18:18:18 <mmichelson> #startmeeting ovn_community_development_meeting
18:18:19 <openstack> Meeting started Thu Feb 18 18:18:18 2021 UTC and is due to finish in 60 minutes.  The chair is mmichelson. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:18:20 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
18:18:22 <openstack> The meeting name has been set to 'ovn_community_development_meeting'
18:18:30 <mmichelson> Sorry for the late start. I got caught up in some code reviews :)
18:18:49 <blp> I have a report.
18:19:15 <blp> I posted v10 of my ddlog patches. I think they should be ready to go, when the project is ready for them.
18:19:26 <blp> That's my basic report.
18:19:34 <mmichelson> blp, OK thanks. 21.03 is scheduled to be branched tomorrow
18:19:48 <mmichelson> So we could probably put the DDLog patches in after the branch is created.
18:20:01 <blp> Does anyone want to object to that?
18:20:19 <blp> I want to make sure that there's a basic consensus on it.
18:20:55 <imaximets> blp: FYI, ovsrobot was not able to build the patch set: make[1]: *** No rule to make target 'northd/OVN_Northbound.dl', needed by 'all-am'.
18:21:14 <imaximets> https://github.com/ovsrobot/ovn/runs/1923661603?check_suite_focus=true#step:10:3110
18:21:16 <blp> imaximets: OK, I'll figure out what's going on there.
18:24:18 <mmichelson> Who wants to go next?
18:24:58 <imaximets> I have a small update
18:25:33 <imaximets> some time ago we removed inactivity probes for raft connections.  And that was a mistake. :)
18:26:12 <imaximets> since heartbeats are kind of unidirectional and leader doesn't do anything if it doesn't receive replies dead connections remains open
18:26:55 <imaximets> leader will continue to send messages to dead connections growing the jsonrpc backlog
18:27:20 <imaximets> I'm going to reintroduce inactivity probes to solve that issue.
18:27:48 <imaximets> time interval will depend on election timeout to avoid disconnections on highly loaded clusters.
18:28:31 <imaximets> One more related issue is that cluster/status reports "cluster member" for servers that are retrying election.
18:28:51 <imaximets> i.e. first election attempt timed out.
18:29:28 <imaximets> is_connected filed in _Server database is correctly updated for this case saying that storage is disconnected
18:30:00 <imaximets> but cluster/status still reports "cluster member".  I'll send a patch to print more accurate status.
18:30:28 <imaximets> And yes, OVS 2.15 officially released. :)
18:30:34 <imaximets> That's it from me.
18:30:45 <fnordahl> \o/
18:30:52 <fnordahl> I'll go next if that's ok
18:31:46 <fnordahl> first some stable maintenance stuff, I have proposed a number of patches to branch-20.06 and branch-20.03, and I wonder if the form of the proposed backports are acceptable?
18:33:36 <mmichelson> fnordahl, I noticed the patch series but haven't had a close look at the patches. Are they all just bug fixes?
18:34:41 <fnordahl> the end product of the series is to fix the predictable resolution for conflicting flow actions bug
18:35:27 <dceara> fnordahl: As I mentioned on the mailing list, one of the worrying things is that we're still chasing some bugs that were introduced by some of those patches.  I'm a bit worried about the stability of 20.06/20.03 if the backports are accepted.
18:36:53 <fnordahl> I definitively see your point on that, however, 20.03 and 20.06 wihout those patches are not really consumable when OVN is part of a bigger layered system such as an OpenStack deployment with Kubernetes on top
18:37:12 <numans> I'd say if its possible to move to 20.09/20.12, please do so :)
18:37:41 <fnordahl> yeah, that would be easy, wouldn't it ;)
18:38:19 <fnordahl> the thing there is that versions > 20.03 make changes to how CT flows are used
18:39:29 <fnordahl> and that can have ill effects for deployments that make use of hardware offload on kernels without support for CT offload
18:40:08 <dceara> fnordahl: There was a recent patch that fixed the fact that we were matching on ct_state unconditionally, wouldn't that be enough?
18:40:41 <fnordahl> that would definitively help, will check.
18:40:45 <dceara> fnordahl: This one: https://github.com/ovn-org/ovn/commit/5336b5cb342b8f81115299540f3268f734a6d009
18:41:14 <fnordahl> but anyway, we kind of have a commitment to keep 20.03 alive for several years still, and it would make it easier for us to consume upstream patches in form of stable point release updates if we're ajour with what is in upstream
18:42:42 <fnordahl> and since this fix is missing in 03, 06 and 09 branches it kind of feels like a special case. But, let's take this to ML. :)
18:43:06 <numans> May be you can bring this up in the discussion. If there are no objections from others, I'm fine with it.
18:44:06 <numans> I can go real quick once fnordahl is done.
18:44:17 <fnordahl> go for it
18:44:25 <numans> thanks.
18:45:02 <numans> I've bug fix patch for review. Request to to take a look at it - https://patchwork.ozlabs.org/project/ovn/patch/20210212072140.341358-1-numans@ovn.org/
18:45:41 <numans> This past week I've been busy debugging the assertions we are seeing in ofctrl.
18:45:50 <numans> Found the cause for one and submitted the fix.
18:46:02 <numans> thanks to dceara for that.
18:46:05 <numans> Still trying to figure out the other one.
18:46:22 <numans> That's it from me.
18:48:37 <dceara> I have a related update as well: while investigating with numans the ofctrl crashes we spotted 3 more issues so I sent patches to fix them.  One of them is RFC because I'm not sure about the best way to fix the problem.
18:49:35 <dceara> That's it from my side, thanks!
18:50:21 <numans> dceara, thanks for fixing them.
18:50:31 <numans> I'll take a look tomorrow.
18:50:37 <dceara> numans: Thanks!
18:51:26 <mmichelson> Anybody else? Not sure who else there is to go.
18:51:49 <mmichelson> I guess I've been quiet, but that's because I've been working on fixing some niche issues that OpenStack found (patches will probably hit the mailing list next weeK)
18:52:09 <mmichelson> And today I'm hitting the reviews hard since as I mentioned earlier we plan to branch 21.03 tomorrow.
18:52:28 <mmichelson> And that's all from me.
18:54:07 <blp> Does anyone know how I can get the robot to try another variation? Do I have to repost the whole series?
18:54:10 <blp> I think I have a fix.
18:54:30 <imaximets> blp, you could just commit your patches to your github
18:54:46 <imaximets> blp, GitHub actions will be triggered.
18:55:47 <blp> imaximets: Oh, sure enough. https://github.com/blp/ovs-reviews/actions/runs/579105628
18:56:29 <imaximets> blp, since it worked, I think, you need to re-post the series.
18:56:47 <mmichelson> imaximets, blp I'm not sure the github actions and the bot do the same thing
18:56:47 <blp> OK, I'll do that then.
18:57:05 <mmichelson> The github actions make use of the scripts in the .ci dir
18:57:18 <mmichelson> The bot is, uh, its own thing.
18:57:29 <imaximets> mmichelson, blp: robot does a bit more, but last time both robot and ci failed.
18:57:36 <mmichelson> ah, ok
18:57:42 <mmichelson> https://github.com/blp/ovs-reviews/actions
18:57:45 <mmichelson> That shows some failed runs
18:58:00 <mmichelson> Ah but most recent succeeded
18:58:06 <imaximets> I guess, if CI worked, robot should be OK too.
18:58:20 <imaximets> unless there are some strange issues.
18:59:00 * dceara has to run, bye everyone
18:59:33 <blp> We'll find out soon.
19:01:10 <mmichelson> All right, I guess that's all for today.
19:01:20 <mmichelson> Bye everyone.
19:01:28 <mmichelson> #endmeeting