Wednesday, 2024-01-10

opendevreviewXuQi proposed openstack/cinder master: Fujitsu Driver: Add parameter fujitsu_use_cli_copy  https://review.opendev.org/c/openstack/cinder/+/90496703:35
opendevreviewRaghavendra Tilay proposed openstack/cinder master: WIP - HPE 3par - PP - Return LUN ids from both arrays  https://review.opendev.org/c/openstack/cinder/+/90158807:09
*** geguileo is now known as Guest1357909:19
opendevreviewXuQi proposed openstack/cinder master: Fujitsu Driver: Add support for revert to snapshot  https://review.opendev.org/c/openstack/cinder/+/90497110:08
*** chuanm9 is now known as chuanm10:50
opendevreviewMerged openstack/cinder-tempest-plugin master: Update stable jobs on master gate  https://review.opendev.org/c/openstack/cinder-tempest-plugin/+/88566910:58
wncsllnhi, morning, cinder o/13:33
wncsllnabout https://bugs.launchpad.net/cinder/+bug/2008043 is there a way to know if qemu is really installed or not?13:33
wncsllnconsidering that function raises an exception even when its present13:33
whoami-rajatclarkb, rosmaita i think we did debug the memory issues during PTG and in one job there were 14 qemu processes running killing mysql eventually13:42
whoami-rajatalso based on the performance data in the gate job, cinder processes doesn't seem to be consuming too much memory13:42
whoami-rajathttps://wiki.openstack.org/wiki/CinderCaracalPTGSummary#Gate_Issues13:43
whoami-rajathttps://wiki.openstack.org/wiki/CinderCaracalPTGSummary#Gate_Issues_213:43
rosmaitawncslln: you may need to parse the stdout/stderr from the call; i think all the errors coming from an installed qemu-img will begin with 'qemu-img:'13:46
wncsllnrosmaita: so a possible solution is to parse and use that infos to improve log message?13:55
whoami-rajatCinder meeting in #openstack-meeting-alt at 1400 UTC14:00
whoami-rajatjungleboyj rosmaita smcginnis tosky whoami-rajat m5z e0ne geguileo eharney jbernard hemna fabiooliveira yuval tobias-urdin adiare happystacker dosaboy hillpd msaravan sp-bmilanov14:00
rosmaitawncslln: i think so14:00
sp-bmilanovo/14:04
*** Guest13579 is now known as geguileo14:08
opendevreviewMerged openstack/os-brick master: NVME-oF: Fix to support new "address" in sysfs  https://review.opendev.org/c/openstack/os-brick/+/89519414:09
opendevreviewMerged openstack/os-brick master: NVMe-oF: Support nvme cli v2  https://review.opendev.org/c/openstack/os-brick/+/89519514:12
opendevreviewMerged openstack/cinder-specs master: Remove placeholder for 2024.1  https://review.opendev.org/c/openstack/cinder-specs/+/90312714:48
opendevreviewGorka Eguileor proposed openstack/os-brick stable/2023.2: NVMe-oF: Fix generation of hostnqn file  https://review.opendev.org/c/openstack/os-brick/+/90522814:50
opendevreviewGorka Eguileor proposed openstack/os-brick stable/2023.2: NVMe-oF: Improve hostnqn creation  https://review.opendev.org/c/openstack/os-brick/+/90522914:50
opendevreviewGorka Eguileor proposed openstack/os-brick stable/2023.2: NVMe-oF: Fix attach when reconnecting  https://review.opendev.org/c/openstack/os-brick/+/90523014:50
opendevreviewGorka Eguileor proposed openstack/os-brick stable/2023.2: NVME-oF: Fix to support new "address" in sysfs  https://review.opendev.org/c/openstack/os-brick/+/90523114:50
opendevreviewGorka Eguileor proposed openstack/os-brick stable/2023.2: NVMe-oF: Support nvme cli v2  https://review.opendev.org/c/openstack/os-brick/+/90523214:50
sp-bmilanovzaitcev:  Hi! trying to follow up on your comment here: https://review.opendev.org/c/openstack/os-brick/+/887576 we discussed this last month ( https://etherpad.opendev.org/p/cinder-caracal-midcycles ) and the consensus is that throwing in __init__() in this context was not an issue15:04
zaitcevsp-bmilanov: let me take a look15:05
sp-bmilanovthanks15:09
opendevreviewBrian Rosmaita proposed openstack/cinder-specs master: Reorganize specs table of contents  https://review.opendev.org/c/openstack/cinder-specs/+/90523915:24
clarkbwhoami-rajat: ya with as many services as openstack runs it often isn't your specific service at fault. But cinder is part of a larger entity and can work within openstack to imrpove things generally too16:15
clarkbhistorically cinder backup was one of the worst services for memory consumption fwiw. I believe that did get addressed though16:15
rosmaitaclarkb: it still uses a lot of memory, but it releases it faster16:16
whoami-rajatclarkb, yes, i wasn't trying to pin point the issue in other projects, was just trying to mention that we did take a look at the memory consumption issue, found few things and even discussed with the nova team about it16:31
whoami-rajatmaybe i didn't phrase it correctly16:31
whoami-rajatwe sure want to improve any issue that arises from the cinder services16:32
clarkbmy main concern is that I believe io load generated by swap usage in our jobs acts as a major part of noisy neighbor problems in the clouds we have for CI resources16:32
clarkbmy hunch is that the more rely on swap the worse our general unexplained job failures and timeouts problem becomes16:33
hemna-so FYI, I know this isn't our issue directly here in cinder, but this has affected my actual cinder deployment.20:26
hemna-https://github.com/openstack4j/openstack4j/issues/22420:26
hemna-we have an internal team that has some automations using that openstack4j client library and this has caused us to have inconsistent volume attachments, because that openstack4j java client exposes a nova <---> cinder api to try to attach a volume.20:28
hemna-they used this, which boils down to an 'os-attach' call.  That simply created an attachment record and marked the volume 'in-use' and an attachment record as attached, with actually doesn't do anything else.20:31
hemna-Then someone called nova to attach the volume, which used the v3 protocol for attachment, (attachment_create, etc.) 20:31
hemna-and because of this code in cinder: https://github.com/openstack/cinder/blob/master/cinder/volume/api.py#L2407-L241420:32
hemna-we ended up with 2 active volume attachments.20:33
hemna-with the first (from the direct os-attach call) completely empty and bogus.20:33
hemna-I think the old pre-v3 api attach calls should be deprecated and eventually removed.   20:34
hemna-volume inconsistencies are some of our biggest headaches.20:34
hemna-volume attachment inconsistencies that is20:34
opendevreviewBryan Neumann proposed openstack/cinder master: Dell EMC: PowerMax - Configurable SRDF snapshots  https://review.opendev.org/c/openstack/cinder/+/89905122:33
*** jph4 is now known as jph23:48

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!