| opendevreview | Merged openstack/cinder stable/xena: Reject unsafe delete attachment calls https://review.opendev.org/c/openstack/cinder/+/882839 | 02:23 |
|---|---|---|
| opendevreview | Gorka Eguileor proposed openstack/cinder master: Doc: Improve service token https://review.opendev.org/c/openstack/cinder/+/883360 | 08:17 |
| geguileo | rosmaita: done, thanks | 08:17 |
| yuval | geguileo: I need to install the DLM on my deployment? I see my driver do use coordination decorator while creating snapshots | 08:23 |
| geguileo | yuval: yes yo need to install and configure the DLM | 08:54 |
| geguileo | s/yo/you | 08:54 |
| geguileo | yuval: what installation tool are you using? | 08:54 |
| yuval | I am running using kolla-ansible | 08:54 |
| yuval | is there a suggested DLM? | 08:55 |
| yuval | also all the clients using this driver need to install a DLM? | 08:55 |
| geguileo | yuval: if customer wants to do A/A they need to install the DLM and configure cinder to use it | 08:55 |
| yuval | I see | 08:56 |
| geguileo | yuval: if they are running A/P then the default in Cinder is to use file locks for those decorators, so no need to configure it | 08:56 |
| geguileo | yuval: there seems to exist some support for tooz and DLM in kolla-ansible, at least by the looks of this bug report: https://bugs.launchpad.net/kolla-ansible/+bug/1840070 | 08:58 |
| yuval | yea I also see this: https://opendev.org/openstack/kolla-ansible/commit/03b4c706fae54a381b78c0f516499bfac986fb26 | 09:06 |
| yuval | geguileo: I know that if I configure the cluster=yuval_cluster in cinder.conf I will get an a/a cluster. how do I set a a/p cluster? | 09:22 |
| geguileo | yuval: for the a/a cluster you do the cluster config and the DLM | 09:23 |
| geguileo | yuval: for a/p you just do the cluster config | 09:23 |
| geguileo | and for a/p non-cluster you don't do the cluster config | 09:23 |
| yuval | now I just configured the cluster in cinder.conf. I am running operations I see it round robin between the cinder's | 09:24 |
| yuval | so its running as a/a right? | 09:24 |
| yuval | is it possible I see it roundrobin because the backend name is the same in the 2 nodes? | 09:25 |
| yuval | and its not related to cluster? I do see in the logs - "moving all resources to cluster yuval_cluster@<backend name> | 09:26 |
| geguileo | yuval: it's important to understand each part of the puzzle on the A/A madness | 09:34 |
| yuval | :)))) | 09:34 |
| yuval | lol | 09:34 |
| geguileo | yuval: the round robin is done because they are using the same message queue in RabbitMQ, and they use the same one because you have configured the cluster | 09:35 |
| geguileo | so that part is probably working correctly | 09:35 |
| yuval | got it | 09:35 |
| geguileo | now, if your DLM is configured to use file locks | 09:35 |
| yuval | the coordination section is empty | 09:35 |
| geguileo | and you have your 2 cinder-volume services in different hosts, then you are not actually doing the locking you want | 09:35 |
| yuval | so I think it is files locks | 09:35 |
| geguileo | yup | 09:36 |
| yuval | I see | 09:36 |
| opendevreview | Sofia Enriquez proposed openstack/cinder master: Automate generation of backups api-ref samples https://review.opendev.org/c/openstack/cinder/+/861491 | 09:36 |
| geguileo | So this is likely to have race conditions in the critical sections that you define in your code using the coordination lock decorators | 09:36 |
| yuval | yes I see | 09:36 |
| yuval | ok, regarding the a/p and a/a. if the coordination is empty but exists that mean it is using the default DLM and have the cluster enabled - so its running a/a? | 09:38 |
| geguileo | yuval: it is running a/a in an unsafe manner | 09:39 |
| yuval | yes | 09:39 |
| yuval | geguileo: so if I would like to run it in an a/p manner? I would need to disable the DLM? how can I do that? | 09:40 |
| geguileo | the cinder core code also expects the DLM to be properly configured when running a/a, so there will be operations that cinder will allow (and shouldn't) if incorrectly configured | 09:40 |
| geguileo | yuval: you don't need to disable anything, you just need to run 1 single cinder-volume service at a time | 09:40 |
| yuval | ah, I see. so the active-passive mode. is not like there is only 1 writer node which take care of all the operations and one passive which waits for the first 1 to die. its just mean I have 2 cinder nodes, some kind of a heartbeat on the active node and if its dead, I will init the passive node. is that correct? | 09:43 |
| geguileo | yuval: yes that is correct, we don't have the feature to do leader election | 09:48 |
| geguileo | (in Cinder I mean) | 09:49 |
| yuval | I see now what you mean. if the passive node was not part of the cluster (and also the active one) the data of what happen would be saved only on the first cinder node, not in the cluster database. once I set it to be in the cluster mode, all the nodes know which data to look for. | 09:52 |
| yuval | I would appreciate another review on this one liner: https://review.opendev.org/c/openstack/cinder/+/883797 | 10:14 |
| yuval | if I can help someone else with a review I am gladly open for requests | 10:14 |
| opendevreview | Raghavendra Tilay proposed openstack/cinder stable/2023.1: HPE 3PAR: Few issues with new WSAPI (of 2023) https://review.opendev.org/c/openstack/cinder/+/882781 | 10:15 |
| raghavendrat | hi, for below patch i received two +2 and +W on May 17. | 12:27 |
| raghavendrat | https://review.opendev.org/c/openstack/cinder/+/882266/ | 12:27 |
| raghavendrat | However, zuul didn't pass in one week. Observed failure in below sub-tests of zuul: | 12:27 |
| raghavendrat | 4 times by cinder-plugin-ceph-tempest, 2 times by grenade, and | 12:27 |
| raghavendrat | 1 time each by cinder-grenade-mn-sub-volbak, tempest-integrated-storage, cinder-tempest-plugin-lvm-lio-barbican, tempest-integrated-storage-ubuntu-focal, tempest-slow-py3 | 12:27 |
| raghavendrat | Not sure if anybody else is facing this issue. | 12:28 |
| raghavendrat | It would be great if someone can provide pointers to resolve this. Thanks. | 12:28 |
| raghavendrat | ------- | 12:28 |
| raghavendrat | Please note: From IRC logs of past one week, i observed few patches got merged into master(Bobcat). | 12:28 |
| raghavendrat | However, those were trivial patches (having change in .rst/.txt files). | 12:28 |
| raghavendrat | Didn't observe a patch having atleast 20+ lines of python code getting merged. | 12:28 |
| opendevreview | Eric Harney proposed openstack/cinder master: zuul: Extend timeout for cinder-plugin-ceph-tempest https://review.opendev.org/c/openstack/cinder/+/884194 | 12:42 |
| eharney | this patch should help at least one of raghavendrat's zuul failures ^ | 12:43 |
| raghavendrat | thank you eharney: for quick help | 12:45 |
| opendevreview | Merged openstack/cinder master: Doc: Improve service token https://review.opendev.org/c/openstack/cinder/+/883360 | 13:30 |
| opendevreview | Nahim Alves de Souza proposed openstack/cinder master: NetApp ONTAP: Fix create FlexVol pool replica https://review.opendev.org/c/openstack/cinder/+/798384 | 13:45 |
| whoami-rajat | Cinder meeting in #openstack-meeting-alt at 1400 UTC | 13:59 |
| whoami-rajat | jungleboyj rosmaita smcginnis tosky whoami-rajat m5z e0ne geguileo eharney walshh_ jbernard sfernand enriquetaso hemna fabiooliveira yuval tobias-urdin adiare happystacker dosaboy hillpd | 13:59 |
| enriquetaso | Hi I had a personal issue and I havent finished the bug report. I will send it later today but I wont be able to run the cinder bug meeting after the normal meeting | 14:01 |
| hillpd | hello | 14:01 |
| rosmaita | enriquetaso: that's ok, let's have a bug-free week in cinder! | 14:05 |
| enriquetaso | thanks | 14:05 |
| opendevreview | Eric Harney proposed openstack/cinder-tempest-plugin master: zuul: Extend timeout for cinder-tempest-plugin-lvm-multiattach https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/884204 | 14:09 |
| opendevreview | Merged openstack/cinder master: Allow lightos driver to run as active-active https://review.opendev.org/c/openstack/cinder/+/883797 | 14:22 |
| opendevreview | Eric Harney proposed openstack/cinder-tempest-plugin master: zuul: Exclude unneeded tests from multiattach job https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/884209 | 14:51 |
| eharney | tosky: ^ | 14:51 |
| tosky | uhhh | 14:56 |
| eharney | yeah | 14:57 |
| opendevreview | Eric Harney proposed openstack/cinder master: Tests: Provide filter arg to VolumeAttachmentNotFound() https://review.opendev.org/c/openstack/cinder/+/884253 | 16:43 |
| opendevreview | Eric Harney proposed openstack/cinder-tempest-plugin master: zuul: Extend timeout for cbak-ceph job https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/884258 | 17:43 |
| opendevreview | Gorka Eguileor proposed openstack/os-brick master: SCSI: Cleanup leftover devices and validate result https://review.opendev.org/c/openstack/os-brick/+/882841 | 18:40 |
| opendevreview | Ashley Rodriguez proposed openstack/devstack-plugin-ceph master: [WIP][DNM] Remote Ceph with cephadm https://review.opendev.org/c/openstack/devstack-plugin-ceph/+/876747 | 20:56 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!