- Thursday, May 13, 2021

release-team@conference.openafs.org

Thursday, May 13, 2021< ^ >

kaduk@jabber.openafs.org/barnowl has set the subject to: openafs release team

Room Configuration

Room Occupants

GMT+0
[00:05:08] mbarbosa joins the room
[00:05:32] mbarbosa leaves the room
[01:07:08] Cheyenne leaves the room
[01:20:10] mvita2 joins the room
[11:57:11] mbarbosa joins the room
[13:49:12] mbarbosa leaves the room
[13:52:01] mbarbosa joins the room
[13:58:44] Cheyenne joins the room
[14:49:44] mvita2 leaves the room
[15:11:43] meffie joins the room
[15:22:43] mvita2 joins the room
[15:40:53] <kaduk@jabber.openafs.org/barnowl> Hmm, did I break gerrit by requesting too many pages at once?  Now I
have spinny browser tabs...
[15:55:02] mbarbosa leaves the room
[15:57:52] mbarbosa joins the room
[15:58:43] yadayada joins the room
[16:00:20] <mvita2> spinny browser tabs could be considered a feature in some circles
[16:00:42] <meffie> greetings.
[16:00:49] <yadayada> hi all
[16:01:00] <Cheyenne> Hello
[16:01:27] <kaduk@jabber.openafs.org/barnowl> greetings
[16:06:25] <meffie> fyi, i'll need someone to release the www.workshop <http://www.workshop> volume soon
[16:07:45] <meffie> (insert self deprecating remark about lack of html/css/jekyll skills here)
[16:13:13] <kaduk@jabber.openafs.org/barnowl> We didn't hear anything from Stephan about expected attendance, right?
[16:13:25] <meffie> i have not.
[16:13:38] <kaduk@jabber.openafs.org/barnowl> Anyway, I was a bit distracted earlier in the week but did manage to
get some review in (notably, the CTD stats that Mike was interested
in)
[16:13:59] <meffie> thank you! and the bozo bssize thing
[16:14:39] <kaduk@jabber.openafs.org/barnowl> yes ... finally
[16:14:49] <mvita2> thank you
[16:15:40] <kaduk@jabber.openafs.org/barnowl> Also the unoquorum stack, which is almost ready
[16:15:54] <Cheyenne> Yes -- thank you.  I'm reworking some of the earlier commits in the stack to address the feedback there and I will replace the new RPC commits with the size increase
[16:16:03] <kaduk@jabber.openafs.org/barnowl> And I've got a few other things open to look at, that saw activity
recently.
[16:19:42] <Cheyenne> Oh.. I wonder if Stephan is out for holiday?
[16:20:00] <mvita2> ah, good point
[16:20:27] <kaduk@jabber.openafs.org/barnowl> could be, I did have a few European colleagues note that today was a
holiday for them
[16:20:42] <meffie> ah, yes, it is a holiday in germany
[16:20:43] <kaduk@jabber.openafs.org/barnowl> it looks like a few things did get merged to 1.8.x after last week's
meeting
[16:21:07] <Cheyenne> Anyway .. Linux-5.13-rc1+ -- master and 1.8.x build cleanly
[16:22:19] <kaduk@jabber.openafs.org/barnowl> Looks like currently open against 1.8.x are ih_PkgDefaults,
cold_shutdown, bigsur, "warn when starting without keys", and the pre1
version/NEWS changes
[16:23:06] <kaduk@jabber.openafs.org/barnowl> I forget which ones we already talked about :(
[16:23:27] <kaduk@jabber.openafs.org/barnowl> but IIUC ih_PkgDefaults is not critical for 1.8.8, nor is
cold_shutdown
[16:23:28] <mvita2> all of the above, I think
[16:23:50] wiesand joins the room
[16:24:07] <mvita2> yay!
[16:25:07] <wiesand> Hello. Sorry I'm very late.
[16:25:15] <kaduk@jabber.openafs.org/barnowl> I think we should be trying to get 1.8.8 out with the support for
newer linux as soon as we can, though, so maybe we want to defer some
of this stuff to the next release.
[16:25:28] <mvita2> we assumed you were observing the holiday
[16:26:18] <wiesand> While it is a public holiday here today, my absence was simply a mistake. Sorry again.
[16:26:48] <kaduk@jabber.openafs.org/barnowl> I think I managed to blather on and kill some time
[16:27:23] <wiesand> I think Ben summarized the state of affairs rather nicely.
[16:27:58] <kaduk@jabber.openafs.org/barnowl> I did not have a summary for whether "warn when starting without keys"
should go in 1.8.8 or not, though
[16:29:07] <kaduk@jabber.openafs.org/barnowl> It looks like only Mark has +1'd the bigsur changes for 1.8.x; do you
want an additional reviewer for those?
[16:30:05] <wiesand> "warn when starting without keys" has some interesting path conflicts with other changes on the wish list and may have to wait a bit - nothing serious though
[16:31:40] <wiesand> Mark indeed reviewed the bigsur stack after last meeting, and very thoroughly IMO. While more review would be nice, I'm inclined to accept the stack now.
[16:32:06] <kaduk@jabber.openafs.org/barnowl> ok
[16:32:31] <wiesand> That is, unless anyone would like to object…
[16:33:27] <kaduk@jabber.openafs.org/barnowl> There are a lot of moving parts in "delegate sock_* calls to bkg
daemons", but the churn is constrained to macos versions that would
otherwise just be broken, so I think merging now is ok
[16:34:19] <mvita2> I did NOT examine that stuff closely, because I don't know it.
[16:34:30] <kaduk@jabber.openafs.org/barnowl> (or do we also set it for darwin_190 that might work otherwise?)
[16:34:31] <mvita2> I was aiming to verify some stuff in my tests
[16:34:52] <yadayada> We are also trying BigSur patches for sometime now and didn;t faced any issues on delegate dock call to background daemon, so changes looks fine
[16:34:56] <kaduk@jabber.openafs.org/barnowl> Yeah, it's also the sort of thing where if it doesn't work it should
be really obvious right away
[16:36:31] <kaduk@jabber.openafs.org/barnowl> So then are we proposing to merge bigsur and cut pre1?
[16:37:24] <wiesand> Looks like it's either that or a 1.8.8 w/o bigsur.
[16:38:35] <kaduk@jabber.openafs.org/barnowl> My vote is to include bigsur :)
[16:38:41] <mvita2> +1
[16:38:47] <yadayada> I also vote to include bigsur
[16:39:11] <meffie> +1
[16:39:20] <Cheyenne> +1
[16:40:15] <wiesand> Ok, let's do that then.
[16:40:47] <kaduk@jabber.openafs.org/barnowl> Sounds good.
[16:41:08] <kaduk@jabber.openafs.org/barnowl> Other 1.8 topics, or should we move on to ... "other master/1.9
topics"? ;)
[16:41:22] <yadayada> One generic question
[16:42:01] <wiesand> yes?
[16:42:10] <yadayada> recently we were seeing on certain deployment 1.8 code performance is less then 1.6. Did anyone heard of something similar ?
[16:42:24] <mvita2> no
[16:42:30] <mvita2> can you be more specific?
[16:43:08] <yadayada> mainly user is trying to do lots of "git fetch" in parallel and that's where they saw performance dip
[16:43:20] <mvita2> ahh
[16:43:20] <kaduk@jabber.openafs.org/barnowl> do you know if client side or server side was the changing factor?
[16:43:33] <yadayada> client side
[16:43:36] <mvita2> was this on Linux?
[16:43:59] <yadayada> yes Linux
[16:44:28] <mvita2> okay, it's probably changes to ensure coherence between Linux dentry cache and AFS cache
[16:45:18] <yadayada> I still need to dig into it, but thought of asking if someone has seen something similar.  Could be dentry cache + AFS cache. Will get more info
[16:46:11] <mvita2> I haven't noticed a performance difference myself, even though I use git operations frequently as a stress test for AFS
[16:46:26] <yadayada> One thing we saw recently was on 1.6 code base we got lots of SIGBUS error
[16:46:30] <mvita2> but then, I haven't done A:B with 1.6.x vs 1.8.x lately
[16:46:53] <mvita2> sigbus sounds familiar, checking my notes
[16:47:00] <yadayada> but when we use 1.8 it went away. I thought of 14171
[16:48:38] <yadayada> On 1.8 we use kernel funnction to wait compared to 1.6 where we manualy add task to wait list and call schedule. So 1.8 did helped for SIGBUS, but I think 14171 issue can pop up again. Anything on 14171 , was any other attenpt made to fix SIGBUS
[16:49:17] <kaduk@jabber.openafs.org/barnowl> 14171 says "abandoned in favor of 14189" which is still open and has a
-1 from Andrew
[16:49:52] <wiesand> right
[16:50:29] <mvita2> https://gerrit.openafs.org/#/q/topic:linux_signals
[16:50:51] <mvita2> a difficult patchset - I haven't had time to get back to it for a while
[16:52:43] <yadayada> I think 1.8 did help to avoid SIGBUS but it can pop up again. As you say no simple way to fix.
[16:55:09] <mvita2> yadayada:   it's possible that https://gerrit.openafs.org/#/c/11794/ would help with performance issues - it is in the area of code I suspect
[16:55:41] <mvita2> and has not been merged to master yet, so not in 1.8.x either of course
[16:56:09] <yadayada> Thanks will check on this patch and revert
[16:56:32] <mvita2> that work was done for a site that did heavy parallel workloads in AFS
[16:57:53] <kaduk@jabber.openafs.org/barnowl> That one is one that will need careful review, I think...
[16:58:04] <mvita2> you might also try to characterize with BPF or other tools, especially in OpenAFS validation/invalidation/revalidation of dentry
[16:58:17] <mvita2> systemtap, e.g.
[16:58:29] <mvita2> I know you know how to do that!
[16:58:36] <meffie> 11794 has been in use for some time, as a patch
[16:59:56] <yadayada> right, was planning to that. Thanks for all pointers, had a glance at patch and it looks it will definatley help. Will try it and will see if I can provide some feedback on it
[17:00:10] <mvita2> thank you!
[17:02:28] <meffie> thanks yadayada
[17:04:04] <mvita2> yes, meffie is correct, that patch has been running in production  for some time on some large cells
[17:04:34] <wiesand> Ok, looking forward to Yadav's findings.
[17:04:41] <kaduk@jabber.openafs.org/barnowl> Yes, me too
[17:04:53] <kaduk@jabber.openafs.org/barnowl> Were there other topics people wanted to cover today?
[17:04:58] <wiesand> On to 1.9/master, or is it too late for that already?
[17:05:13] <mvita2> I'm still good for time
[17:05:32] <kaduk@jabber.openafs.org/barnowl> We talked about a couple 1.9/master things already ... I don't think I
had anything else I wanted to cover
[17:06:29] <wiesand> I have no further topics for today. My agenda is  clearly "get 1.8.8pre1 done, now that we seem to have all the pieces in place"
[17:06:45] <mvita2> +1
[17:06:48] <kaduk@jabber.openafs.org/barnowl> +1
[17:07:15] <yadayada> +1
[17:08:51] <wiesand> Which translates into "merge bigsur + maybe some other stuff already reviewed, update NEWS". Will start with that later today.
[17:09:58] <wiesand> More FBSD and bzo stuff may still make pre2, or have to wait for 1.8.9
[17:10:25] <wiesand> Anything else to discuss today?
[17:10:49] <Cheyenne> I mentioned earlier everything is good with linux-5.13-rc1+
[17:11:05] <wiesand> Very good, thanks!
[17:11:12] <meffie> wiesand: i will need someone to do a vos release of www.workshop <http://www.workshop>, hopefully tomorrow
[17:11:43] <wiesand> I should be standing by.
[17:11:59] <meffie> thanks!
[17:12:41] <wiesand> Let's adjourn?
[17:12:44] <kaduk@jabber.openafs.org/barnowl> +1
[17:12:59] <yadayada> +1
[17:13:01] <mvita2> thank you everyone
[17:13:09] <yadayada> thanks all
[17:13:10] <wiesand> Thanks a lot everybody!
[17:13:30] wiesand leaves the room
[17:13:48] <meffie> thanks all!
[17:14:09] <kaduk@jabber.openafs.org/barnowl> thanks everyone!
[17:14:53] meffie leaves the room
[18:23:21] yadayada leaves the room
[21:53:14] mbarbosa leaves the room
[21:58:28] Cheyenne leaves the room