15:00:15 <ricolin> #startmeeting multi_arch
15:00:16 <openstack> Meeting started Tue May 11 15:00:15 2021 UTC and is due to finish in 60 minutes.  The chair is ricolin. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:17 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
15:00:20 <openstack> The meeting name has been set to 'multi_arch'
15:00:29 <ricolin> #topic roll call
15:00:35 <ricolin> o/
15:03:26 <jeremyfreudberg> o/
15:03:50 <ricolin> hey jeremyfreudberg :)
15:04:08 <jeremyfreudberg> hey!
15:04:59 <hrw> [o]
15:06:18 <ricolin> Hi hrw :)
15:06:48 <ricolin> https://etherpad.opendev.org/p/Multi-Arch-agenda
15:08:16 <ricolin> #topic Arm64 job performance tunning
15:08:28 <ricolin> So for Arm64 job
15:09:48 <ricolin> I have run tests on my local environment, and I can't really reproduce the volume backup/restore issue(long running time)
15:10:35 <ricolin> But my set up is different from what's in CI(but still Devstack)
15:11:35 <ricolin> On the other hand, current CI environment seems to reach better performance now from around 3hrs to 2hrs
15:12:11 <hrw> did you checked where it was run?
15:12:22 <hrw> maybe it is difference between linaro and osuosl nodes
15:12:23 <ricolin> on linaro-us
15:12:52 <ricolin> didn't check all recent runs, but I'm sure two of them from linaro-us
15:12:56 <hrw> ok
15:12:59 <jeremyfreudberg> interesting
15:13:06 <ricolin> and it runs 2 hrs this time
15:13:22 <ricolin> I mean the running time is around 2 hrs
15:14:41 <rh-jelabarre> 2 hours deploy time, or 2 hours for a test/validation run?
15:15:00 <ricolin> rh-jelabarre, 2 hours totally
15:16:02 <ricolin> 1200+ secs to run all tests
15:16:33 <ricolin> and 1656 secs for set up devstack
15:17:18 <ricolin> Also, besides performance tunning, we probably need to decide which services/tests we can drop from tempest arm64 job
15:17:44 <hrw> do we fit in 3h time?
15:18:13 <hrw> and how often is this job run?
15:19:04 <ricolin> hrw, everything is still in review patch
15:19:12 <hrw> url?
15:19:13 <ricolin> https://review.opendev.org/c/openstack/devstack/+/708317/100/.zuul.yaml
15:19:21 <ricolin> #link https://review.opendev.org/c/openstack/devstack/+/708317
15:19:55 <hrw> o, num_pcie_ports ;d
15:20:01 <ricolin> I think now is also a good time to discuss what can we moving forward from for job
15:20:24 <ricolin> #link https://zuul.openstack.org/builds?job_name=devstack-platform-arm64+
15:20:28 <ricolin> this is job status
15:21:26 <ricolin> for running time less than 1 hrs, that's me try and error runs:(
15:21:28 <hrw> looks like fit in 3h without problem (at least recently)
15:21:57 <ricolin> hrw, yeah, I will keep put in few more recheck to see how stable is that result
15:23:49 <ricolin> Few follow up ideas for this job, like we can try to use Ceph as backend
15:24:15 <ricolin> Or ideas like we should try to get rid of non-arch related part of tests
15:24:26 <hrw> also rename job to tempest-platform-NAMEOFDISTRO-arm64 (or -aarch64 even)
15:24:47 <ricolin> hrw, that's good idea
15:26:32 <ricolin> One question I have now is which services can be consider non-arch related
15:26:51 <ricolin> Keystone is one
15:27:20 <ricolin> but Neutron sounds arch related to me
15:28:55 <ricolin> I feels like Nova, Neuton, Cinder, are three services which we should keep tempest tests on
15:29:36 <hrw> nova is the most arch related. and it needs cinder and neutron to run. and keystone to auth
15:29:52 <hrw> you may skip horizon for example
15:30:41 <ricolin> hrw yes, but do we need to run tempest tests which design to test Neuton or Cinder specifically
15:31:30 <hrw> how often do those tests run? every change? daily? weekly?
15:31:42 <jeremyfreudberg> i would recommend to run neutron and cinder tests unless there are resource or time constraints
15:31:46 <jeremyfreudberg> it's good to have parity
15:31:53 <hrw> +1
15:32:57 <ricolin> hrw, I plan to propose this job at least as daily/weekly periodic job
15:33:32 <hrw> if daily/weekly then as much as possible
15:33:43 <ricolin> also Think it make sense to run as gate(if possible) or post merge job for Nova
15:33:44 <hrw> as long as it fit in 3h
15:34:17 <hrw> if project accepts then it can be non-voting run-on-change one
15:34:38 <hrw> does not block anything and allows to see when change breaks
15:34:58 <ricolin> I can propose to Nova I think
15:35:22 <ricolin> hrw, do you think the current CI can take on such load?
15:36:42 <hrw> probably
15:36:45 <ricolin> or we can push what kevinz currently proposed in Nova (devstack+nova-tempest)
15:37:15 <ricolin> Let's find out!:)
15:37:31 <hrw> There are several nodes on CI. In worst case you will get queued
15:37:47 <hrw> now we have 82 jobs in check and just 4 in check-arm64
15:37:58 <hrw> and 24 in gate
15:38:14 <hrw> so 106 against 4 ;D
15:39:35 <hrw> https://grafana.opendev.org/d/S1zTp6EGz/nodepool-linaro?orgId=1 shows small use
15:39:41 <ricolin> sounds like good to go!
15:40:36 <ricolin> Okay, guess we can move on:)
15:40:52 <ricolin> #topic arm64 UT
15:41:05 <ricolin> #link https://review.opendev.org/q/topic:%22story%252F2007938%22+(status:open%20OR%20status:merged)
15:41:07 <hrw> UT?
15:41:35 <hrw> ah. unit tests
15:41:35 <ricolin> sorry, I mean Unit test
15:42:08 <ricolin> I didn't yet propose on for Swift, which I should
15:42:18 <ricolin> so will do one today
15:42:43 <hrw> will fix https://review.opendev.org/c/openstack/kolla/+/777062 (kolla) one
15:43:03 <ricolin> hrw, thanks
15:43:50 <ricolin> we might need to change the name in patch to openstack-python3-xena-jobs-arm64
15:44:16 <hrw> ok
15:44:26 <ricolin> That's all I have for Unit test
15:44:38 <ricolin> Kolla and Swift
15:45:22 <ricolin> Let's move on:)
15:45:27 <ricolin> #topic SIG report
15:46:36 <ricolin> jeremyfreudberg, I get the data for how many page vew for our report
15:46:45 <ricolin> the data is for the first 6 days
15:46:47 <jeremyfreudberg> cool
15:46:55 <ricolin> so now should be way more
15:49:16 <ricolin> I think we definitely will have report at least 1 per year
15:49:53 <ricolin> we can also consider doing that as a cycle work, but only if we have enough information to share
15:50:19 <jeremyfreudberg> thanks for getting the data
15:50:38 <jeremyfreudberg> i think additional reports is a good idea
15:50:38 <ricolin> NP
15:50:51 <hrw> ricolin: can it be xena-jobs-aarch64?
15:51:22 <hrw> instead of -arm64
15:51:42 <ricolin> I actually prefer to use aarch64 indeed
15:52:02 <ricolin> forgot why I use arm64 at first place
15:52:06 <hrw> ;D
15:52:45 <ricolin> #topic Libvirt 7 chances for Multi-arch
15:53:39 <hrw> cgroups...
15:53:49 <hrw> 17:53 < openstackgerrit> Marcin Juszkiewicz proposed openstack/kolla master: add openstack-python3-xena-jobs-aarch64 job  https://review.opendev.org/c/openstack/kolla/+/777062
15:54:58 <ricolin> I didn't put much time on this yet
15:56:06 <ricolin> will spend more time in following weeks
15:56:30 <ricolin> hrw, I guess cgroups is where I should starts on?:D
15:58:25 <hrw> yeah
15:58:51 <ricolin> We're 2 mins left
15:59:34 <hrw> we can postpone that part to next week
15:59:57 <ricolin> okay
16:00:12 <hrw> and start from it ;D
16:00:41 <ricolin> hrw, BTW, I like to mention that jinyuanliu reach out to me today and appears he is plan to introduce multi-arch support for openstack helm images
16:01:11 <jeremyfreudberg> nice
16:01:18 <hrw> good to know that someone still uses helm
16:01:22 <ricolin> nothing in progress yet
16:01:55 <ricolin> haha
16:02:16 <ricolin> will see if we can have something in progress from there:)
16:03:19 <ricolin> Okay, I think we can call this meeting off
16:03:28 <ricolin> Thanks all for join:)
16:03:34 <ricolin> #endmeeting