Thursday, 2016-05-05

« Wednesday, 2016-05-04 Index Friday, 2016-05-06 »
*** yangyapeng has quit IRC00:05
*** vbellur has joined #openstack-manila00:15
*** gouthamr has joined #openstack-manila00:27
*** amit213 has quit IRC00:42
*** suyash has quit IRC00:49
*** esker has joined #openstack-manila00:52
*** esker has quit IRC00:57
*** absubram has joined #openstack-manila00:57
*** mtanino has quit IRC01:05
*** absubram has quit IRC01:12
*** yangyapeng has joined #openstack-manila01:13
*** amit213 has joined #openstack-manila01:13
*** yangyapeng has quit IRC01:15
*** amit213 has quit IRC01:16
*** esker has joined #openstack-manila01:28
*** esker has quit IRC01:33
*** gouthamr has quit IRC01:53
*** esker has joined #openstack-manila02:04
*** esker has quit IRC02:09
*** esker has joined #openstack-manila02:40
*** esker has quit IRC02:44
*** wangqun has joined #openstack-manila02:45
*** zhongjun_ has quit IRC03:03
*** zhongjun_ has joined #openstack-manila03:04
*** esker has joined #openstack-manila03:15
*** esker has quit IRC03:20
*** darkwsh has quit IRC03:29
*** esker has joined #openstack-manila03:51
*** esker has quit IRC03:57
*** esker has joined #openstack-manila04:27
*** esker has quit IRC04:32
*** esker has joined #openstack-manila05:03
*** esker has quit IRC05:08
openstackgerritMerged openstack/manila-ui: Updated from global requirements  https://review.openstack.org/31153905:20
openstackgerritMerged openstack/manila: Added docs for commit message tags  https://review.openstack.org/31000905:23
*** rraja has joined #openstack-manila05:26
*** xiaoxi has joined #openstack-manila05:27
*** gaurangt1 has joined #openstack-manila05:47
*** nobse has joined #openstack-manila05:55
*** openstackgerrit has quit IRC06:03
*** openstackgerrit has joined #openstack-manila06:03
openstackgerritRamana Raja proposed openstack/manila: cephfs_native: doc fixes  https://review.openstack.org/29544806:04
*** nobse has quit IRC06:05
*** gaurangt1 has left #openstack-manila06:14
*** nobse has joined #openstack-manila06:15
*** nobse has quit IRC06:16
*** yangyapeng has joined #openstack-manila06:31
*** soumitm has quit IRC06:33
*** soumitm has joined #openstack-manila06:34
*** soumitm has quit IRC06:39
*** rraja has quit IRC07:29
*** gaurangt has joined #openstack-manila07:33
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893007:51
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893007:54
*** lpetrut has joined #openstack-manila07:54
*** dmk0202 has joined #openstack-manila07:57
*** dmk0202 has quit IRC08:03
*** dmk0202 has joined #openstack-manila08:11
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893008:37
openstackgerritNidhiMittalHada proposed openstack/manila: Delete Snapshot: status wrongly set when busy  https://review.openstack.org/31199608:41
*** rraja has joined #openstack-manila08:49
vponomaryovtbarron: you are missing that it does not use tempest09:07
*** lpetrut has quit IRC09:07
*** lpetrut has joined #openstack-manila09:08
vponomaryovtbarron: it has its own config file09:08
*** lpetrut has quit IRC09:14
rrajaganso: vponomaryov: it'd be great if you can https://review.openstack.org/#/c/295448/ help in merging the patch sometime soon. the doc fixes would be very useful for the early testers of the cephfs driver. thanks!09:22
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893009:31
vponomaryovrraja: +2'ed, seek for another +209:39
rrajavponomaryov: super! once again thanks a lot! :)09:42
*** lpetrut has joined #openstack-manila09:47
*** ishu has joined #openstack-manila09:57
ishuHi all, can anyone help me in https://bugs.launchpad.net/manila/+bug/157853909:57
openstackLaunchpad bug 1578539 in Manila "Pop ssh instance in pool when calling remove function" [Undecided,New] - Assigned to iswarya vakati (v-iswarya)09:57
*** Guest22636 has joined #openstack-manila10:02
*** iswarya has joined #openstack-manila10:03
vponomaryovishu: what kind of help do you need?10:06
vponomaryovishu: confirm that there is dead-never-reached-insane-code? ))10:07
vponomaryovishu: bug topic is incomplete, it should what happens when described condition happens10:10
vponomaryovit should say*10:10
iswaryayes i am also about to ask this10:10
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893010:27
*** wangqun has quit IRC10:31
*** iswarya has quit IRC10:31
*** iswarya has joined #openstack-manila10:32
*** iswarya has quit IRC10:32
*** ishu has quit IRC10:33
openstackgerritValeriy Ponomaryov proposed openstack/manila-ui: Update detailed page for share servers  https://review.openstack.org/31263910:34
*** ishu has joined #openstack-manila10:35
ishuhi all10:36
*** yangyapeng has quit IRC10:46
*** Guest22636 has quit IRC11:24
openstackgerritCedric Zhuang proposed openstack/manila: Add EMC Unity Driver for Manila  https://review.openstack.org/31242311:31
openstackgerritMerged openstack/manila: cephfs_native: doc fixes  https://review.openstack.org/29544811:32
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893011:37
openstackgerritCedric Zhuang proposed openstack/manila: Add EMC Unity Driver for Manila  https://review.openstack.org/31242311:42
openstackgerritValeriy Ponomaryov proposed openstack/manila-ui: Update detailed page for share servers  https://review.openstack.org/31263911:50
*** tpsilva has joined #openstack-manila11:55
*** akerr has joined #openstack-manila11:57
*** aovchinnikov has joined #openstack-manila12:03
*** gaurangt has left #openstack-manila12:05
*** ekarlso has quit IRC12:06
*** ekarlso has joined #openstack-manila12:06
*** Guest22636 has joined #openstack-manila12:08
*** xiaoxi has quit IRC12:16
*** wangqun has joined #openstack-manila12:28
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893012:28
*** akerr_ has joined #openstack-manila12:30
openstackgerritNidhiMittalHada proposed openstack/python-manilaclient: manila list --all-tenants should display projectID  https://review.openstack.org/31201712:31
*** timcl has joined #openstack-manila12:32
*** akerr has quit IRC12:33
bswartzvponomaryov: can we switch the grenade job from experimental to check queue now or are we waiting on something still?12:41
vponomaryovbswartz: it is not merged12:41
vponomaryovbswartz: we are waiting for review only12:41
*** Guest22636 has quit IRC12:55
*** xyang1 has joined #openstack-manila12:58
*** yangyapeng has joined #openstack-manila12:59
*** akerr_ is now known as akerr12:59
*** gouthamr has joined #openstack-manila13:03
*** eharney has joined #openstack-manila13:04
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293913:11
*** yangyapeng has quit IRC13:14
*** ishu has quit IRC13:15
*** esker has joined #openstack-manila13:17
bswartzvponomaryov: link to review?13:23
vponomaryovbswartz: https://review.openstack.org/#/c/302801/13:23
bswartzvponomaryov: that's a manila change13:24
vponomaryovbswartz: yes13:24
bswartzif it hasn't merged, then why do grenade jobs pass on other manila changes?13:24
vponomaryovbswartz: because grenade runs without manila13:25
vponomaryovbswartz: only other projects succeed grenade in this job while that commit is not merged13:26
bswartzokay13:26
bswartzganso xyang1 toabctl markstur: https://review.openstack.org/#/c/302801/13:26
*** esker has quit IRC13:27
*** esker has joined #openstack-manila13:28
*** akshai has joined #openstack-manila13:31
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293913:32
*** yangyapeng has joined #openstack-manila13:35
*** lpetrut has quit IRC13:37
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893013:40
*** porrua has joined #openstack-manila13:40
*** mtanino has joined #openstack-manila13:41
tbarronvponomaryov: thanks for the reply regarding functional tests13:45
tbarronvponomaryov: in cinder I just do 'tox -efunctional' with no special setup13:45
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293913:45
tbarronvponomaryov: that didn't work in python-manilaclient, I got 'could not find requested endpoint in service catalog'13:46
tbarronvponomaryov: so i tried following iinstructions from commit 8760b9cd13:46
tbarronvponomaryov: but clearly I am confused13:47
*** openstackgerrit has quit IRC13:47
vponomaryovtbarron: cinder uses only one set of creds13:48
*** akapil has joined #openstack-manila13:48
vponomaryovcinderclient*13:48
vponomaryovcinderclient functional tests*13:48
*** openstackgerrit has joined #openstack-manila13:48
vponomaryovthat is why it works for you13:48
vponomaryovyou have env vars that are enough13:48
vponomaryovin manila we requrie two different sets for creds13:48
vponomaryovadmin and user13:48
vponomaryovyou cannot satisfy it only with env vars that are used for clients in common way13:49
vponomaryovfor that reason config file is used13:49
vponomaryovtbarron: our tests cover much more cases than cinderclient does13:50
tbarronvponomaryov: and 'tox -egenconfig' will make that file for me?13:50
*** lpetrut has joined #openstack-manila13:50
vponomaryovtbarron: it generates sample13:51
vponomaryovtbarron: setting of proper data stil lrequired13:51
vponomaryovtbarron: the only thing I would say is lost for now - it is generation of such config on devstack installation step13:51
vponomaryovbut, if you do it from different-host-than-devstack you will have problems anyway ))13:52
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Fix share server info in CGs created from CGs  https://review.openstack.org/31296313:52
tbarronvponomaryov: ok, I'll look at devstack setup, thanks.13:52
vponomaryovhttp://logs.openstack.org/49/304449/6/check/gate-manilaclient-dsvm-neutron-functional/4bb3418/logs/devstack-gate-tbarron: post_test_hook.txt.gz#_2016-05-03_13_04_42_65013:53
vponomaryovtbarron: http://logs.openstack.org/49/304449/6/check/gate-manilaclient-dsvm-neutron-functional/4bb3418/logs/devstack-gate-post_test_hook.txt.gz#_2016-05-03_13_04_42_65013:54
*** dustins has joined #openstack-manila13:55
*** martyturner has joined #openstack-manila13:56
vponomaryovtbarron: and be aware there are not only read-only tests!13:56
vponomaryovtbarron: lots of read-write tests too13:56
tbarronvponomaryov: kk, thanks!13:56
gansogouthamr: ping13:57
gouthamrganso: pong13:57
gansogouthamr: Hi Goutham, good morning13:57
gouthamrganso: Goodmorning! :)13:57
gansogouthamr: in your patch proposed just above ^, shouldnt reno be renamed according to bug ID?13:58
gansogouthamr: s/renamed/named13:58
gouthamrganso: other renos aren't so far.. would that be helpful?13:59
gansogouthamr: if we have another fix related to consistency groups api then the name would be the same, just a different hash13:59
gouthamrganso: https://github.com/openstack/cinder/tree/master/releasenotes/notes https://github.com/openstack/nova/tree/master/releasenotes/notes14:00
gouthamrganso: yes.. that's quite a problem14:00
gouthamrganso: i just noticed that issue this morning with a cinder patch..14:00
gansogouthamr: nobody is using the bug id :\14:00
*** cknight has joined #openstack-manila14:07
*** absubram has joined #openstack-manila14:18
dustinsgouthamr: ping14:23
*** krotscheck has quit IRC14:27
*** krotscheck has joined #openstack-manila14:33
gouthamrdustins: pong14:34
dustinsgouthamr: Good morning!14:35
dustinsgouthamr: Want to sync up real quick before the meeting?14:35
gouthamrdustins: Goodmorning! yep..14:35
*** akapil has quit IRC14:41
*** akapil has joined #openstack-manila14:45
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893014:45
*** absubram has quit IRC14:51
*** soumitm has joined #openstack-manila14:52
*** markstur_ has joined #openstack-manila14:54
*** a_ta has quit IRC14:57
*** lpetrut has quit IRC14:57
*** a_ta has joined #openstack-manila14:58
vponomaryovbswartz: meeting?15:01
gouthamrbswartz: meeting time check15:01
*** wangqun has quit IRC15:03
*** absubram has joined #openstack-manila15:05
*** esker has quit IRC15:05
*** soumitm has quit IRC15:05
*** soumitm has joined #openstack-manila15:14
*** lpetrut has joined #openstack-manila15:21
*** absubram has quit IRC15:26
*** absubram has joined #openstack-manila15:27
*** lpetrut has quit IRC15:35
*** lpetrut has joined #openstack-manila15:35
*** dmk0202 has quit IRC15:38
*** akapil has quit IRC15:40
*** adrianofr has quit IRC15:42
*** a_ta has quit IRC15:50
*** a_ta has joined #openstack-manila15:50
*** a_ta_ has joined #openstack-manila15:51
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893015:54
*** a_ta has quit IRC15:55
*** akshai has quit IRC15:58
*** akshai has joined #openstack-manila15:59
*** martyturner has quit IRC16:00
*** lpetrut has quit IRC16:01
*** rraja has quit IRC16:01
bswartzvponomaryov: it should be defined to False when not set16:01
bswartzI think the docs says that16:01
bswartzthat's important for backwards compatibility16:01
xyang1bswartz: You have said we should not mix the two16:02
gouthamrbswartz: it defaults to False16:02
xyang1bswartz: is that changed?16:02
vponomaryovI am talking about extra spec for "thin_provisioning"16:02
xyang1bswartz: you talked to me about it some time ago16:02
vponomaryovit is not set by default16:02
xyang1bswartz: I mean thin and thick16:03
vponomaryovbswartz: for your comment "(19:00:13) bswartz: you use the value from the share type, when creating the new share" - we do not use it is not set16:03
gouthamrvponomaryov: ah.. i meant teh capability reported.. if the driver doens't report anything, it will be set to False16:03
vponomaryovs/is/if/16:03
bswartzthere is no thick in manila16:04
bswartzjust thin=false16:04
xyang1bswartz: there is both, True and False16:04
markstur_the original stipulatino was that if a driver did both, they would need to report as separate pools16:04
bswartzthin can be both true and false16:04
*** martyturner has joined #openstack-manila16:04
bswartzthere is no extra_spec/capability called thick though16:04
bswartzthin=false MEANS thick16:05
xyang1bswartz: we had a discussion about calculation16:05
markstur_stipulation, I meant (but stipulatino is probably latin for that :) )16:05
gouthamryes.. lists upset max_over_subscription_ratio16:05
bswartzgouthamr: is there a bug?16:05
gouthamrbswartz: nope.. i thought it was intended behavipr16:06
gouthamrbehavior16:06
xyang1bswartz: you said we should not let a pool supports both true and false for thin16:06
bswartzoh wair16:06
bswartzoh wait16:06
bswartzyes I think you're right16:06
bswartzwhile most capabilities can be lists of [True, False], thin_provisioning is one that shouldn't be16:07
gouthamrbswartz: https://github.com/openstack/manila/blob/master/manila/scheduler/weighers/capacity.py#L6616:07
markstur_tell sched you'll do 20x, but then allocate full anyway16:07
bswartzsome backends want it to be, but it breaks manila's space reservation logic16:07
xyang1bswartz: we probably shoukd take another look at this16:07
gouthamrbswartz: if a driver reports [True,False] for thin_provisioning, scheduler assumes it is thin_provisioning=True16:07
markstur_I believe bswartz insisted that it would be bad to mix them in one pool16:08
gouthamrbswartz: sry, that particular weigher.. not the scheduler16:08
bswartzyeah.... if you mix them you create the following situation16:08
markstur_drivers can make their pool appear as 2 in reporting to manila16:08
*** yangyapeng has quit IRC16:08
xyang1bswartz: if we take share type into consideration, we can get the calculation right16:08
*** yangyape_ has joined #openstack-manila16:08
bswartzyou can fill up a pool with the maximum amount of thin provisioned shares (up to the max oversubscription ratio)16:08
bswartzbut then you can start to put thick shares on the same pool, which results in the oversub ratio going much much higher than it should16:09
gouthamrbswartz: exactly..16:09
xyang1bswartz: in reality, there are backends that can support both true and false in the same pool16:09
markstur_what if share type lets driver decide?16:09
bswartzxyang1: yes we have that with netapp16:09
bswartzwhich is why we have a long argument about thin provisioning after paris16:10
xyang1bswartz: the problem is in our calculation, we do not look at share types16:10
vponomaryovmarkstur_: share_type is used to filter backends16:10
*** akapil has joined #openstack-manila16:10
bswartzultimately I was convinced though that if we want to oversub ratio in manila to mean something, then we shouldn't mix thick and thin16:10
markstur_what if there is no extra-spec for thin_provisioning -- do we automatically set one?16:10
bswartzmarkstur_: IMO it should default to thick if not set16:10
xyang1bswartz: we can check whether it is true or false from a type in scheduler and calculation should be right16:10
markstur_I think that one is left as "whatever"16:10
vponomaryovmarkstur_: yes, not set equals to any16:11
xyang1markstur: we need to require that in extra specs if we do this16:11
bswartzokay it sounds like there's a subtle bug here16:11
bswartzin at least one place the default isn't being set16:11
bswartzand the filter is being ignored16:11
markstur_So are we backtracking on the initial agreement that a pool can only be thin or thick?16:12
markstur_as reported to manila16:12
bswartzand a share which *should* be thick is getting sent to a thin-only pool16:12
xyang1bswartz: markstur_ anyway, I think this is worth considering as an enhancement16:12
bswartzmarkstur_: I'm not16:12
bswartzmarkstur_: I think others might be trying to though16:13
xyang1bswartz: we get into that situation because we don't check if extra specs says true or false16:13
markstur_OK.  Sounds like we are at least "revisiting" it.  It's worth discussing, but I don't want it to just sneakily evolve16:13
xyang1bswartz: if we make that a requirement, we won't pick the wrong pool16:14
xyang1markstur_: sneakly?16:14
*** akapil has quit IRC16:14
markstur_yeah. if the scheduler considered the extra-spec with a required default that would help16:14
markstur_Re: sneakily -- what happened was we agreed on one thing and then patches tried to change that (a while ago)16:15
markstur_and now the same thing is happening -- not really sneaky... just because we are perhaps not remembering the old conversations16:15
markstur_That's what I meant.16:16
bswartzyeah we should probably put in patches which enforce what we say so that attempts to sneak around it are either obvious or don't work16:16
xyang1markstur_: we found problems in the original assumption and try to fix it16:16
* bswartz needs to run to lunch16:16
bswartzbbl16:16
xyang1bswartz: see you16:16
markstur_xyang1, Today we are discussing fixing it.  Yes.16:16
markstur_xyang1, In the past there was one patch that simply didn't follow the spec and would have broken things (for other drivers too I think?)16:17
markstur_So let's not allow this w/o fixing the scheduler to be aware.16:18
xyang1markstur_: I don't know.  we do not have spec in manila16:18
markstur_Spec is IRC logs  ???16:18
markstur_actually we have documented this for capabilities16:18
xyang1or wiki?16:18
markstur_better than spec.  living doc16:18
gouthamrmarkstur_ xyang1 bswartz: https://bugs.launchpad.net/manila/+bug/157871816:19
openstackLaunchpad bug 1578718 in Manila "max_over_subscription_ratio is miscalculated for drivers reporting list [True, False] for thin_provisioning" [Undecided,New]16:19
gouthamrwe need to triage ^16:19
markstur_which drivers report that?16:20
gouthamrmarkstur_: i was going to write some documentation on the usage of max_over_subscription_ratio for the admin guide16:20
gouthamrmarkstur_: netapp :|16:20
markstur_Today that is a NetApp bug.  Don't report both.16:21
gouthamrmarkstur_: yep.. on the tree, just the netapp driver16:21
akerrbut we can support both16:21
gouthamrmarkstur_: but it seems weird that i can do capability lists and then not do it for thin_provisioning16:22
xyang1gouthamr: it only works right if you have one pool support true or false but not both16:22
markstur_The comon capabilities spec says -- must be one or the other for this16:22
gouthamrxyang1: that's correct..16:22
xyang1gouthamr: bswartz talks to me about this a while back16:23
markstur_And if a driver wants to do both, they need to report as 2 separate pools16:23
xyang1gouthamr: to solve this, we should look at extra specs as I said earlier16:23
markstur_Was originally a NetApp proposal to require this (not naming names)  :)16:23
*** jcsp has quit IRC16:23
gouthamrmarkstur_: that could be a solution..16:23
gouthamrmarkstur_: lol :D16:23
markstur_That _was_ the requried solution16:23
markstur_But as discussed today -- if we also fix the scheduler to be smarter, then maybe we could revisit this restriction16:24
gouthamrxyang1: yes.. either we look at extra_specs and use it appropriately to do the max_over_subscription_ratio or we stop drivers from reporting a list16:24
gouthamrmarkstur_: yes..16:24
markstur_Unfortunately at least one driver was forced to use the dual-pool thing which they disagreed with16:25
akerror we take the cinder approach and make thin_provisioning one capability and thick_provisioning another16:25
markstur_Ack!16:25
markstur_Did he say "take the cinder approach"?16:25
xyang1akerr: cinder has the similar problem16:25
markstur_Yeah.  I guess consistency with Cinder and fixing this problem a really 2 different things.16:26
xyang1akerr: the thick provisioning one is not really used16:26
akerrxyang1: I guess that depends on the admin making the volume-types16:27
xyang1akerr: in cinder you need to create two pools too16:27
xyang1akerr: similar to manila16:27
markstur_If the scheduler looked at the extra-spec that would fix it.  I don't think there would be a problem as long as the extra-spec is reaquired to have a value or default and not "don't care" like most extra-specs16:28
xyang1akerr: I mean the cinder approach does not solve this problem either16:28
akerri'm not sure how 2 pools fixes this problem, you still have 1 of the pools reporting incorrect capacity if they're both actually pointing to the same set of storage16:28
markstur_If there was another reason why mixing thin/think in one pool was bad, I don't remember what it was16:28
xyang1akerr: one pool report true and another false, not to mix them16:29
*** rraja has joined #openstack-manila16:29
akerrxyang1: but then i have 1 set of disks with 2 values for capacity in the scheduler, which is correct?16:30
markstur_akerr, One pool reports itself 2 ways is hacky. Solves this problem somewhat but then you are double-reporting and relying on the status update to fix up stats after each create16:30
akerri agree, its a crazy situation16:30
*** rraja has quit IRC16:30
akerri'm just saying I don't think the 2 pools solution is correct either16:31
gouthamrmarkstur_: +1 the simplest solution is stop supporting both, and just report one16:31
*** dustins has quit IRC16:31
markstur_gouthamr, Yes.16:31
gouthamrmarkstur_: we may be able to live with that.. we'll go back to our drawing boards when bswartz comes back from lunch :P16:31
*** dustins has joined #openstack-manila16:32
* akerr wishes everything was thick provisioned. Then we don't have to worry about lying about capacity at all16:32
markstur_akerr, blasphemy16:33
* gouthamr finds that rusty lamp16:33
akerrmath is so much easier when 2-1 = 1, not 2 -1 = 2 sometimes16:33
xyang1markstur_: even if we check extra specs, weighing is still a challenge if the pool supports both16:33
markstur_akerr, I somewhat agree with you, but our companies really like their thin feature16:34
markstur_and in many cases it works great16:34
akerrwell ya, thin provisioning makes all this storage stuff have the proper ROI16:34
xyang1markstur_: check type can fix the filter, but which capacity to use for weighing16:34
akerrstill means you're lying to yourself :)16:34
markstur_It is a math conundrum though16:34
markstur_Fixed limits vs probabilities and statistics16:35
*** dustins has quit IRC16:36
akerrgouthamr: maybe we could make thick/thin a driver setting in the conf file?16:36
akerrstill nothing to stop them from making 2 backends pointing to the same disks, but then its on the admin for doing that16:37
xyang1bswartz: I'd like to know what you think, we should talk about this more16:37
markstur_xyang1, Not sure about weighing.  I thought we had a problem getting the spec value, but I haven't looked at it in a long time16:37
markstur_I'v got to go.  Be back in the office soon.16:38
xyang1markstur_: I thought that is available but I need to check too16:38
*** markstur_ has quit IRC16:38
akerrcould you do the weighing based on what the share is asking for?  so if its asking to be thin provisioned use that set of math, and if its asking to be thick use that set of math?  And then default to something (thin?) if its unspecified?16:38
xyang1markstur_: maybe weighing can be configurable16:38
xyang1akerr: weighing is for a entire backend, not a share16:39
akerrxyang1: but weight isn't calculated until a request comes in16:39
xyang1akerr: yes16:39
akerrwe know what the share's requirements are, so use that to help the weigher16:39
xyang1we could do that, I don't if that is the best way16:40
akerrI'm not sure it is either, but just spit-balling ideas here16:41
openstackgerritCedric Zhuang proposed openstack/manila: Add EMC Unity Driver for Manila  https://review.openstack.org/31242316:42
*** akshai has quit IRC16:44
*** akshai has joined #openstack-manila16:47
gouthamrakerr: yes, in the simplest solution, we can let the admin configure which pools are thin and which ones are meant to be thick16:50
*** akshai has quit IRC16:52
*** martyturner has quit IRC16:52
*** martyturner has joined #openstack-manila16:52
*** akshai has joined #openstack-manila16:54
*** martyturner has quit IRC16:57
*** adrianofr has joined #openstack-manila16:57
*** lpetrut has joined #openstack-manila16:58
*** cknight has quit IRC17:00
*** cknight has joined #openstack-manila17:01
*** martyturner has joined #openstack-manila17:03
openstackgerritDaniel Gonzalez Nothnagel proposed openstack/manila: Fix IPv6 standalone network plugin test  https://review.openstack.org/31306117:04
*** akapil has joined #openstack-manila17:11
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293917:15
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293917:15
openstackgerritGoutham Pacha Ravi proposed openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293917:15
* gouthamr doh17:15
*** suyash has joined #openstack-manila17:21
*** gluytium has quit IRC17:24
*** dustins has joined #openstack-manila17:27
gouthamrtbarron: and looks fancy, no? :)17:34
tbarrongouthamr: yeah, it looks good.  Now for some contnent :)17:35
gouthamrtbarron: yep.. on it :)17:35
tbarrongouthamr: i know, just teasing you.  Thanks for what you are doing!17:35
gouthamrtbarron: :)17:35
ameadegonna brain dump here17:39
ameadeso for porting messages17:39
ameadeI would actually really like all of the messages for all projects defined in one place17:39
ameadeso that the event_id of '000002' always means the same thing17:39
ameadeand we dont overload the ids17:39
ameadesmcginnis: you should help me think about this ^ :P17:40
ameadeeharney:  you too17:40
ameadefor context, i am gonna port messages to manila17:41
smcginnisameade: I think that makes sense, but I'm not sure how practical it would end up being.17:41
smcginnisameade: I was actually thinking at the summit, maybe we should do like we did with hacking checks.17:41
eharneyameade: just prefix them with another per-project number?17:41
ameademaybe?17:41
smcginnisSo C0001 is a Cinder specific thing, M0001 is a manila specific thing.17:41
ameadeor maybe we make event ids more like a uuid?17:41
smcginnisThen it's easy-ish to tell from the message ID where it came from.17:42
eharneysmcginnis: pretty much what i was thinking, just that instead of "C" i was imagining "100"17:42
smcginnisAnd we could potentially suck all of them in to one reference page or catalog.17:42
smcginniseharney: That works too.17:42
eharneybut maybe they don't need to be numeric, not sure17:42
smcginnisThe thing I like about the C|M|N is it's some kind of indication without needing to look up what it means.17:42
smcginnisBut we do have multiple projects with the same first initial...17:43
ameadeyou kind of get that with the resource_type in the error though17:43
smcginnisKeeping all numeric would actually be just fine with me.17:43
ameadein the message object i mean17:43
ameadethe main question is then, how do we avoid collisions?17:43
ameadei could move stuff to oslo?17:43
smcginnisameade: uuid's are just harder to parse if someones trying to match things up manually.17:44
ameadebe really careful?17:44
eharneytechnically if they're standardized error codes they should be prefixed with "volume" instead of "cinder" probably17:44
ameadesmcginnis: +117:44
smcginniseharney: Another good point.17:44
smcginnisameade: Haha, yeah - be really careful. ;)17:44
ameadenot hard to db migrate that at least17:44
ameadeok i tell you what, i'll do volume_...17:45
ameadethat way it's easy to do in other projects until we figure out what can be abstracted (if anything) to oslo17:45
smcginnisSounds reasonable to me.17:45
ameadeso that patch will be incoming soon17:45
eharneyameade: but what prefix do you use for brick? :)17:45
ameadeheh17:45
ameadeeharney: brick never talks to the db directly right?17:46
ameadeeharney: so likely we will have it bubble up to cinder or nova17:46
eharneyameade: sure but that doesn't mean it won't throw interesting errors for us to create messages from17:46
eharneyyeah, i suppose so17:46
*** akapil has quit IRC17:54
jseiler__We're working on a new manila driver. We've been referencing the manila minimum requirements page - http://docs.openstack.org/developer/manila/devref/driver_requirements.html18:00
jseiler__We've also found the manila API page - https://wiki.openstack.org/wiki/Manila/API. But that states that is is for how to develop applications that use manila.18:00
jseiler__In today's manila IRC meeting, I saw discussion of updating documentation.18:00
jseiler__Is there documentation for the interface that drivers must implement?18:00
*** cknight has quit IRC18:03
*** jcsp has joined #openstack-manila18:03
*** soumitm has quit IRC18:04
*** cknight has joined #openstack-manila18:04
*** akapil has joined #openstack-manila18:04
* bswartz scrolls back18:05
*** jasonsb has joined #openstack-manila18:07
*** martyturner has quit IRC18:07
bswartzxyang1: still here?18:09
xyang1bswartz: yes18:10
bswartzxyang1: so some more work is needed I think on max_oversubscribe_ratio -- if we want to allow mixing of thin and thick18:11
bswartzI think we could make the scheduler a bit smarter, and get a sane result18:11
*** martyturner has joined #openstack-manila18:11
bswartzwe would need filters in addition to weighers though18:11
xyang1bswartz: yes, look at type in filter and decide which logic to go?18:12
*** soumitm has joined #openstack-manila18:12
bswartzwell for one, I think it makes sense to have a filter even if the pool is just thin18:12
xyang1bswartz: there is already a filter, what else do you need18:13
bswartzbecause otherwise you could still exceed your max oversub ratio with just a lot of empty shares all on teh same pool18:13
bswartzoh18:13
bswartzgouthamr told me there wasn't18:13
xyang1we have capacity filter18:13
* bswartz goes to double check18:13
xyang1it is there18:13
bswartzdoes the capacity filter correctly handle virtual capacity?18:14
xyang1I think we just need to enhance it18:14
xyang1for one type, yes18:14
xyang1there is a problem if the pool supports both currently18:14
bswartzokay18:14
bswartzyes well we could fix that18:15
bswartzI guess that's probably the smartest thing to do18:15
*** lpetrut has quit IRC18:15
xyang1so if we say, if thin is true in extra specs, use the logic for thin, otherwise use the other logic18:15
bswartzhowever I'd feel safer if we had test code that verified the correctness of these filters and weighers18:15
xyang1there are actually lots of tests18:16
bswartzxyang1: well if we have both, we need to run 2 types of checks18:16
bswartzthere are unit tests but I'd like functional tests for the scheduler18:16
xyang1do we have any for scheduler? I don't remember18:17
bswartzprobably not18:17
bswartzwell we do for share types18:17
xyang1ok18:17
bswartzwe validate that share types are filtered properly I think18:17
bswartzbut I don't think we check that the capacity filter/weighers do the correct things18:17
xyang1probably not18:18
bswartzthat's one reason it's easy to get into a situation where we disagree about what the correct thing is18:18
bswartzif we just had a test, then we could point to it18:18
xyang1sure18:18
bswartzwe could argue about changing the test, but we could never argue about what the test does18:18
akerrbswartz: challenge accepted18:19
bswartzakerr: what challenge?18:19
akerr we could never argue about what the test does18:19
bswartzakerr: are you going to write the scheduler tests? or are you going to prove that it's possible to argue about what a tests does?18:20
bswartzlol18:20
*** dustins has quit IRC18:24
bswartzameade: who is doing the work on defining better error response at the HTTP level (the step before you get to user messages)18:25
*** akapil has quit IRC18:25
*** lpetrut has joined #openstack-manila18:26
*** akapil has joined #openstack-manila18:26
*** soumitm has quit IRC18:31
openstackgerritAlexey Ovchinnikov proposed openstack/manila: Container driver  https://review.openstack.org/30893018:37
*** dustins has joined #openstack-manila18:37
openstackgerritSumit Kumar proposed openstack/manila: Fix documentation for REST API version history and Scheduler  https://review.openstack.org/31310718:40
*** esker has joined #openstack-manila18:40
*** esker has quit IRC18:40
*** esker has joined #openstack-manila18:45
*** jcsp has quit IRC18:47
*** martyturner has quit IRC18:55
gouthamrjseiler__: hi, still around?18:58
*** martyturner has joined #openstack-manila18:58
jseiler__gouthamr: Yes, still here18:58
gouthamrjseiler__: not sure if anyone answered that question.. but yes, you found the driver minimum requirements doc... we're enhancing the driver interface docs.. so its easier to write code around those interfaces18:58
gouthamrjseiler__: i mean, the docstrings we'll add/change will help you do the right thing for each interface..18:59
*** timcl1 has joined #openstack-manila19:00
*** akapil has quit IRC19:00
jseiler__gouthamr: So, for reference, I should look at the docstrings in the driver code? Or is there a webpage somewhere too?19:01
*** timcl has quit IRC19:03
openstackgerritMerged openstack/manila: Add Grenade support to Manila  https://review.openstack.org/30280119:06
*** akapil has joined #openstack-manila19:09
*** akapil has joined #openstack-manila19:11
*** akapil has joined #openstack-manila19:11
*** akshai has quit IRC19:14
*** akshai has joined #openstack-manila19:14
*** akapil has quit IRC19:14
-openstackstatus- NOTICE: Gerrit is restarting to address performance issues related to a suspected memory leak19:21
*** akapil has joined #openstack-manila19:26
*** akshai has quit IRC19:29
*** akshai has joined #openstack-manila19:30
*** jcsp has joined #openstack-manila19:34
gouthamrjseiler__: sorry for the latency.. as of today, its not all in one place.. but http://docs.openstack.org/developer/manila/devref/index.html is a place to start19:43
gouthamrjseiler__: you may look at some of the reference drivers.. the individual driver documentation is on that page as well19:43
jseiler__gouthamr, OK. Thank your for the info and help. We were looking at the reference and other drivers too.19:45
gouthamrjseiler__: we'll have it all in one place soon-ish.. dustins is currently working on that :)19:45
* dustins looks around19:46
jseiler__gouthamr: My colleague that developer the cinder driver also used devstack to gather some info.  i.e. add print statements to gather info on the calls the come down to the driver19:46
gouthamrjseiler__: gee :P19:47
*** absubram has quit IRC19:47
dustinsjseiler__: Yeah, I'm going through the driver interface now and tidying it up, expect some clarifications soon :)19:47
jseiler__dustins: Thank you. I appreciate the info :)19:48
dustinsjseiler__: No problem!19:49
*** absubram has joined #openstack-manila19:52
*** gouthamr has quit IRC19:53
*** gouthamr has joined #openstack-manila19:56
*** akapil has quit IRC19:56
*** martyturner has quit IRC19:56
*** gouthamr has quit IRC19:56
*** martyturner has joined #openstack-manila19:59
*** akapil has joined #openstack-manila20:00
*** jcsp has quit IRC20:00
*** timcl has joined #openstack-manila20:01
*** akapil has quit IRC20:01
*** timcl1 has quit IRC20:03
*** martyturner has quit IRC20:03
gansocknight, bswartz, xyang1, markstur: Could you please take a look at https://review.openstack.org/#/c/299332/ ?20:06
gansocknight, bswartz, xyang1, markstur: that ^ needs to be backported as soon as it merges20:06
gansoxyang1: thanks! =D20:10
xyang1ganso: no problem:)20:10
openstackgerritRodrigo Barbieri proposed openstack/manila: Fix HNAS error with unconfined filesystems  https://review.openstack.org/30740720:16
marksturameade, Doesn't devstack already solve the single letter thing with m- c- n- q- (neutron stuck with old q-)?  Or do we already have issues with that? To devstacky?20:16
marksturs/To/Too/20:16
ameademeh we may as well be more explicit20:16
marksturameade, Yeah.20:20
-openstackstatus- NOTICE: Gerrit is restarting to revert incorrect changes to test result displays20:29
*** dmk0202 has joined #openstack-manila20:29
*** lpetrut has quit IRC20:32
*** akapil has joined #openstack-manila20:47
*** amit213 has joined #openstack-manila20:50
*** dustins has quit IRC20:51
*** akapil has quit IRC20:51
*** esker has quit IRC20:52
*** porrua has quit IRC20:56
openstackgerritMerged openstack/manila: Developer Reference: Adopt the openstackdocstheme  https://review.openstack.org/31293920:56
*** esker has joined #openstack-manila20:57
*** akapil has joined #openstack-manila21:00
*** lpetrut has joined #openstack-manila21:03
*** lpetrut has quit IRC21:08
*** akapil has quit IRC21:09
*** dustins has joined #openstack-manila21:11
*** gouthamr has joined #openstack-manila21:15
*** timcl has quit IRC21:21
*** jasonsb has quit IRC21:55
*** jcsp has joined #openstack-manila21:55
*** jasonsb has joined #openstack-manila21:58
*** akshai has quit IRC22:12
*** eharney has quit IRC22:15
*** absubram has quit IRC22:22
*** cknight has quit IRC22:22
*** jcsp has quit IRC22:28
*** a_ta_ has quit IRC22:36
*** a_ta has joined #openstack-manila22:36
*** akerr has quit IRC22:37
*** a_ta has quit IRC22:41
*** dustins has quit IRC22:42
*** dustins has joined #openstack-manila22:43
*** dustins has quit IRC22:43
*** david-lyle has quit IRC22:51
*** david-lyle has joined #openstack-manila22:53
*** david-lyle has quit IRC22:58
*** tpsilva has quit IRC23:07
*** sgotliv has quit IRC23:21
*** mtanino has quit IRC23:30
*** mtanino has joined #openstack-manila23:30
*** krotscheck has quit IRC23:31
*** krotscheck has joined #openstack-manila23:31
*** gouthamr_ has joined #openstack-manila23:31
*** gouthamr has quit IRC23:36
*** dmk0202 has quit IRC23:42
« Wednesday, 2016-05-04 Index Friday, 2016-05-06 »

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