapeng has quit IRC01:07
*** a_ta has joined #openstack-manila01:14
*** chlong has quit IRC01:18
*** a_ta has quit IRC01:19
*** chlong has joined #openstack-manila01:31
bswartzmarkstur_: ty01:31
*** yangyapeng has joined #openstack-manila01:45
*** number80 has joined #openstack-manila02:01
*** tobe has joined #openstack-manila02:05
*** baojg has joined #openstack-manila02:11
*** bwolfe_ has joined #openstack-manila02:14
*** cknight has joined #openstack-manila02:18
*** darrenc has quit IRC02:19
*** akshai_ has quit IRC02:22
*** darrenc has joined #openstack-manila02:25
*** akshai has joined #openstack-manila02:32
*** number80 has quit IRC02:36
*** houming has joined #openstack-manila02:38
*** cknight has quit IRC02:38
*** cknight has joined #openstack-manila02:47
*** number80 has joined #openstack-manila02:50
*** baojg has quit IRC02:57
*** baojg has joined #openstack-manila03:00
*** bwolfe_ has quit IRC03:14
*** merooney has quit IRC03:22
openstackgerritMerged openstack/manila: 3PAR remove file tree on delete when using nested shares  https://review.openstack.org/29020903:32
*** martyturner has joined #openstack-manila03:33
*** cknight has quit IRC03:54
*** a_ta has joined #openstack-manila04:01
openstackgerritMerged openstack/manila: Fix Hitachi HNAS driver version  https://review.openstack.org/29317504:03
*** a_ta has quit IRC04:05
*** martyturner has quit IRC04:17
openstackgerritdaiki kato proposed openstack/manila: Fix typos  https://review.openstack.org/29330205:06
*** shausy has joined #openstack-manila05:22
*** mdnadeem has quit IRC05:33
*** a_ta has joined #openstack-manila05:40
*** a_ta has quit IRC05:46
*** tobe has quit IRC06:02
*** sgotliv has joined #openstack-manila06:06
*** chlong has quit IRC06:08
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Ensure Snapshots across replicas  https://review.openstack.org/29265906:10
*** baojg has quit IRC06:14
*** baojg has joined #openstack-manila06:15
*** chlong has joined #openstack-manila06:21
*** tobe has joined #openstack-manila06:29
*** tobe has quit IRC06:31
*** ociuhandu has quit IRC06:34
*** lpetrut has joined #openstack-manila06:36
*** ociuhandu has joined #openstack-manila06:36
*** baojg has quit IRC06:46
*** baojg has joined #openstack-manila06:46
*** nkrinner has joined #openstack-manila06:53
*** gouthamr_ has quit IRC07:11
openstackgerritMerged openstack/manila: HDS-HNAS: Fix exception in update_access not found  https://review.openstack.org/29135807:31
*** a_ta has joined #openstack-manila07:34
*** a_ta has quit IRC07:39
*** lpetrut has quit IRC07:44
*** chlong has quit IRC07:57
*** pcaruana has joined #openstack-manila08:24
*** rraja has joined #openstack-manila08:55
openstackgerritzhongjun proposed openstack/manila: Fix the scheduler choose a disable share service  https://review.openstack.org/29388809:02
openstackgerritMarc Koderer proposed openstack/manila: Fix binding in neutron driver  https://review.openstack.org/28349409:10
openstackgerritzhongjun proposed openstack/manila: Fix the scheduler choose a disable share service  https://review.openstack.org/29324609:14
openstackgerritDaniel Gonzalez Nothnagel proposed openstack/manila: [DOC] Fix use of keystone client in quick start guide  https://review.openstack.org/27413609:27
*** lpetrut has joined #openstack-manila09:36
*** rraja has quit IRC09:46
*** baojg has quit IRC09:49
*** baojg has joined #openstack-manila09:51
*** baojg has quit IRC09:54
*** baojg has joined #openstack-manila09:55
*** baojg has quit IRC10:01
*** rraja has joined #openstack-manila10:03
*** tpsilva has joined #openstack-manila10:04
*** baojg has joined #openstack-manila10:06
*** baojg has quit IRC10:10
*** aovchinnikov has quit IRC10:17
*** martyturner has joined #openstack-manila10:17
*** martyturner has quit IRC10:22
*** ociuhandu has quit IRC10:24
*** absubram has quit IRC10:34
*** ganso has joined #openstack-manila10:54
*** ociuhandu has joined #openstack-manila10:57
openstackgerritzhongjun proposed openstack/manila: Fix the scheduler choose a disable share service  https://review.openstack.org/29324611:03
*** aovchinnikov has joined #openstack-manila11:15
*** dalgaaf has joined #openstack-manila11:33
zhongjunvponomaryov: ping11:34
openstackgerritValeriy Ponomaryov proposed openstack/manila: Fix report of ZFSonLinux driver capabilities  https://review.openstack.org/29302511:44
vponomaryovzhongjun: pong11:44
zhongjunvponomaryov: When I run command: manila quota-show --tenant=not_exist --user=not_exist,11:46
zhongjuneven the tenant and user does not exist,11:46
zhongjunthe command still return success and echo specific values(quota-default).11:46
vponomaryovzhongjun: ше шы сщккусе иурфмшщгк11:47
zhongjunvponomaryov: Is it the originally design like this? or does we need do some change?11:47
vponomaryov zhongjun: sorry11:47
vponomaryov zhongjun:it is correct behavior11:47
zhongjunvponomaryov: It makes me confused, It makes me think the tenant(not_exist) exists.11:48
*** chlong has joined #openstack-manila11:48
*** openstackgerrit has quit IRC11:48
vponomaryovzhongjun:  manila doe not verify whether user or tenant exist, if it has no redefinition of quotas it returnes default values11:48
vponomaryovzhongjun: it should make you think , that user with ID will have such quota if it exists11:48
*** openstackgerrit has joined #openstack-manila11:48
*** yangyape_ has joined #openstack-manila11:49
zhongjunDoes it let the user feel the tenant(that is not exist) exist.11:49
vponomaryovzhongjun: no11:50
vponomaryovzhongjun: it let user think that if tenant exists it will have such quota11:50
vponomaryovzhongjun: if tenant does not exist you just won't be able to use i11:50
vponomaryovs/i/it/11:50
vponomaryovzhongjun: also, CLI uses direct IDs, so usernames and tenant/project names are not proper values to use11:52
*** mtanino has joined #openstack-manila11:52
vponomaryovzhongjun: you will have default values be returned if use names11:52
vponomaryovzhongjun: and it is not the ones that will be used calculating usage when you consume some resources11:53
*** yangyapeng has quit IRC11:53
*** yangyape_ has quit IRC11:53
*** lpetrut has quit IRC11:54
zhongjunvponomaryov: ok, this logic sounds ok. thanks11:55
*** lpetrut has joined #openstack-manila11:58
tpsilvavponomaryov: who's responsible for stable/liberty?11:58
vponomaryovtpsilva: ?11:59
tpsilvavponomaryov: for keeping it working... is there anyone who does that?11:59
vponomaryovtpsilva: you mean who has +2 power?11:59
*** mlima has joined #openstack-manila11:59
tpsilvavponomaryov:  ganso and I are worried it might be broken11:59
tpsilvavponomaryov: https://review.openstack.org/#/c/293389/12:00
tpsilvavponomaryov: it's failing the same test12:00
gansostable/liberty CI at least is12:00
tpsilvavponomaryov: we sent another debug patch and it failed on the exactly same test12:01
vponomaryovtpsilva: I am not surprised, since Tempest and Devstack changes tend to break stable branches12:01
vponomaryovtpsilva: anyone can volunteer fixing of stable branch12:02
vponomaryovtpsilva: but standard core manila group does not have +2 power there12:02
tpsilvavponomaryov: makes sense that anyone can fix it, but I thought maybe there was a guy who watches it to make sure it's working... some of the cores12:03
tpsilvavponomaryov: anyway, we'll look into that12:03
vponomaryovtpsilva: maybe there is such guys for other projects12:03
vponomaryovtpsilva: but definitely no such guy for Manila project12:03
vponomaryovtpsilva: kilo branch is broken for long time ))12:04
vponomaryovtpsilva: every year-old stable branch is broken always12:05
vponomaryovtpsilva: in Manila12:05
vponomaryovtpsilva: broken in scope of CI12:05
tpsilvavponomaryov: but then how do you backport if CI is broken?12:05
tpsilvavponomaryov: must fix CI first?12:05
vponomaryovtpsilva: year-old branches accept only security fixes12:06
vponomaryovtpsilva: I remember only one such fix for Manila, where user context could be elevated to having admin rights12:07
vponomaryovtpsilva: but yes, to be able to merge something, CI should be fixed first12:07
vponomaryovtpsilva: obviously ))12:07
tpsilvavponomaryov: so looks like stable/liberty currently needs some attention :)12:08
tpsilvavponomaryov: we'll take on that12:08
tpsilvavponomaryov: thank you!12:08
tpsilvavponomaryov: s/take on that/take a look on that12:09
vponomaryovtpsilva: let the force be with you ))12:09
tpsilvavponomaryov: always :)12:09
*** mlima_ has joined #openstack-manila12:13
*** mlima has quit IRC12:16
*** martyturner has joined #openstack-manila12:20
*** merooney has joined #openstack-manila12:23
openstackgerritMerged openstack/manila: Fix error logged for wrong HPE 3par client  https://review.openstack.org/29376312:32
*** gouthamr has joined #openstack-manila12:44
*** mlima__ has joined #openstack-manila12:45
*** eharney has joined #openstack-manila12:47
*** mlima_ has quit IRC12:48
*** mlima_ has joined #openstack-manila12:50
*** alyson_ has quit IRC12:52
*** mlima__ has quit IRC12:53
*** timcl has joined #openstack-manila12:55
*** yangyapeng has joined #openstack-manila12:57
*** houming has quit IRC13:02
*** dustins has joined #openstack-manila13:03
*** martyturner has quit IRC13:03
*** porrua has joined #openstack-manila13:06
*** mtanino has quit IRC13:07
tpsilvabswartz: ping13:10
*** aovchinnikov has quit IRC13:12
gouthamrvponomaryov: ping13:16
*** dustins_ has joined #openstack-manila13:17
*** dustins has quit IRC13:17
vponomaryovgouthamr: pong13:17
*** akshai has quit IRC13:19
vponomaryovgouthamr: just added comments to your change of snapshot replication13:19
gouthamrvponomaryov: sorry about breaking zfsonlinux.. I can quickly update the driver interfaces, i was hoping i can talk to you about the approach before that.13:19
vponomaryovgouthamr: it still has no population of provider location, is it todo or not todo thing?13:19
gouthamrvponomaryov: do you think this would help you avoid the race condition we were talking about yesterday?13:20
vponomaryovgouthamr: what this?13:20
gouthamrvponomaryov: sending all the snapshot info through replica crud interfaces13:20
vponomaryovgouthamr: and why help me? NetApp driver suffers from it too13:20
vponomaryovgouthamr: I cam to idea how I can workaround it13:21
gouthamrvponomaryov: the netapp driver uses the provider location from the snapshot instance list provided13:21
vponomaryovgouthamr: using private data storage13:21
gouthamrvponomaryov: +1, that's one approach.13:21
vponomaryovgouthamr: I describe scenario yesterday13:21
vponomaryovgouthamr: where we lose the only available source of provider location13:22
vponomaryovgouthamr: it means, NetApp driver will suffer from this scenario13:22
bswartztpsilva: pong13:22
vponomaryovgouthamr: each creation of new replica means creation of new snapshot instances per snapshot, it means that share manager should be aware about data updates for them13:23
gouthamrvponomaryov: this was your situation.. 1) create share; 2) create snapshot; 3) create replica; 4) promote replica not waiting periodic update; 5) delete old active replica; 6) try delete snapshot13:23
bswartzgouthamr vponomaryov: you on the same page?13:23
gouthamrbswartz: trying to get on it13:23
bswartzlmk if you need me to join in13:23
vponomaryovgouthamr: on step 6 you do not have provider location anymore13:23
vponomaryovgouthamr: because it was not populated13:24
vponomaryovgouthamr: and the only source was already removed13:24
* bswartz feels uncomfortable about using provider location in replication when driver-private-share-data is available13:24
gouthamrvponomaryov: that's why you're getting the snapshot instances in the create_replica call13:24
vponomaryovbswartz: I think it is design problem if we feature "requires" usage of private data storage13:25
vponomaryovs/we//13:25
*** timcl has quit IRC13:25
gouthamrvponomaryov: the driver should not be reporting that the replica is in_sync until the snapshots are available on it13:25
bswartzvponomaryov: that's a reasonable point13:25
gouthamrvponomaryov: until the 'available' snapshots are available on it.13:26
bswartzvponomaryov: I'm not sure it's required but it seems like the best way to tie the snapshot instances together13:26
* gouthamr tries hard not to sound confusing13:26
vponomaryovgouthamr: you ignore the fact of absense of provider location ))13:26
gouthamrvponomaryov: oh wait, one other thing here -> since this patch, every netapp snapshot will have a provider location13:27
vponomaryovgouthamr: mentioned scenario makes "provider_location" approach blow up13:27
vponomaryovgouthamr: only original one has13:27
gouthamrvponomaryov: so you create your first snapshot, the provider location exists on it.13:27
vponomaryovgouthamr: that is the problem13:27
gouthamrvponomaryov: yes.. and you're getting the original snapshot in the create_replica call13:27
vponomaryovgouthamr: and what? driver cannot save info to DB13:28
vponomaryovgouthamr: with latest code13:28
*** martyturner has joined #openstack-manila13:28
vponomaryovgouthamr: in scope of "create_replica" call13:28
gouthamrvponomaryov: doesn't need to.. look into the 'active_replica_snapshot', grab its provider_location13:28
*** dencaval has joined #openstack-manila13:29
gouthamrvponomaryov: https://review.openstack.org/#/c/292659/5/manila/share/driver.py ... see the 'replica_snapshots' field in the create_replica call13:29
vponomaryovgouthamr: ok, I will make working variant in ZFS that I already can do13:29
vponomaryovgouthamr: I am talking about different hting13:30
vponomaryovgouthamr: about return data from driver "create_replica" call13:30
gouthamrvponomaryov: oh.. you mean, for the newly created replica, you can't update the snapshot_instance's provider location?13:31
vponomaryovgouthamr: finally13:31
gouthamr:)13:31
*** akshai has joined #openstack-manila13:32
gouthamrvponomaryov: that is a bummer, i understand that situation... that's why we have the update_replicated_snapshot call.. right, bswartz?13:32
vponomaryovgouthamr: may pass long time before first call of it13:32
ameadehmmm13:33
gouthamrvponomaryov: how about storing it in driver_private_data and updating it in the update_replicated_snapshot call..13:33
*** Yogi1_ has joined #openstack-manila13:33
*** timcl has joined #openstack-manila13:33
ameadewhat would you look it up by in the private data?13:33
gouthamrameade: you always have the active replica's snapshot instance provided to you13:33
ameadebut if that's the cast you have the provider_location13:34
ameadecase*13:34
ameadenot seeing how provider_location and private storage are diff13:35
ameadejust two places we could store the same information13:35
ameadehow does that solve the problem?13:35
tpsilvabswartz: hi, good morning13:35
gouthamrameade: the active replica snapshot should always have the provider location. remember that the snapshot instance of the replica itself stays in 'creating' until it is ensured. when you promote a replica that has any snapshot instance in 'creating', the snapshot instance is set to 'error'..13:36
ameadeif there are no instances with provider location then you know it doesnt exist right?13:36
ameadeyeah13:36
ameadeso what's wrong?13:36
tpsilvabswartz: so, I was implementing the solution that you and cknight came with and I actually noticed something we haven't though yesterday13:36
bswartzk13:37
ameadeyou could have an issue updating private storage just the same as you do updating provider_locaiton13:37
tpsilvabswartz: the manager is supposed to drop the requests when the status is 'updating'13:37
ameadevponomaryov: sorry to jump in the middle of the convo, but am I making sense?13:38
bswartzwhat are all the possible statuses?13:38
tpsilvabswartz: but the api is supposed to always change the status to 'needs_update' and always send the request to manager13:38
tpsilvabswartz: so it will never be 'updating'13:38
tpsilvabswartz: 'needs_update', 'updating', 'active' and 'error'13:38
bswartzokay that all sounds accruate13:39
tpsilvabswartz: but did you understand the problem?13:39
bswartzthe manager changes the state from needs_update to updating after it receives the RPC13:39
tpsilvabswartz: yes, but the api always changes it to 'needs_update', so the manager will never drop13:40
bswartzone second13:40
tpsilvak13:41
bswartzokay I see the problem13:41
tpsilvaI actually thought of another possible solution13:42
tpsilvabswartz: https://review.openstack.org/#/c/287758/13/manila/share/access.py13:42
tpsilvabswartz: if we could make lines 104 to 123 to be an atomic operation, this would solve the race condition, right?13:43
vponomaryovameade: problem is that, goutham thinks that NetApp driver cannot face case when provider_location is not set for any of snapshot instances13:43
vponomaryovameade: and all snapshot will be "available"13:43
ameadenobody can? that means there is nowhere the snapshot exists13:43
vponomaryovameade: all can13:44
vponomaryovameade: it is the problem, we can have lots of "available" snapshot instances and only one will have provider location13:45
ameadein the netapp driver, we wont say it's available without setting the provider locaiton13:45
vponomaryovameade: so if we manage to delete one that has provider location before first call of "snapshot instance update" we face case when no providre location can be found13:45
ameadeall available snapshots should have provider location set if we depend on it13:46
ameadewe only return available and provider location together13:47
ameade            return {13:47
ameade                'status': constants.STATUS_AVAILABLE,13:47
ameade                'provider_location': snapshot_name,13:47
ameade            }13:47
ameadeiiuc, I think it's not really an issue13:47
ameadebswartz: and gouthamr  are chatting irl about something13:50
*** xyang1 has joined #openstack-manila13:54
*** mtanino has joined #openstack-manila13:57
*** a_ta has joined #openstack-manila14:00
*** shausy has quit IRC14:05
*** Yogi1_ has quit IRC14:05
*** shausy has joined #openstack-manila14:06
gansogouthamr: ping14:07
*** rraja has quit IRC14:12
openstackgerritMarc Koderer proposed openstack/manila: Fix keystone v3 issues for all clients  https://review.openstack.org/29051114:14
bswartzsorry tpsilva14:14
bswartzI got interrupted14:14
bswartzlooking now14:14
*** shausy has quit IRC14:14
tpsilvabswartz: no problem14:14
tpsilvabswartz: I have the team daily meeting right know, bbiab14:15
gouthamrganso: pong14:15
bswartztpsilva: you are correct that we need 2 different pieces of state14:15
bswartzthere needs to be a bit that says: the api services changed something recently14:16
bswartzand there needs to be another bit that says: the manager is working on an update currently14:16
*** rickyrem has joined #openstack-manila14:17
bswartzactually the first piece of state needs more than a bit -- it needs 3 states14:18
* bswartz goes to draw state diagram on whiteboard14:19
*** aovchinnikov has joined #openstack-manila14:19
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Add tests to ensure snapshots across replicas  https://review.openstack.org/29311314:21
bswartztpsilva: this could also be addressed by adding 1 more state -- I'm not sure what to call it though14:24
bswartzif there was an additional state which indicated a combination of updating and needs_update -- let's call it "multiple_update" then the logic could work like this:14:25
vponomaryovgouthamr: https://bugs.launchpad.net/manila/+bug/155858614:27
openstackLaunchpad bug 1558586 in Manila "Replica scheduling considers errored replicas" [Undecided,New]14:27
bswartzwhen the API service gets a change, it writes it into the database and then looks at the state -- if the state is active it changes it to needs_update and sends an RPC -- if the state is updating it changes it to "multiple_update" and sends an RPC14:27
*** pcaruana has quit IRC14:28
gouthamrvponomaryov: that's by design :P14:28
vponomaryovgouthamr: nice14:28
bswartzwhen the manager receives an RPC it looks at the state and if its updating or multiple_update it ignores it -- if it's needs_update it sets it to updating and begins an update14:28
gouthamrvponomaryov: haha. kidding. hey but i know the problem, and it isn't really a problem since we don't know why exactly it failed on that particular host14:29
bswartzwhen the manager completes and update it again looks at the state and if it's updating it sets it back to active, and if it's multiple_update it sets the state back to updating an restarts the update process14:29
bswartztpsilva: what do you think of that ^ approach?14:29
bswartzso the state machine grows from 4 states to 5 states (including error) and we need to do some test-and-set type updates14:30
vponomaryovgouthamr: informing you that I do not have reason to use list of snapshot instances that are added to interface of replicas14:30
vponomaryovgouthamr: since I do not use provider_location as was lanned14:31
vponomaryov*planned14:31
gouthamrvponomaryov: oh.. okay.. not even for the update_replica call?14:31
tpsilvabswartz: hi, I'm back from daily meeting.. i'm reading it, just a sec14:31
vponomaryovgouthamr: it was required for storing information only14:31
vponomaryovgouthamr: so, I workarounded it using private data storage14:32
*** dustins_ has quit IRC14:32
gouthamrvponomaryov: nice. yes, that is a solution..14:32
vponomaryovgouthamr: does NetApp driver need that new args?14:32
gouthamrvponomaryov: as long as we don't tie up the data that we're storing in any way with the 'active' replica at the time.. it is fine14:33
tpsilvabswartz: this might actually work14:33
tpsilvabswartz: it's simple enough to implement14:33
tpsilvabswartz: I'll add it to my current code and test it14:34
bswartztpsilva: I think in an ideal case we'd have more DB columns, but this should be a quick fix14:34
gouthamrvponomaryov: because the 'active' replica can ultimately change.. i mean, use the 'active' one at the time to come up with the name/data you want to store.. but don't rely on that being the 'active' one forever.. that was one consideration ameade and i had at the earlier discussions around this fix14:34
vponomaryovgouthamr: To finish driver implementation we need to get nailed interfaces14:35
vponomaryovgouthamr: so, I am asking, does this new arg in replica interfaces requried for NetApp driver?14:35
gouthamrvponomaryov: the NetApp driver doesn't need the snapshots to be sent in the create and delete calls14:35
vponomaryovgouthamr: ok, then, we need to revert it14:35
gouthamrvponomaryov: ameade ignores those args14:35
vponomaryovgouthamr: ZFS will do the same14:36
gouthamrvponomaryov: are we sure no other driver needs it?14:36
vponomaryovgouthamr: it has no useful info14:36
gouthamrvponomaryov: we realized they are essential in the update_replica call as akerr pointed out on the patch yesterday14:36
vponomaryovgouthamr: only provider location was useful14:36
bswartzvponomaryov gouthamr: I'm not following your conversation and I'm in a meeting -- but my request is that we pick an implementation which will work well enough and that we can build on in Newton -- the driver interface can/should be revisited14:37
gouthamrvponomaryov: then do we keep it in Mitaka, and revisit the design in Newton?14:38
vponomaryovgouthamr: moreover, why update of snapshots and replicas are two different updates?14:38
gouthamrvponomaryov: because snapshots need not be updated all the time14:38
gouthamrvponomaryov: we only poll for 'creating' and 'deleting' snapshots14:38
vponomaryovgouthamr: replicas not all the time? ))14:38
gouthamrvponomaryov: tell me about it14:38
vponomaryovgouthamr: wow-wow, easy ))14:39
gouthamrhaha14:39
gouthamrvponomaryov: replying on your bug..14:39
gouthamr https://bugs.launchpad.net/manila/+bug/155858614:40
openstackLaunchpad bug 1558586 in Manila "Replica scheduling considers errored replicas" [Undecided,New]14:40
*** erlon has joined #openstack-manila14:40
*** timcl has quit IRC14:40
*** timcl has joined #openstack-manila14:42
*** a_ta has quit IRC14:42
*** kaisers_ has joined #openstack-manila14:42
*** a_ta has joined #openstack-manila14:43
vponomaryovgouthamr: replica creation is memberadmin API, force deletion is admin only API14:43
vponomaryov*merber and admin14:43
vponomaryovgouthamr: so, you make member be dependent on admin14:44
vponomaryovgouthamr: it is admin's problem to remove errored replicas14:44
vponomaryovgouthamr: so, what about interfaces?14:45
gouthamrvponomaryov: what if there's a genuine problem that needs admin intervention14:45
vponomaryovgouthamr: that scheduling bug is low priority now14:46
vponomaryovgouthamr: what about interfaces? We update them or not?14:46
gouthamrvponomaryov: i'll defer to bswartz.. he suggested sending all the snapshot instances in the crud calls.14:46
vponomaryovgouthamr: none of driver is going to use them14:46
dencavalGuys, in case the share -network is created and passed by parameter to create the share, but the appears: Share server handling is enabled. But 'share_server' is not provided. Make sure you used 'share_network'. What else can be?14:47
gouthamrvponomaryov: none of the drivers that currently implement replication.14:47
gouthamr:)14:47
vponomaryovgouthamr: it is enough14:47
vponomaryovgouthamr: to not doing thing that can never be used14:47
*** a_ta has quit IRC14:47
gouthamrvponomaryov: i agree.. bswartz? ameade?14:48
dencaval*Guys, in case the share-network is created and passed by parameter to create the share, but it appears: Share server handling is enabled. But 'share_server' is not provided. Make sure you used 'share_network'. What else can be?14:48
vponomaryovdencaval: one you, either driver you are wrong14:48
vponomaryovdencaval: oh, my above statemtn is not english, sorry ))14:49
vponomaryovdencaval: how do you make sure you provided it?14:49
vponomaryovdencaval: scenario you use?14:49
ameadeposted on the bug gouthamr vponomaryov14:49
ameadelet me review some code already14:49
ameadelol14:49
gouthamrameade: posted a reply to your reply. users can delete replicas in 'error' state.. but if the backend misbehaves, only the force can help you.14:50
vponomaryovgouthamr: yeah, the Force can help everyone ))14:51
ameadecorrect14:51
*** a_ta has joined #openstack-manila14:51
ameadeuse the force!14:51
*** a_ta has quit IRC14:52
dencavalvponomaryov, this command was executed: manila create NFS 1 --name share1 --share-network private-share1. Both share-network and share were creating using demo project.14:52
*** a_ta has joined #openstack-manila14:53
*** cknight has joined #openstack-manila14:53
gansogouthamr: ping14:55
gouthamrhi ganso!14:55
gansogouthamr: hi Goutham!14:55
gansogouthamr: are you working on https://bugs.launchpad.net/manila/+bug/1544725 ?14:55
openstackLaunchpad bug 1544725 in Manila "manage API ignores snapshot_support in the share_type" [High,New] - Assigned to Goutham Pacha Ravi (gouthamr)14:55
gouthamrganso: no i'm not, currently...14:55
ameadegouthamr: so much test code in this patch14:56
gansogouthamr: do you intend to?14:56
*** Zhongjun_ has joined #openstack-manila14:56
gouthamrganso: that's currently low on my priority list.. got another couple bugfixes that take the higher priority :(14:57
gouthamrganso: i can get around to it.. but are you offering help? :)14:57
gouthamrameade: yeah, thorough..14:58
gouthamr:P14:58
ameadei didnt even know about test_models.py14:58
gouthamrameade: ganso created it.14:58
gouthamrameade: we had these properties that had so much logic go into them recently..14:58
ameadeyeah14:59
gansogouthamr: sure, let's discuss it after the meeting, don't know if it is going to be postponed or not14:59
gouthamrsure thing ganso14:59
*** markstur has joined #openstack-manila15:01
*** mlima_ has quit IRC15:04
*** alejandrito has joined #openstack-manila15:05
*** yangyapeng has quit IRC15:06
*** yangyapeng has joined #openstack-manila15:07
*** rraja has joined #openstack-manila15:08
*** yangyape_ has joined #openstack-manila15:13
*** yangyapeng has quit IRC15:16
*** bwolfe has joined #openstack-manila15:18
*** mlima has joined #openstack-manila15:20
*** baojg has joined #openstack-manila15:23
*** dustins has joined #openstack-manila15:32
*** rickyrem has quit IRC15:32
*** rickyrem has joined #openstack-manila15:36
*** rraja has quit IRC15:37
*** rickyrem has quit IRC15:44
*** rickyrem has joined #openstack-manila15:49
*** nkrinner has quit IRC15:49
*** merooney has quit IRC15:50
*** bwolfe has quit IRC15:54
openstackgerritMerged openstack/manila: Fix typos  https://review.openstack.org/29330215:55
openstackgerritMerged openstack/manila: Fix report of ZFSonLinux driver capabilities  https://review.openstack.org/29302515:56
gansogouthamr: ping15:56
*** aovchinnikov has quit IRC15:56
*** martyturner has quit IRC15:57
*** ChanServ changes topic to "12 bugs left until RC1"15:57
*** martyturner has joined #openstack-manila16:01
openstackgerritTin Lam proposed openstack/manila: Add hacking check to ensure not to use xrange()  https://review.openstack.org/29412416:02
*** markstur has quit IRC16:05
*** jseiler_ has quit IRC16:08
gansocknight: ping16:09
*** timcl has quit IRC16:10
openstackgerritMadhu Mohan proposed openstack/manila: Lock operations for share_instance, snapshot  https://review.openstack.org/28944116:11
*** dustins has quit IRC16:15
*** dustins_ has joined #openstack-manila16:15
*** dustins has joined #openstack-manila16:19
*** dustins_ has quit IRC16:20
*** pcaruana has joined #openstack-manila16:22
*** dustins has quit IRC16:26
openstackgerritThomas Bechtold proposed openstack/manila: Fix generic and LVM driver access rules for CIDRs  https://review.openstack.org/29414016:30
toabctlbswartz, vponomaryov: ^^ that's the fix for lp#155252616:30
toabctlvponomaryov: I can add you as co-author if you want. it was basically copy&paste from your ZFS fix :)16:31
gansotoabctl: is this just to fix 0.0.0.0/24? I have tested 20.0.0.0/24 and it works16:31
toabctlganso: I tested 0.0.0.0/016:32
toabctlwhich works with the patch but doesn't without16:32
gansotoabctl: 0.0.0.0/0 does not work, I have tested that16:32
gansotoabctl: as commit message implies 20.0.0.0/24 would not work, since it is short-form16:33
toabctlganso: you mean 20.0.0.0/24 works also without any patch?16:34
*** baojg has quit IRC16:35
*** martyturner has quit IRC16:37
gansotoabctl: yes16:37
*** martyturner has joined #openstack-manila16:38
toabctlganso: true. tested with short form (192.168.0.0/16) and it works16:39
toabctlI basically copied the commit message from the zfsonlinux cidr fix16:39
toabctlbut I have to leave now :(16:39
toabctlso if somebody wants to fix it, just go ahead16:39
*** merooney has joined #openstack-manila16:39
toabctlganso: so the fix fixes the 0.0.0.0/0 case which the bug report is about16:41
toabctlanyway. bye. maybe back later.otherwise tomorrow16:42
gouthamrganso: pong16:44
gansogouthamr: about that bug16:44
gansogouthamr: I am configuring my lab to repro it16:44
gouthamrganso: yep..16:45
gansogouthamr: if you specify a share-type, does it work?16:45
gouthamrganso: manage a share with share_type with snapshot_support set to False16:45
gansogouthamr: in your example, you are using default share type, right?16:45
gouthamrganso: i changed the default share type to have the snapshot_support extra_spec set to False16:46
gouthamrganso: I guess the real question is.. do we need to allow managing with conflicting extra-specs..16:46
gouthamrganso: manage does not go through teh scheduler..16:47
gouthamrganso: so we never do all of these extra-spec tallying...16:47
gansogouthamr: I believe it should respect the share_type definition16:47
gouthamrganso: we always know the host in the manage operation, don't we?16:48
*** timcl has joined #openstack-manila16:48
cknightganso: pong16:48
gansogouthamr: oh wait, you are saying that it is derived from the host?16:48
gansocknight: https://review.openstack.org/#/c/293708/116:48
cknightganso: saw your comments, haven't gotten back to it yet16:48
*** alejandrito has quit IRC16:49
gouthamrganso: for the bugfix, i was just hoping we could do what we're doing in create, consistently in manage as well16:49
gouthamrganso: copy the snapshot_support key from the share type to the share model16:49
gansogouthamr: exactly, but what if the host says snapshot_support is the opposite value?16:50
gouthamrganso: but i don't like the idea about not enforcing it.. in manage, we know the host.. the host may/may not support snapshots.. but we add this thing from the share_type?16:50
gansogouthamr: I have not looked at the code yet, but if share_type extra specs are not being validated, then they should be16:50
gouthamrganso: yes16:50
gouthamrganso: +116:50
*** alejandrito has joined #openstack-manila16:51
*** nkrinner has joined #openstack-manila16:53
openstackgerritMerged openstack/python-manilaclient: Fix typos  https://review.openstack.org/29318216:55
*** alejandrito has quit IRC17:01
cknightganso: OK, I see what you mean.  No force-delete needed, which is what I wanted to avoid.  I'll update my patch to just raise the exception.17:02
*** jasonsb has quit IRC17:02
*** erlon has quit IRC17:17
*** zhongjun has quit IRC17:29
*** Zhongjun_ is now known as Zhongjun17:29
*** zhongjun2 has joined #openstack-manila17:31
gansogouthamr: I believe there is no way to do it in manager without re-implementing scheduler logic17:32
gouthamrganso: yessir. i agree.. so maybe we should live with accepting that that disparity exists for now17:32
gansogouthamr: going to have to call scheduler to validate, like I do in migration17:32
gouthamrganso: this isn't true just about snapshot_support though17:34
gouthamrganso: for all the extra specs that are in the share type that are conflicting17:34
gansogouthamr: yes17:34
gouthamrganso: i'm in favor of pushing this bug to newton... since its existed since liberty and we need a little more time to discuss this17:36
gansogouthamr: I will try to push patch anyway, then if other reviews think it is too risky, then we postpone17:36
gouthamrganso: sure thing. thank you!17:37
ameadehttps://review.openstack.org/#/c/291803/ cknight xyang1 markstur_17:38
ameadethanks for the reviews folks17:39
*** lpetrut has quit IRC17:42
*** rickyrem has quit IRC17:47
openstackgerritMadhu Mohan proposed openstack/python-manilaclient: Use os_client_config for functional tests  https://review.openstack.org/26701817:47
*** rickyrem has joined #openstack-manila17:50
openstackgerritTin Lam proposed openstack/manila: Add hacking check to ensure not to use xrange()  https://review.openstack.org/29412418:04
openstackgerritClinton Knight proposed openstack/manila: NetApp cDOT: Raise ShareResourceNotFound in update_access  https://review.openstack.org/29370818:07
cknightganso: ^^ Much smaller change this time.18:07
openstackgerritValeriy Ponomaryov proposed openstack/manila: Make ZFSonLinux driver handle snapshots of replicated shares properly  https://review.openstack.org/29337718:08
gansocknight: nice! I will give +2 when Jenkins and NetApp CI runs18:09
cknightganso: ok, thanks18:09
mkoderer___bswartz: cknight: https://review.openstack.org/#/c/290511/ is again ready for review.. hope I addressed all findings18:16
*** pcaruana has quit IRC18:18
cknightmkoderer: looking now18:18
*** nkrinner has quit IRC18:24
*** pcaruana has joined #openstack-manila18:30
*** eharney has quit IRC18:37
*** Zhongjun has quit IRC18:38
*** eharney has joined #openstack-manila18:41
*** ociuhandu has quit IRC18:41
mkoderer___bswartz: I will address the things from cknight tomorrow mornig (CET)18:43
mkoderer___hope that works to get it in RC118:44
cknightmkoderer, bswartz: it's more important to hear back from Jamie Lennox.  My requests are fairly simple.18:45
*** markstur has joined #openstack-manila18:45
*** markstur_ has quit IRC18:46
mkoderer___cknight: I will ping him again18:46
openstackgerritMerged openstack/manila: Fix the scheduler choose a disable share service  https://review.openstack.org/29324618:56
*** martyturner has quit IRC18:56
*** ChanServ changes topic to "11 bugs left until RC1"18:58
*** lpetrut has joined #openstack-manila18:59
cknighttpsilva: Hi, Tiago.  How's the concurrency patch coming?19:02
tpsilvacknight: hi clinton19:02
tpsilvacknight: actually I've ran into some other problems19:02
tpsilvacknight: but with the deny_access now19:02
tpsilvacknight: trying to fix those, but without success19:03
cknighttpsilva: ok, let us know if you need to discuss further19:03
*** martyturner has joined #openstack-manila19:03
tpsilvacknight: okay... I'll try a few more things and if those don't work, I'll ping you again19:04
cknightmarkstur, xyang1: This one seems ready to merge: https://review.openstack.org/#/c/294140/19:06
xyang1cknight: done19:09
cknightxyang1: thanks19:09
xyang1cknight: np19:09
*** kaisers_ has quit IRC19:11
openstackgerritValeriy Ponomaryov proposed openstack/manila: Make ZFSonLinux driver handle snapshots of replicated shares properly  https://review.openstack.org/29337719:23
openstackgerritValeriy Ponomaryov proposed openstack/manila: Make ZFSonLinux driver handle snapshots of replicated shares properly  https://review.openstack.org/29337719:26
tpsilvacknight: ping19:30
cknighttpsilva: pong19:30
tpsilvacknight: If I change a property from one of the db models19:30
tpsilvacknight: do I need a migration?19:30
tpsilvacknight: no, right?19:30
cknighttpsilva: change it how?19:30
tpsilvajust a sec19:31
tpsilvacknight: https://github.com/openstack/manila/blob/master/manila/db/sqlalchemy/models.py#L55319:31
tpsilvacknight: it's not just out_of_sync now19:31
tpsilvacknight: needs to be changed to out_of_sync or updating or multiple_updates19:32
tpsilvacknight: I already did change, but it does not seem to work19:32
cknighttpsilva: no, you're not changing the schema, and you're not changing any data in existing db data.19:32
tpsilvacknight: oh, I see that it calls another function19:33
tpsilvacknight: I missed that19:33
tpsilvacknight: found the error, thanks19:33
tpsilvarubber duck debugging :P19:34
bswartzhttps://en.wikipedia.org/wiki/Rubber_duck_debugging19:37
*** ociuhandu has joined #openstack-manila19:38
openstackgerritEmilien Macchi proposed openstack/puppet-manila: releasenotes: add notes for Mitaka cycle  https://review.openstack.org/29425919:49
*** martyturner has quit IRC19:50
openstackgerritTiago Pasqualini da Silva proposed openstack/manila: Fix update_access concurrency issue  https://review.openstack.org/28775819:50
tpsilvacknight: ^19:50
*** jseiler has joined #openstack-manila19:51
openstackgerritTiago Pasqualini da Silva proposed openstack/manila: Fix update_access concurrency issue  https://review.openstack.org/28775819:52
tpsilvahehe forgot to remove some debug logs :)19:52
*** martyturner has joined #openstack-manila19:54
*** dustins has joined #openstack-manila20:03
openstackgerritRodrigo Barbieri proposed openstack/manila: Fix manage api ignoring type extra specs  https://review.openstack.org/29426920:10
gansogouthamr: ^20:11
gouthamrganso: looking20:12
*** sgotliv has quit IRC20:16
openstackgerritMerged openstack/manila: Fix generic and LVM driver access rules for CIDRs  https://review.openstack.org/29414020:16
*** pcaruana has quit IRC20:17
*** porrua has quit IRC20:17
*** eharney has quit IRC20:21
*** timcl has quit IRC20:26
*** pcaruana has joined #openstack-manila20:30
*** ChanServ changes topic to "10 bugs left until RC1"20:34
*** mlima has quit IRC20:47
*** kaisers_ has joined #openstack-manila21:00
*** akshai_ has joined #openstack-manila21:03
bswartzganso: did you write all that in the last 4 hours?21:03
bswartzor is it copied from cinder?21:03
gansobswartz: I wrote, and partially copied from migration in scheduler21:04
*** akshai has quit IRC21:05
bswartzdoes cinder use the scheduler for manage it this way?21:05
*** kaisers_ has quit IRC21:05
gansobswartz: scheduler migration code just validates host, which is the same purpose for manage21:05
gansobswartz: I have no idea, haven't checked cinder21:05
bswartzit seems very smart21:05
bswartzthis idea didn't occur to me21:05
smcginnisIIRC, we just check host in the scheduler.21:05
gansobswartz: I have not run tempest locally, waiting for jenkins21:09
*** gouthamr has quit IRC21:23
*** gouthamr has joined #openstack-manila21:23
*** gouthamr has quit IRC21:23
openstackgerritGoutham Pacha Ravi proposed openstack/manila: NetApp cDOT: Fix status updates for replicas  https://review.openstack.org/29180321:24
openstackgerritGoutham Pacha Ravi proposed openstack/manila: NetApp cDOT: Handle replicated snapshots  https://review.openstack.org/29324321:36
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Ensure Snapshots across replicas  https://review.openstack.org/29265921:36
*** xyang1 has quit IRC21:47
*** rickyrem has left #openstack-manila21:49
*** divab has joined #openstack-manila21:51
*** divab is now known as abhi21:51
*** dustins has quit IRC21:51
*** a_ta has quit IRC21:52
*** a_ta has joined #openstack-manila21:53
cknightbswartz, markstur: This is ready to merge: https://review.openstack.org/#/c/293708/21:53
marksturcknight, OK +221:55
cknightmarkstur: thanks21:55
*** porrua has joined #openstack-manila21:56
*** a_ta has quit IRC21:58
*** cknight has quit IRC22:07
*** akshai_ has quit IRC22:13
*** jasonsb has joined #openstack-manila22:36
*** porrua has quit IRC22:41
*** kaisers_ has joined #openstack-manila22:49
*** kaisers_ has quit IRC22:54
openstackgerritRodrigo Barbieri proposed openstack/manila: Fix manage API ignoring type extra specs  https://review.openstack.org/29426922:54
*** martyturner has quit IRC22:57
*** kaisers_ has joined #openstack-manila23:00
*** kaisers_ has quit IRC23:06
*** jasonsb has quit IRC23:22
openstackgerritMerged openstack/manila: NetApp cDOT: Raise ShareResourceNotFound in update_access  https://review.openstack.org/29370823:35
*** tpsilva has quit IRC23:47
*** a_ta has joined #openstack-manila23:54
*** dalgaaf has quit IRC23:58
*** a_ta has quit IRC23:59
« Wednesday, 2016-03-16 Index Friday, 2016-03-18 »

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!