| *** yangyapeng has quit IRC | 00:06 | |
| *** yangyapeng has joined #openstack-manila | 00:06 | |
| *** baojg has joined #openstack-manila | 00:07 | |
| *** baojg has quit IRC | 00:08 | |
| *** zhongjun has quit IRC | 00:11 | |
| *** yangyapeng has quit IRC | 00:11 | |
| *** zhongjun has joined #openstack-manila | 00:12 | |
| *** krotscheck has joined #openstack-manila | 00:42 | |
| *** erlon has quit IRC | 00:47 | |
| *** a_ta has joined #openstack-manila | 00:50 | |
| *** a_ta has quit IRC | 00:56 | |
| *** yangyapeng has joined #openstack-manila | 01:09 | |
| *** yangyapeng has quit IRC | 01:14 | |
| *** mtanino has joined #openstack-manila | 01:23 | |
| *** yangyapeng has joined #openstack-manila | 01:28 | |
| *** swamireddy1 has joined #openstack-manila | 01:31 | |
| *** jasonsb has joined #openstack-manila | 01:46 | |
| *** baojg has joined #openstack-manila | 01:46 | |
| *** swamireddy1 has quit IRC | 01:49 | |
| *** gouthamr has joined #openstack-manila | 01:52 | |
| *** jasonsb has quit IRC | 01:54 | |
| *** gouthamr_ has quit IRC | 01:55 | |
| *** tpsilva has quit IRC | 01:57 | |
| *** houming has joined #openstack-manila | 02:15 | |
| *** qeelee has joined #openstack-manila | 02:16 | |
| *** akshai_ has quit IRC | 02:32 | |
| *** akshai has joined #openstack-manila | 02:40 | |
| *** baojg has quit IRC | 02:48 | |
| *** Ben__ has joined #openstack-manila | 02:49 | |
| *** baojg has joined #openstack-manila | 02:50 | |
| *** gouthamr has quit IRC | 03:10 | |
| Ben__ | I try to use glusterfs volume on Manila but keep failing. I mount it on the ubuntu 14.04 host with nfs v3 and the host is set through Devstack .The m-shr.log shows [ Error encountered during initialization of driver 'GlusterfsShareDriver \ DriverNotInitialized: Share driver 'GlusterfsShareDriver' not initialized. ]. Anyone has idea about this? | 03:17 |
|---|---|---|
| *** cknight has joined #openstack-manila | 03:42 | |
| *** qeelee has quit IRC | 03:51 | |
| *** swamireddy1 has joined #openstack-manila | 04:02 | |
| *** jasonsb has joined #openstack-manila | 04:22 | |
| *** cknight has quit IRC | 04:28 | |
| *** qeelee has joined #openstack-manila | 04:58 | |
| *** akshai has quit IRC | 05:07 | |
| *** qeelee has quit IRC | 05:11 | |
| *** qeelee has joined #openstack-manila | 05:17 | |
| openstackgerrit | Felix Ma proposed openstack/manila: Fix wrong indent bug in _get_private_router https://review.openstack.org/296183 | 05:29 |
| *** Ben__ has quit IRC | 05:33 | |
| *** sgotliv has joined #openstack-manila | 05:34 | |
| *** lpetrut has joined #openstack-manila | 05:37 | |
| *** qeelee has quit IRC | 05:54 | |
| *** qeelee has joined #openstack-manila | 05:57 | |
| *** jasonsb has quit IRC | 06:06 | |
| *** lpetrut has quit IRC | 06:15 | |
| *** nkrinner has joined #openstack-manila | 06:17 | |
| *** qeelee has quit IRC | 06:21 | |
| *** lpetrut has joined #openstack-manila | 06:35 | |
| *** lpetrut has quit IRC | 06:41 | |
| *** sgotliv has quit IRC | 06:42 | |
| *** sgotliv has joined #openstack-manila | 06:42 | |
| openstackgerrit | daiki kato proposed openstack/manila: Fix typos https://review.openstack.org/296215 | 06:55 |
| *** qeelee has joined #openstack-manila | 07:22 | |
| *** qeelee has quit IRC | 07:32 | |
| *** furlongm_ is now known as furlongm | 07:36 | |
| *** sgotliv has quit IRC | 07:40 | |
| *** a_ta has joined #openstack-manila | 08:05 | |
| *** a_ta has quit IRC | 08:09 | |
| *** nkrinner has quit IRC | 08:12 | |
| *** nkrinner has joined #openstack-manila | 08:17 | |
| *** aovchinnikov has joined #openstack-manila | 08:21 | |
| *** permalac has joined #openstack-manila | 08:43 | |
| *** qeelee has joined #openstack-manila | 08:44 | |
| openstackgerrit | Marc Koderer proposed openstack/manila: Capitalize global var for clients https://review.openstack.org/296292 | 09:01 |
| *** openstackgerrit has quit IRC | 09:03 | |
| *** openstackgerrit has joined #openstack-manila | 09:04 | |
| *** sgotliv has joined #openstack-manila | 09:08 | |
| *** houming has quit IRC | 09:12 | |
| *** houming has joined #openstack-manila | 09:14 | |
| *** qeelee has quit IRC | 09:49 | |
| *** lpetrut has joined #openstack-manila | 10:00 | |
| *** baojg has quit IRC | 10:06 | |
| *** yangyapeng has quit IRC | 10:13 | |
| *** tpsilva has joined #openstack-manila | 10:26 | |
| *** qeelee has joined #openstack-manila | 10:32 | |
| openstackgerrit | Marc Koderer proposed openstack/manila: Increase logging for driver initialization https://review.openstack.org/296367 | 10:37 |
| *** baojg has joined #openstack-manila | 10:42 | |
| *** furlongm_ has joined #openstack-manila | 10:44 | |
| *** furlongm has quit IRC | 10:45 | |
| *** baojg has quit IRC | 10:47 | |
| *** furlongm has joined #openstack-manila | 10:55 | |
| *** furlongm_ has quit IRC | 10:55 | |
| *** ganso has joined #openstack-manila | 10:58 | |
| *** a_ta has joined #openstack-manila | 11:10 | |
| *** a_ta has quit IRC | 11:15 | |
| *** zhongjun has quit IRC | 11:25 | |
| *** zhongjun has joined #openstack-manila | 11:26 | |
| *** ociuhandu has quit IRC | 11:29 | |
| *** rraja has joined #openstack-manila | 11:40 | |
| *** yangyapeng has joined #openstack-manila | 11:42 | |
| mkoderer | ganso: ping | 11:51 |
| ganso | mkoderer: pong | 11:51 |
| mkoderer | ganso: I have some questions regarding the data service if you have a sec | 11:51 |
| ganso | mkoderer: sure ;) | 11:51 |
| mkoderer | ganso: how do you ensure that you don't have concurrent write access during the copy? | 11:52 |
| ganso | mkoderer: wait, are you talking about migration or this new patch I uploaded yesterday? https://review.openstack.org/295836 | 11:53 |
| mkoderer | ganso: the new patch basically | 11:53 |
| mkoderer | but can be also valid for mirgation | 11:53 |
| ganso | mkoderer: in migration I change the access rules to read only, so any client cannot perform any addional writes while it is migrating | 11:54 |
| mkoderer | ganso: ok that's fine | 11:54 |
| ganso | mkoderer: in the new patch I have not done that yet | 11:54 |
| mkoderer | ganso: ok I see | 11:54 |
| ganso | mkoderer: but it will be the same thing | 11:55 |
| mkoderer | ganso: so you change the access rules automatically? | 11:56 |
| mkoderer | I guess this could cause failures in application runnding in the cloud | 11:56 |
| ganso | mkoderer: https://github.com/openstack/manila/blob/master/manila/share/manager.py#L679 | 11:57 |
| ganso | mkoderer: what failures? | 11:57 |
| mkoderer | ganso: e.g. you run a database on a share fs and by setting it to read/only it could crash | 11:57 |
| mkoderer | ganso: ok I see.. we need to make sure that this aspect is well documented | 11:58 |
| mkoderer | I will ask our ops team about some real-live expriences for this topic | 11:58 |
| ganso | mkoderer: the only way so far to work around that is to perform migration through replication | 11:58 |
| *** david-lyle_ has joined #openstack-manila | 11:58 | |
| ganso | mkoderer: but then it needs to use a driver that implements replication | 11:59 |
| *** john_bar has joined #openstack-manila | 11:59 | |
| ganso | mkoderer: and implement optimized migration using such driver | 11:59 |
| mkoderer | ganso: ok I understand it know | 11:59 |
| *** ociuhandu has joined #openstack-manila | 11:59 | |
| mkoderer | ganso: I will try to test that feautre in our lab | 12:00 |
| ganso | mkoderer: great! :D | 12:00 |
| *** david-lyle has quit IRC | 12:00 | |
| *** yangyapeng has quit IRC | 12:01 | |
| mkoderer | ganso: IMHO manila need to have a spec process.. with that I could also include our ops people during design phase | 12:01 |
| *** yangyapeng has joined #openstack-manila | 12:02 | |
| ganso | mkoderer: I agree, there has been discussion a while back whether we should or not have it, we agreed that we should not, but now I think we should | 12:02 |
| ganso | mkoderer: the main problem raised was that after spec is approved, the feature ends up implemented differently due to necessary changes found during coding, and then the spec is never updated again | 12:03 |
| *** yangyape_ has joined #openstack-manila | 12:04 | |
| *** lpetrut has quit IRC | 12:04 | |
| mkoderer | ganso: yeah that's also correct.. but a design document can be read from ppl that aren't devs and I have some folks that would love to review manila specs | 12:05 |
| ganso | mkoderer: we are in the best moment to debate that again in our weekly meeting | 12:05 |
| mkoderer | ganso: correct | 12:06 |
| *** yangyapeng has quit IRC | 12:07 | |
| ganso | mkoderer: btw, I just remembered why I have not implemented the changing of access rules to read only in the new patch | 12:07 |
| ganso | mkoderer: I coded this patch before feature freeze so I had to review some details now | 12:08 |
| ganso | mkoderer: the reason is that drivers are going to be invoking those methods in the new patch | 12:08 |
| ganso | mkoderer: and the drivers themselves are able to change rules to read only | 12:08 |
| ganso | mkoderer: so the helper class does not need to do it | 12:09 |
| ganso | mkoderer: so let's say the driver wants to implement "create share from snapshot" using the data helper | 12:09 |
| ganso | mkoderer: it will first change rules to read only (or in this case won't be needed because it is a snapshot), and then call the helper to copy data from the snapshot to the new share | 12:10 |
| *** houming has quit IRC | 12:17 | |
| *** lpetrut has joined #openstack-manila | 12:19 | |
| ganso | mkoderer: I replied to your comment in https://review.openstack.org/295836 | 12:20 |
| mkoderer | ganso: ok I see .. can you add a check before coping? or a force option.. I just hesitant becuase manila could potential corrupt a share data which would be a worst case | 12:21 |
| mkoderer | ganso: ah btw if you have time to review some trivial fixes: https://review.openstack.org/#/c/296292/ and https://review.openstack.org/#/c/296367/ | 12:23 |
| openstackgerrit | Marc Koderer proposed openstack/manila: Increase logging for driver initialization https://review.openstack.org/296367 | 12:25 |
| ganso | mkoderer: sure | 12:27 |
| mkoderer | ganso: forget about https://review.openstack.org/296367 needed to push a new patchset | 12:28 |
| ganso | mkoderer: ok, I will keep an eye on it | 12:28 |
| *** qeelee has quit IRC | 12:31 | |
| *** ociuhandu has quit IRC | 12:34 | |
| *** ociuhandu has joined #openstack-manila | 12:35 | |
| openstackgerrit | Marc Koderer proposed openstack/manila: Increase logging for driver initialization https://review.openstack.org/296367 | 12:37 |
| *** qeelee has joined #openstack-manila | 12:39 | |
| *** a_ta has joined #openstack-manila | 12:40 | |
| *** a_ta has quit IRC | 12:45 | |
| *** lpetrut has quit IRC | 12:46 | |
| *** lpetrut has joined #openstack-manila | 12:46 | |
| *** baojg has joined #openstack-manila | 12:50 | |
| *** a_ta has joined #openstack-manila | 12:51 | |
| *** ociuhandu has quit IRC | 12:52 | |
| *** ociuhandu has joined #openstack-manila | 12:53 | |
| *** mtanino has quit IRC | 12:53 | |
| *** a_ta has quit IRC | 12:56 | |
| *** porrua has joined #openstack-manila | 12:57 | |
| *** akerr has joined #openstack-manila | 13:02 | |
| *** lpetrut has quit IRC | 13:03 | |
| *** david-lyle_ is now known as david-lyle | 13:06 | |
| *** akerr has quit IRC | 13:09 | |
| openstackgerrit | Marc Koderer proposed openstack/manila: Increase logging for driver initialization https://review.openstack.org/296367 | 13:16 |
| openstackgerrit | Merged openstack/manila: Capitalize global var for clients https://review.openstack.org/296292 | 13:17 |
| openstackgerrit | Marc Koderer proposed openstack/manila: Add binding_profile option for backends https://review.openstack.org/284034 | 13:24 |
| openstackgerrit | Marc Koderer proposed openstack/manila: Add neutron driver for binding https://review.openstack.org/283494 | 13:24 |
| bswartz | csaba: ping | 13:27 |
| bswartz | csaba rraja: have you noticed that glusterfs CI has been voting UNSTABLE for a while now? | 13:28 |
| *** qeelee has quit IRC | 13:28 | |
| rraja | bswartz: yes. thanks! i'll fix that soon. | 13:29 |
| bswartz | do you know the cause? | 13:29 |
| bswartz | I'm curious because you don't see UNSTABLE that often | 13:29 |
| bswartz | although UNSTABLE is definitely better than FAILURE | 13:30 |
| rraja | bswartz: it's happening while copying the logs to another location at the end of run. | 13:30 |
| bswartz | interesting | 13:30 |
| mkoderer | bswartz: had a short discussion with ganso about manila spec process.. we agreed that it would be beneficial for manila to have it | 13:31 |
| *** absubram has joined #openstack-manila | 13:32 | |
| mkoderer | I have some ops folks that would like to give feedback during design phase | 13:32 |
| *** absubram_ has joined #openstack-manila | 13:33 | |
| openstackgerrit | Rodrigo Barbieri proposed openstack/manila: Add Driver Data Service Helper https://review.openstack.org/295836 | 13:35 |
| *** absubram has quit IRC | 13:37 | |
| *** absubram_ is now known as absubram | 13:37 | |
| mkoderer | tbarron: ha, unbelivable how tricky it can be to add two log lines :) | 13:40 |
| tbarron | mkoderer: well, it's just a thought. I don't have strong feelings one way or another, but I know the manila codebase is still fairly clean and consistent. | 13:41 |
| tbarron | Has anyone looked - or is anyone looking - at how manila behaves if you configure two nodes to actively have the same host name relative to AMPQ? | 13:42 |
| tbarron | As opposed to using something like pacemaker to arbitrate. | 13:42 |
| mkoderer | tbarron: you mean two manila-share instances? | 13:42 |
| tbarron | mkoderer: yes, that's what I meant. | 13:42 |
| *** timcl has joined #openstack-manila | 13:43 | |
| tbarron | I'm asking the analogous question to what has been discussed for cinder. | 13:43 |
| tbarron | I think I know the answer, but am checking with my betters to make sure I give the right report back :-) | 13:43 |
| mkoderer | tbarron: I have it on m list of thing that I want to test | 13:44 |
| tbarron | mkoderer: so are you planning to test for races, etc? | 13:44 |
| mkoderer | tbarron: yes but I have a long todo list :) | 13:44 |
| *** mtanino has joined #openstack-manila | 13:44 | |
| tbarron | mkoderer: migrate? | 13:44 |
| tbarron | mkoderer: replication? | 13:45 |
| mkoderer | tbarron: maybe migrate.. but no replication | 13:45 |
| *** porrua has quit IRC | 13:46 | |
| tbarron | so unless corrected I will report that manila-share is essentially in the same position as cinder-volume: | 13:46 |
| tbarron | one may be able to get away with such a configuration but it has not been vetted. | 13:46 |
| *** Yogi1 has joined #openstack-manila | 13:46 | |
| mkoderer | the question is how are the amqp messages distributed | 13:47 |
| tbarron | There are likely races and potential for data corruption, though the chance of the latter is likely less with shares than volumes b/c filesystems generally know how to handle concurrent access. | 13:47 |
| mkoderer | tbarron: so we would need a stress test and a test env | 13:48 |
| bswartz | mkoderer: each release cycle we've voted on doing spec vs not doing specs, and each time the vote fails | 13:49 |
| bswartz | just to be clear, we do expect people to write functional specs and link them from their blueprints | 13:49 |
| bswartz | however we haven't yet felt the need to create a manila-specs repo and manage the specs process through gerrit | 13:49 |
| *** swamireddy1 has quit IRC | 13:49 | |
| ganso | vponomaryov, bswartz: Could you please take a look at this backport patch? https://review.openstack.org/#/c/294575/ | 13:49 |
| vponomaryov | ganso: I set +1 long time ago, there is no +2 power for common manila-core group | 13:50 |
| bswartz | the concern is always that specs are constantly out of date, so while they add value for discussion, they should not be formalized | 13:51 |
| ganso | vponomaryov: no? =( | 13:51 |
| ganso | vponomaryov: who can +2 to backport this? | 13:51 |
| bswartz | ganso vponomaryov: yeah the gerrit acls have a problem | 13:51 |
| vponomaryov | ganso: bswartz can and release team | 13:51 |
| bswartz | I can +2A things but nobody else can | 13:51 |
| bswartz | need to talk to stable PTL about getting the ACLs corrected | 13:52 |
| *** cknight has joined #openstack-manila | 13:53 | |
| *** dustins has joined #openstack-manila | 13:53 | |
| bswartz | vponomaryov ganso: there is no bug in the commit message -- is liberty gate broken? | 13:53 |
| ganso | bswartz: yes | 13:53 |
| ganso | bswartz: thanks | 13:54 |
| zhongjun | bswartz: hi :) Does the Netapp driver or manila has a plan to implement the feature about "Restrict the number of files in the share"? | 13:58 |
| bswartz | zhongjun: I haven't seen that feature suggested anywhere | 13:59 |
| bswartz | not all backends could support that | 13:59 |
| zhongjun | bswartz: Do you know which backend is interested in this? | 14:01 |
| bswartz | zhongjun: no I don't -- where is the feature proposed? | 14:01 |
| *** absubram has quit IRC | 14:02 | |
| zhongjun | bswartz: Does the Netapp could support this if it is needed? | 14:02 |
| mkoderer | cknight: can I get your feedback on https://review.openstack.org/#/c/274136/ | 14:03 |
| cknight | mkoderer: sure | 14:04 |
| zhongjun | This feature enables you to restrict the number of files in the share. | 14:05 |
| zhongjun | If the file quantity quota exceeds the quota, the system immediately | 14:05 |
| zhongjun | forbids write operations and prevents users from using extra files. | 14:05 |
| *** akerr has joined #openstack-manila | 14:05 | |
| *** a_ta has joined #openstack-manila | 14:06 | |
| bswartz | cknight: do you know if the qtree APIs on cDOT support file limits as well as byte limits? I know that 7-mode does both | 14:06 |
| zhongjun | bswartz: I just put it on the https://blueprints.launchpad.net/manila/+spec/file-quantity-quota-in-share, but not sure it is suitable for all backend. | 14:06 |
| bswartz | zhongjun: It's an interesting idea but I'm not sure if it has enough value to implement it -- especially if it's going to only work on a few backends | 14:07 |
| bswartz | zhongjun: are there users out there who really need this? | 14:07 |
| cknight | bswartz: I thought so. I will confirm. | 14:07 |
| mkoderer | bswartz: maybe we need it.. let me ask my folks tomorrow | 14:07 |
| bswartz | mkoderer: you need file limits in share? | 14:08 |
| cknight | bswartz: Yes. You can optionally specify the following additional attributes for the quota policy rule: | 14:08 |
| cknight | •User mapping | 14:08 |
| cknight | •Hard disk limit | 14:08 |
| cknight | •Hard file limit | 14:08 |
| cknight | •Threshold for disk limit | 14:08 |
| cknight | •Soft disk limit | 14:08 |
| cknight | •Soft file limit | 14:08 |
| bswartz | mkoderer: what is the point of that? is the default limit too low? too high? | 14:08 |
| *** openstack has joined #openstack-manila | 14:24 | |
| bswartz | zhongjun: tomorrow we'll be discussing topics for newton (austin) design summit | 14:25 |
| bswartz | zhongjun: if we have time we could also cover this topic, because we need convincing use cases before we can say the feature is something we want | 14:26 |
| zhongjun | bswartz: just add use case first :) | 14:27 |
| bswartz | thx zhongjun | 14:28 |
| zhongjun | thx you bswartz ;) | 14:29 |
| cknight | vponomaryov: ping | 14:30 |
| vponomaryov | cknight: pong | 14:31 |
| cknight | vponomaryov: The Manila service names in Keystone are manila and manilav2. But aren't the endpoint names share and sharev2? That's what devstack does. | 14:31 |
| vponomaryov | yes | 14:33 |
| vponomaryov | why? | 14:33 |
| cknight | vponomaryov: The quick start guide under review sets both service & endpoint names to manila and manilav2. I need to know if that is correct. https://review.openstack.org/#/c/274136/3..4/doc/source/adminref/quick_start.rst | 14:33 |
| vponomaryov | cknight: manilaclient has dependency on "sharev2" value | 14:36 |
| vponomaryov | cknight: so, service type should stay "share" and "sharev2" | 14:36 |
| cknight | vponomaryov: OK, thanks, I'll -1 the review (again). | 14:37 |
| *** vbellur has quit IRC | 14:39 | |
| cknight | mkoderer: ^^ | 14:40 |
| zhongjun | vponomaryov: Do you know what is the original reason, when we delete a share, | 14:43 |
| zhongjun | we just change the status to "deleted" in db table, Not deleted the row in the db table. | 14:43 |
| vponomaryov | zhongjun: it is "soft" deletion | 14:44 |
| vponomaryov | zhongjun: common approach in OpenStack | 14:44 |
| openstackgerrit | Marc Koderer proposed openstack/manila: Fix quick start guide to new auth mechanism https://review.openstack.org/296498 | 14:45 |
| zhongjun | vponomaryov: What can be used with this "deleted" value. | 14:46 |
| *** david-lyle_ is now known as david-lyle | 14:47 | |
| mkoderer | dgonzalez: can you have a look at cknight comment.. I was so sure that it worked.. (only when you have time) | 14:47 |
| zhongjun | vponomaryov: Does it will cause the database too big? | 14:47 |
| vponomaryov | zhongjun: yes, it makes table grow | 14:48 |
| *** akshai has joined #openstack-manila | 14:48 | |
| dgonzalez | mkoderer: i will have a look at it | 14:49 |
| vponomaryov | zhongjun: and I am aware about such concern for operators that are forced to write scripts with DB sleanups | 14:50 |
| zhongjun | vponomaryov: Does this way have the advantage of this "deleted" data in db? | 14:50 |
| vponomaryov | zhongjun: I can think about only two reasons: 1) keep consistency of all foreign keys, that implement some kind of matrix where lots of elements refer to each other | 14:51 |
| vponomaryov | zhongjun: 2) debug reasons | 14:52 |
| bswartz | 3) soft deletes are way faster than hard deletes | 14:53 |
| vponomaryov | zhongjun: I guess there should be such discussion in openstack mailing-list or some summits notes | 14:53 |
| *** baojg has quit IRC | 15:03 | |
| zhongjun | vponomaryov, bswartz: thx, This is one of the original problem, Maybe before experienced many discussions in OpenStack.I try to find it. Maybe we could open discuss it in mailling-list. | 15:03 |
| *** vbellur has joined #openstack-manila | 15:03 | |
| bswartz | zhongjun: you're proposing switching away from soft deletes to something else? | 15:04 |
| bswartz | zhongjun: that's a discussion that's larger than just Manila -- we inherited soft deletes from cinder and it's a widely-used technique in the community | 15:04 |
| *** absubram has joined #openstack-manila | 15:05 | |
| zhongjun | bswartz: Oh, that is the real reason. | 15:06 |
| zhongjun | bswartz: I just want to find the reason, why the choose this, and explain to other people. | 15:07 |
| zhongjun | s/the/they | 15:07 |
| bswartz | oh | 15:07 |
| bswartz | you'll have to look back very far in mail list archives | 15:07 |
| bswartz | that decision was made years ago | 15:07 |
| bswartz | but I'm sure there's a record of the discussion with an explanation of the benefits | 15:07 |
| zhongjun | Oh, that is must have the evidence. I believe it will be find. | 15:10 |
| *** openstackstatus has joined #openstack-manila | 15:12 | |
| *** ChanServ sets mode: +v openstackstatus | 15:12 | |
| *** lpetrut has joined #openstack-manila | 15:12 | |
| *** sgotliv has quit IRC | 15:17 | |
| *** adrianofr_ has joined #openstack-manila | 15:52 | |
| *** lpetrut has quit IRC | 15:59 | |
| *** lpetrut has joined #openstack-manila | 16:03 | |
| *** erlon has joined #openstack-manila | 16:08 | |
| *** sage has quit IRC | 16:18 | |
| *** sage has joined #openstack-manila | 16:19 | |
| *** timcl has quit IRC | 16:19 | |
| *** jasonsb has joined #openstack-manila | 16:30 | |
| *** david-lyle has quit IRC | 16:33 | |
| *** david-lyle has joined #openstack-manila | 16:33 | |
| *** jasonsb has quit IRC | 16:38 | |
| *** jasonsb has joined #openstack-manila | 16:39 | |
| openstackgerrit | Thomas Bechtold proposed openstack/manila: Remove PyMySQL and psycopg2 from test-requirements.txt https://review.openstack.org/296600 | 16:40 |
| *** timcl has joined #openstack-manila | 16:47 | |
| *** vbellur has quit IRC | 16:47 | |
| *** lpetrut has quit IRC | 16:49 | |
| *** gouthamr has quit IRC | 16:53 | |
| ameade | why are we still ignoring test cleanup errors in the gate? | 16:54 |
| ameade | bswartz, vponomaryov ^^? | 16:54 |
| bswartz | ameade: because the gate tests wouldn't pass otherwise | 16:56 |
| ameade | should we not fix that? | 16:56 |
| bswartz | ameade: yes we should | 16:56 |
| vponomaryov | ameade: it is designed for generic driver | 16:57 |
| ameade | our CI is broken due to a patch that caused cleanup errors | 16:57 |
| vponomaryov | ameade: and instability of its deps | 16:57 |
| ameade | mm | 16:57 |
| vponomaryov | ameade: all other driverss dhould not ignore cleanup errors | 16:57 |
| bswartz | we assume, but we do not know, that generic driver fails sometimes due to slowness of nova, cinder, and neutron | 16:57 |
| vponomaryov | ameade: let me guess, cleanup of replicas broke somethng? | 16:58 |
| ameade | this patch https://review.openstack.org/#/c/294269/ | 16:58 |
| smcginnis | Who you calling slow? | 16:58 |
| smcginnis | :) | 16:58 |
| vponomaryov | big brother is watching us )) | 17:00 |
| bswartz | it's the interactions between nova and cinder | 17:00 |
| *** lpetrut has joined #openstack-manila | 17:01 | |
| * bswartz notes that smcginnis has an IRC pounce for "cinder" and we must use a different name if we want to disparage them in our channel | 17:01 | |
| bswartz | lol | 17:01 |
| ganso | ameade: how did it break? | 17:01 |
| smcginnis | Haha | 17:01 |
| vponomaryov | bswartz: "that slow service"? | 17:02 |
| *** a_ta has quit IRC | 17:03 | |
| *** a_ta has joined #openstack-manila | 17:03 | |
| *** a_ta has quit IRC | 17:03 | |
| *** a_ta has joined #openstack-manila | 17:04 | |
| vponomaryov | ganso: test stopped cleaning up all created shares and share type cannot be removed | 17:05 |
| vponomaryov | ganso: as "in use" | 17:05 |
| akerr | ganso: I think I found a bug in one of your patches that is causing our CI to fail constantly with cleanup errors: https://review.openstack.org/#/c/294269/4/manila_tempest_tests/tests/api/admin/test_share_manage.py The assertRaises call will create a share in error status but never cleans it up, then when tempest goes to clean up the share type we get a 400 error because the share type is still in use | 17:06 |
| ganso | akerr: it is expected to not remove share in assertRaises because manage reuses share if it is managed again with proper parameters | 17:06 |
| ganso | akerr: but I think I see where the problem is | 17:06 |
| *** sedwards has joined #openstack-manila | 17:07 | |
| ganso | akerr, vponomaryov: indeed it is not removing the share type... I mistakenly removed that, now looking closer, it was the share type that was being removed, not the share | 17:07 |
| vponomaryov | ganso: "manage reuses share if it is managed" where did you hear that? | 17:07 |
| *** rhagarty has quit IRC | 17:07 | |
| ganso | vponomaryov: https://github.com/openstack/manila/blob/master/manila/share/api.py#L537 | 17:08 |
| *** rhagarty has joined #openstack-manila | 17:08 | |
| ganso | akerr: possibly line 174 in original code of that change should not have been removed | 17:09 |
| vponomaryov | ganso: share type ID is used as filter | 17:09 |
| *** jasonsb has quit IRC | 17:09 | |
| vponomaryov | ganso: so, when you use new share type, you create new share | 17:09 |
| vponomaryov | ganso; in this test in anycase | 17:09 |
| *** gouthamr has joined #openstack-manila | 17:10 | |
| ganso | vponomaryov: oh I see, indeed | 17:10 |
| ganso | akerr: what does your CI do differently that fails? | 17:12 |
| *** porrua has joined #openstack-manila | 17:12 | |
| ganso | akerr: manage operation should not create share on backend | 17:12 |
| akerr | ganso: we don't skip cleanup failures | 17:12 |
| *** martyturner has joined #openstack-manila | 17:13 | |
| vponomaryov | ganso: generic driver skips cleanup errors, all other drivers just were not considered merging this commit | 17:14 |
| *** swamireddy1 has joined #openstack-manila | 17:16 | |
| ganso | vponomaryov: only netapp failed | 17:16 |
| ganso | vponomaryov: due to that change | 17:16 |
| vponomaryov | ganso: ZFsonLinux does not have support for "manage" yet | 17:16 |
| vponomaryov | ganso: it is for example | 17:16 |
| vponomaryov | ganso: so, broken test is just skipped there | 17:16 |
| *** sedwards has quit IRC | 17:19 | |
| akerr | ganso: the problem isn't that it makes the share on the backend, its that a DB entry for the share is created in error status and then never cleaned up, so manila says the share type is still in use | 17:19 |
| bswartz | are we cleaning up a whole share type in the test cleanup? | 17:20 |
| bswartz | if so, shouldn't we delete all the shares of that type first? | 17:20 |
| gouthamr | bswartz: logically, the share must be deleted first.. | 17:21 |
| vponomaryov | bswartz: this is the reason of broken CI for NetApp No-ss job | 17:23 |
| bswartz | what I'm trying to say is that if cleanups are failing because stuff still exists, then the cleanups should be more aggressive about cleaning stuff | 17:23 |
| akerr | vponomaryov: is there a manage_share function in base.py that adds cleanup automatically that we could call in the assertRaises() instead of invoking the share_client directly? | 17:24 |
| bswartz | cleanup operations should recursively delete everything they need to to ensure the object in question is actually cleaned | 17:24 |
| vponomaryov | akerr: not yet | 17:24 |
| bswartz | otherwise cleanups will continue to fail | 17:24 |
| vponomaryov | akerr: but it is righttime to add such | 17:24 |
| akerr | vponomaryov: ok, that seems like the right way to handle it | 17:24 |
| vponomaryov | bswartz: tests can break any cleanup logic, it is exactly this case | 17:24 |
| ganso | vponomaryov: curiously, default setting for supressing tests is False, but CI script says default is True... I believed my CI was not supressing errors, but it is =( | 17:25 |
| *** swamireddy1 has quit IRC | 17:34 | |
| *** a_ta has quit IRC | 17:34 | |
| *** a_ta has joined #openstack-manila | 17:34 | |
| akerr | hmm this won't be that straightforward to fix. Since it raises a 409 exception in the client layer we can't easily grab the share_id from the create call... I guess we'd need to do a search on shares of type x in error status and delete those? | 17:37 |
| *** a_ta has quit IRC | 17:39 | |
| vponomaryov | akerr: if we cannot receive it then it should not be created | 17:40 |
| akerr | vponomaryov: I just mean that the exception in python will change the control flow of the code prior to the variable being set | 17:41 |
| vponomaryov | akerr: I don't opposite I say design is wrong, only 201 (at least, 200 and 202) code should be returned after successfully created resource | 17:43 |
| vponomaryov | akerr: if we get 409, then resource should not be created | 17:43 |
| akerr | vponomaryov: even a DB entry shouldn't be created? | 17:43 |
| vponomaryov | akerr: Db is the only thing Manila is aware about | 17:44 |
| vponomaryov | akerr: especially API | 17:44 |
| vponomaryov | akerr: so, creation of DB record should be after check for that error | 17:44 |
| vponomaryov | akerr: we get in broken test | 17:44 |
| akerr | vponomaryov: ok | 17:45 |
| akerr | vponomaryov: makes sense | 17:45 |
| *** lpetrut has quit IRC | 17:47 | |
| *** lpetrut has joined #openstack-manila | 17:47 | |
| ganso | vponomaryov: validation is in the scheduler, DB record is created before sending to the scheduler | 17:48 |
| vponomaryov | ganso: you say it so it cannot be changed | 17:49 |
| vponomaryov | ganso: )) | 17:49 |
| ganso | vponomaryov: I am checking if it is not necessary to send share info to scheduler | 17:49 |
| vponomaryov | ganso: all checks should be performed only before creation of DB record that will mean we return 201 | 17:50 |
| ganso | vponomaryov: if not necessary, then scheduler can validate without needing the share created, and share can be created later | 17:50 |
| ganso | vponomaryov: indeed it is not necessary, it just needs the host | 17:50 |
| vponomaryov | akerr: is there launchpad bug for it? | 17:51 |
| *** lpetrut has quit IRC | 17:52 | |
| akerr | vponomaryov: not yet, ameade and i are verifying that we can repro manually | 17:52 |
| vponomaryov | bswartz: looks like we have candidate for RC2, wrong behaviour of API service using "manage" command | 17:53 |
| bswartz | client or server project? | 17:53 |
| vponomaryov | server | 17:53 |
| akerr | vponomaryov: verified, we get a db entry | 17:53 |
| bswartz | tag bug with "mitaka-rc-potential" and let's hurry and get the fix into newton | 17:54 |
| bswartz | decision for what (if anything) is included in RC2 will happen early next week most likely | 17:54 |
| akerr | bswartz: vponomaryov: creating LP bug | 17:55 |
| vponomaryov | bswartz: who makes decision? | 17:55 |
| bswartz | core team | 17:55 |
| vponomaryov | release or we? | 17:55 |
| bswartz | us | 17:55 |
| vponomaryov | ok | 17:55 |
| bswartz | We have to decide if it's a ship stopper | 17:55 |
| bswartz | because at this point only really bad thing should be fixed | 17:56 |
| vponomaryov | itis not critical, but it is regression | 17:56 |
| bswartz | we can't make the decision to backport until we have a merged fix in newton though | 17:56 |
| bswartz | it helps to see what the change actually looks like so we can assess risk of additional regressions caused by the bugfix | 17:57 |
| ameade | akerr: vponomaryov: https://bugs.launchpad.net/manila/+bug/1561139 | 17:59 |
| openstack | Launchpad bug 1561139 in Manila "Managing a share with an invalid share type results in a share created" [Undecided,New] | 17:59 |
| akerr | ameade: thx | 18:00 |
| akerr | bswartz: ^ | 18:00 |
| bswartz | akerr: this was discovered by a failing test, no? can we mention which test it was in the bug, with perhaps a link to relavant logs? | 18:02 |
| ganso | vponomaryov, akerr: scheduler needs request_spec, which needs share to be created. Unless we create an empty request spec, a possible fix is to delete share in API before returning response | 18:03 |
| akerr | bswartz: sure | 18:04 |
| gouthamr | ganso: can't we delete the DB record on failure for manage? | 18:04 |
| ganso | gouthamr: yes we can, I am suggesting that | 18:05 |
| gouthamr | ganso: its a 409 conflict, so you don't need to save the state.. the request will be re-attempted | 18:05 |
| bswartz | hard delete or soft delete? | 18:05 |
| ganso | gouthamr: save what state? | 18:05 |
| bswartz | it seems crappy if we have only one code path that creates shares and then automatically deletes them | 18:05 |
| gouthamr | you created a record for a share that could not be created because of a conflict.. | 18:06 |
| gouthamr | bswartz: no, the operation is not complete yet | 18:06 |
| bswartz | it would be nice if the API service could detect the error itself without needing RPC at all | 18:06 |
| gouthamr | bswartz: it isn't returning a 202 and failing later.. | 18:06 |
| gouthamr | bswartz: it is failing, with a 409.. | 18:06 |
| ganso | gouthamr: delete DB record here https://github.com/openstack/manila/blob/master/manila/share/api.py#L551 | 18:06 |
| gouthamr | bswartz: so its a failure.. we shouldn't be saving any state in Manila | 18:07 |
| gouthamr | bswartz: its a failed request. | 18:07 |
| gouthamr | API request* | 18:07 |
| gouthamr | ganso: yep | 18:07 |
| *** vbellur has joined #openstack-manila | 18:08 | |
| *** sedwards has joined #openstack-manila | 18:12 | |
| *** krotscheck has quit IRC | 18:14 | |
| bswartz | gouthamr: you mean the scheduler isn't involved at all? | 18:14 |
| *** krotscheck has joined #openstack-manila | 18:15 | |
| bswartz | in that case why do we not reorder the logic to detect this error before writing anything to the DB? | 18:15 |
| openstackgerrit | Merged openstack/puppet-manila: Fix typos https://review.openstack.org/294375 | 18:15 |
| gouthamr | bswartz: we're creating the record before sending the scheduler request.. two ways about this, don't create until the scheduler returns or just delete if the scheduling fails. | 18:15 |
| bswartz | gouthamr: what I was proposing was option #3 -- don't invoke the scheduler RPC and instead import the scheduler code and invoke it from the API service | 18:17 |
| bswartz | ^ could this work or am I missing something | 18:17 |
| *** sedwards has quit IRC | 18:17 | |
| ganso | gouthamr: what's the problem of invoking the scheduler through call_context.call? it is synchronous and you don't need to import scheduler | 18:21 |
| gouthamr | ganso: i think rpc is fine.. the scheduler has the pool info.. im not sure how you can get all of that with importing.. | 18:22 |
| *** esker has joined #openstack-manila | 18:23 | |
| ganso | gouthamr: oh ok | 18:23 |
| ganso | gouthamr: and how can we hard delete it after it is created? | 18:23 |
| ganso | gouthamr: I don't see any db calls or parameters to hard delete | 18:23 |
| gouthamr | ganso: soft delete must be okay.. | 18:23 |
| bswartz | never creating it in the first place would be preferable | 18:24 |
| bswartz | does the scheduler need an share ID in order to respond with the error? | 18:24 |
| ganso | bswartz: scheduler does not need share, but need request_spec, and request_spec needs share | 18:24 |
| vponomaryov | ganso; bswartz is right, it should not be created, if you need to provide something to scheduler, then just use dict with all required data | 18:24 |
| gouthamr | i agree ^ | 18:25 |
| ganso | vponomaryov: this data: https://github.com/openstack/manila/blob/master/manila/share/api.py#L562 | 18:25 |
| *** krotscheck has quit IRC | 18:26 | |
| *** krotscheck has joined #openstack-manila | 18:26 | |
| gouthamr | ganso: don't you just need the share type data? | 18:27 |
| ganso | gouthamr: I don't know, I have never tried request_spec with only share_properties and share_instance_properties empty | 18:28 |
| bswartz | I'm getting the impression that what we need is a new RPC that the API service can call to make this check with minimal inputs | 18:28 |
| bswartz | however a new RPC would be a bad idea for backporting to mitaka | 18:28 |
| ganso | bswartz: if behavior can be the same while supplying a request spec with only share_type then an additional RPC is not needed | 18:29 |
| *** cknight has quit IRC | 18:30 | |
| gouthamr | ganso: going through the logic there.. is there a way you can remove the db create line.. and return the host? | 18:33 |
| gouthamr | ganso: from the scheduler | 18:33 |
| ganso | gouthamr: the scheduler does not create db entry | 18:33 |
| ganso | gouthamr: it calls manager code | 18:33 |
| gouthamr | ganso: we don't need to | 18:33 |
| ganso | gouthamr: API can call manager code once scheduler validates | 18:34 |
| gouthamr | ganso: we can make the scheduler return to the API and the API can call the share manager RPCAPI | 18:34 |
| ganso | gouthamr: yes | 18:34 |
| gouthamr | ganso: that way, you can create the db record only if the scheduler gives you the host. | 18:34 |
| ganso | gouthamr: yes | 18:34 |
| * gouthamr should read ganso's responses before typing away.. :P | 18:35 | |
| ganso | gouthamr: lol | 18:36 |
| *** cknight has joined #openstack-manila | 18:37 | |
| *** krotscheck has quit IRC | 18:38 | |
| *** krotscheck has joined #openstack-manila | 18:39 | |
| *** permalac has quit IRC | 18:39 | |
| openstackgerrit | Valeriy Ponomaryov proposed openstack/manila: Add caution to test-requirements https://review.openstack.org/296665 | 18:39 |
| *** sgotliv has joined #openstack-manila | 18:49 | |
| *** openstack has joined #openstack-manila | 19:23 | |
| openstackgerrit | Rodrigo Barbieri proposed openstack/manila: Fix Manage API creating DB record for invalid type https://review.openstack.org/296706 | 19:24 |
| *** DuncanT_ is now known as DuncanT | 19:24 | |
| *** martyturner has quit IRC | 19:24 | |
| *** timcl1 has joined #openstack-manila | 19:26 | |
| *** timcl has quit IRC | 19:27 | |
| *** martyturner has joined #openstack-manila | 19:30 | |
| *** dustins_ has joined #openstack-manila | 19:34 | |
| *** dustins has quit IRC | 19:36 | |
| *** Yogi1 has quit IRC | 19:55 | |
| openstackgerrit | Emilien Macchi proposed openstack/puppet-manila: Release 8.0.0 https://review.openstack.org/296730 | 20:03 |
| *** ociuhandu has quit IRC | 20:05 | |
| *** swamireddy1 has quit IRC | 20:07 | |
| openstackgerrit | Emilien Macchi proposed openstack/puppet-manila: Release 8.0.0 https://review.openstack.org/296752 | 20:07 |
| gouthamr | ganso: ping | 20:08 |
| ganso | gouthamr: pong | 20:08 |
| gouthamr | ganso: https://github.com/openstack/manila/blob/master/manila/api/v1/share_manage.py#L61 | 20:09 |
| gouthamr | ganso: do you think thats right? | 20:09 |
| gouthamr | ManilaException is a base class for a bunch of exceptions... maybe you should specifically expect the InvalidHost exception? | 20:10 |
| gouthamr | ganso: just increasing the scope of your commit there.. but we should correct that behavior :) | 20:11 |
| *** lpetrut has joined #openstack-manila | 20:12 | |
| ganso | gouthamr: yes, I am checking other exceptions just a sec | 20:12 |
| ganso | gouthamr: share.api.manage raises ManilaException | 20:12 |
| ganso | gouthamr: if share already exists | 20:13 |
| ganso | gouthamr: that would need to be fixed as well | 20:13 |
| ganso | gouthamr: I agree that catch block should capture only specific exception, not the base class | 20:13 |
| ganso | gouthamr: but we need to also remember that if we want the fix in RC2 we need to minimize impact, maybe those changes should not go into same change-id | 20:14 |
| *** timcl1 has quit IRC | 20:14 | |
| gouthamr | ganso: sure... i just thought that was a disparity, | 20:15 |
| gouthamr | ganso: thanks.. will review the patch | 20:15 |
| *** openstack has joined #openstack-manila | 20:33 | |
| *** cknight has joined #openstack-manila | 20:34 | |
| *** gouthamr has quit IRC | 20:35 | |
| *** martyturner has joined #openstack-manila | 20:39 | |
| *** cknight has quit IRC | 20:45 | |
| *** rraja has quit IRC | 20:47 | |
| *** porrua has quit IRC | 20:56 | |
| *** david-lyle has quit IRC | 20:59 | |
| *** cfey has quit IRC | 21:05 | |
| *** david-lyle has joined #openstack-manila | 21:05 | |
| *** a_ta has quit IRC | 21:14 | |
| *** a_ta has joined #openstack-manila | 21:16 | |
| *** tpsilva has quit IRC | 21:17 | |
| *** david-lyle has quit IRC | 21:21 | |
| *** cfey has joined #openstack-manila | 21:22 | |
| *** david-lyle has joined #openstack-manila | 21:23 | |
| openstackgerrit | Merged openstack/puppet-manila: Release 8.0.0 https://review.openstack.org/296752 | 21:24 |
| *** ociuhandu has joined #openstack-manila | 21:25 | |
| *** adrianofr_ has quit IRC | 21:50 | |
| *** a_ta has quit IRC | 21:55 | |
| *** alyson_ has quit IRC | 22:12 | |
| *** martyturner has quit IRC | 22:19 | |
| *** gouthamr has joined #openstack-manila | 22:19 | |
| *** pgbridge has joined #openstack-manila | 22:33 | |
| *** darrenc_ has joined #openstack-manila | 23:05 | |
| *** darrenc has quit IRC | 23:06 | |
| *** darrenc_ is now known as darrenc | 23:07 | |
| *** ganso has quit IRC | 23:09 | |
| *** gouthamr has quit IRC | 23:10 | |
| *** erlon has quit IRC | 23:17 | |
| *** absubram has quit IRC | 23:21 | |
| *** openstack has joined #openstack-manila | 23:25 | |
| *** sgotliv has quit IRC | 23:34 | |
| *** yangyapeng has quit IRC | 23:34 | |
| *** gouthamr has joined #openstack-manila | 23:37 | |
| *** dustins_ has quit IRC | 23:40 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!