Home
release-team@conference.openafs.org
Friday, February 8, 2019< ^ >
Room Configuration
Room Occupants

GMT+0
[12:49:23] meffie leaves the room
[13:44:48] meffie joins the room
[13:45:31] <kaduk@jabber.openafs.org/barnowl> Hey Mark, if you've got (time and) a solaris machine handy, you might
push the 'verified' button for the solaris-nonembed-vnode and/or
solris-vnop-warnings stacks, since they only touch src/afs/SOLARIS and
buildbot is apparently being sad
[13:52:32] <kaduk@jabber.openafs.org/barnowl> Ah, and meffie is here now who I can ask why buildbot hasn't been
reporting feedback on changes for (a day?)
[13:52:39] <meffie> hello
[13:53:01] <kaduk@jabber.openafs.org/barnowl> Hi Mike :)
[13:53:08] <meffie> wha? oh
[13:55:29] <mvita> present
[13:55:55] <mvita> I do have solaris machines but have not built the vnode stuff
[13:55:57] <mvita> I will
[13:57:03] <kaduk@jabber.openafs.org/barnowl> Thank you!
[13:57:04] <meffie> i restarted the master.
[13:59:03] wiesand joins the room
[13:59:03] <mvita> <fires up favorite solaris 11.3 vm>
[14:00:06] <wiesand> Good morning
[14:00:13] <meffie> (i need to fine a way to monitor the connection to gerrit.)
[14:00:39] <meffie> good evening wiesand
[14:01:06] <wiesand> sorry for having been lame again last week
[14:01:22] <kaduk@jabber.openafs.org/barnowl> (greetings)
[14:01:30] <wiesand> but expect preliminary 1.8.3pre1 tarballs to smoke test later today
[14:01:45] <kaduk@jabber.openafs.org/barnowl> Oh, the master wasn't running?
Silly me, I should have checked for that and could even have done
something about it myself :(
[14:03:01] <wiesand> is 13479 something we should still add to 1.8.3pre1?
[14:03:17] <wiesand> (one more dkms fix)
[14:04:07] <meffie> it's a nice to have and seems low risk
[14:04:10] <kaduk@jabber.openafs.org/barnowl> I defer to the people who actually use rpm packages
[14:04:41] <meffie> and would hopefully help cheyenne
[14:04:45] <wiesand> Even I don't use dkms on EL…
[14:05:23] <wiesand> ok, I'll include it
[14:05:27] <meffie> thanks!!
[14:06:06] <wiesand> any other late additions to the 1.8.3 wish list?
[14:08:00] <wiesand> I'll take the silence as a "no"
[14:08:05] <meffie> maybe the solaris vnode fixes
[14:08:14] <meffie> so we can run on solaris 11
[14:08:14] <kaduk@jabber.openafs.org/barnowl> Did 13454 need to get pulled up (or did it already)?
[14:08:40] <kaduk@jabber.openafs.org/barnowl> I don't think it's advisable to hold 1.8.3 for the solaris vnode fixes
[14:08:52] <meffie> those are not merged on master yet, but they seem pretty straightforward
[14:09:10] <kaduk@jabber.openafs.org/barnowl> but if they're merged on master it should be okay to include them in the
pre if the timing works out
[14:09:21] <meffie> sounds good.
[14:09:43] <wiesand> good thing that I'm so slow these days ;-)
[14:10:40] <meffie> i think those would be good candidates for 1.6.x as well, if we want to support solaris
[14:11:02] <meffie> which i think we do. :)
[14:11:14] <kaduk@jabber.openafs.org/barnowl> I guess 13396 also would apply (fixes a memory leak)
[14:11:38] <kaduk@jabber.openafs.org/barnowl> a parent or three of that as well
[14:12:08] <wiesand> I don't think 13454 has been pulled up yet
[14:12:58] <kaduk@jabber.openafs.org/barnowl> I would think about pulling up 12879 as well (vos zap -force cleanup)
[14:15:29] <wiesand> ok, going through the logs searching for changes we discussed for inclusion but have forgotten about is the next thing on my list anyway, so I'll look at those too
[14:15:51] <kaduk@jabber.openafs.org/barnowl> thanks
[14:15:58] <meffie> 13454 is needed after 12753 (convert xstat and friends to pthreads)
[14:16:28] <kaduk@jabber.openafs.org/barnowl> Ah, which we did not take (yet?) on 1.8
[14:16:44] <meffie> yes
[14:16:48] <kaduk@jabber.openafs.org/barnowl> I couldn't remember if the EINTR issue was noted during review for
master or for 1.8
[14:16:57] <meffie> it was master.
[14:17:13] <kaduk@jabber.openafs.org/barnowl> sorry for confusing wiesand :(
[14:17:37] <meffie> sorry for the bug. :(
[14:17:43] <wiesand> Don't worry, I'm so used to that :-)
[14:19:59] <wiesand> I don't think I have more on the stable series for now
[14:20:10] <meffie> ok, thanks wiesand.
[14:20:37] <kaduk@jabber.openafs.org/barnowl> Okay, time to see if I can remember everything I wanted to talk about...
[14:21:07] <kaduk@jabber.openafs.org/barnowl> For "auto-tune fileserver for memory target" (11629), Mike had some good
comments
[14:21:41] <kaduk@jabber.openafs.org/barnowl> The whole scheme is probably half-baked (and is definitely untested),
but I don't know how much we want to fiddle with things before making it
available for people to play with and give us feedback on.
[14:22:02] <kaduk@jabber.openafs.org/barnowl> I'm also happy to hand over change control if Mike thinks he has time to
fix the issues he spotted, directly.
[14:22:49] <meffie> ok, i can update it. i'd like to add some FileLog messages too, to show the values picked.
[14:22:58] <kaduk@jabber.openafs.org/barnowl> A capital idea!
[14:23:01] <meffie> :)
[14:23:31] <meffie> (file logging is deferred, we dont have the log open until later)
[14:23:45] <kaduk@jabber.openafs.org/barnowl> It looks like 13398 is also waiting on an update, FYI
[14:24:18] <kaduk@jabber.openafs.org/barnowl> Er, wait, I am misreading the history, aren't I.
[14:24:42] <kaduk@jabber.openafs.org/barnowl> (not worth the time to untangle during the meeting, I bet)
[14:25:34] <kaduk@jabber.openafs.org/barnowl> Andrew uploaded a new rxgk_token.c rev (10567) so I need to look at that
again and kickstart rxgk-phase1
[14:25:37] <meffie> i have updates for that stack, to be pushed. (was just working on it this morning, not in time to push for the meeting)
[14:25:47] <kaduk@jabber.openafs.org/barnowl> I know the  feeling :)
[14:25:51] <meffie> that stack == bosserver-no-client-dirs
[14:26:28] <kaduk@jabber.openafs.org/barnowl> Several of you saw that I made some progress with gcc8 and clang7 and
--enable-checking
[14:26:41] <meffie> yay! \o/
[14:27:01] <kaduk@jabber.openafs.org/barnowl> But 13477 was my sort of "I'm too lazy" band-aid for a couple things
that I didn't want to deal with at the time.
[14:27:21] <meffie> yes.
[14:27:36] <kaduk@jabber.openafs.org/barnowl> I *think* (but did not check) that 13287 (pull in new v5der from
heimdal) will fix part of it, but there's still some grotty work to do
in volser/restorevol.c
[14:27:55] <kaduk@jabber.openafs.org/barnowl> So I'll toss that out there in case anyone is  enthusiastic about it
[14:28:41] <kaduk@jabber.openafs.org/barnowl> (and if not, it will still be there later when someone else is)
[14:28:56] <meffie> ok, i'll see if cheyenne has time to fix restorevol.c . it is stand alone should be be a good project.
[14:29:17] <kaduk@jabber.openafs.org/barnowl> Thanks.  I was wondering if that might be up cheyenne's alley
[14:30:17] <kaduk@jabber.openafs.org/barnowl> I also saw some more bits of review come in for the static-analysis
topic that Pat started.
Unfortunately, my previous strategy for that series had been to try to
work on it in commit order, and the reviews are not coming in in that
order.  IIRC the changes should all be logically independent (even if
they may have some spurious path conflicts), so do people think I should
just go ahead and merge things out of order, as they are ready?
[14:30:20] <meffie> (also, i have code for something called 'createvol' that's a complement for 'restorevol', maybe that could be done too, scope creep)
[14:30:41] <kaduk@jabber.openafs.org/barnowl> If it's useful...
[14:30:53] <kaduk@jabber.openafs.org/barnowl> in other news, yay, buildbot verified the test change!
[14:31:45] <kaduk@jabber.openafs.org/barnowl> Oh, and back on gcc8/enable-checking -- I spotted a bug in 13463 and
uploaded a fix, if people want to re-review
[14:31:49] <meffie> i've been treating the static-analyzer patches from pat as independent. they dont seem to have dependencies.
[14:32:00] <kaduk@jabber.openafs.org/barnowl> Okay.  I will try to do so as well.
[14:33:29] <kaduk@jabber.openafs.org/barnowl> I also got to look at Andrew's libafs-check-blankdir topic and merged
half of it, but I left a couple of questions on 13430 and 13431.
Ideally we'd  get another pair or two of eyes on  those before I merge
them.
[14:34:16] <meffie> ok
[14:34:24] <mvita> ok
[14:34:35] <kaduk@jabber.openafs.org/barnowl> Thanks!
[14:35:33] <kaduk@jabber.openafs.org/barnowl> In the process of doing my review of that, I noticed some weirdness in
the math for rounding up to fragment boundaries, which resulted in
13478.  Andrew gave me some feedback already (and I expect will look at
the updated version), so I'm not explicitly asking for more review and
this could just be an FYI.
[14:36:56] <meffie> ok. curious, was this from a bug report?
[14:37:21] <kaduk@jabber.openafs.org/barnowl> If I understand correctly the relevant codepaths, it potentially affects
our accounting of how much cache storage is allocated to a given vcache,
but does not affect the file size or data buffer lengths that we pass
back to userspace.  (When I was first looking at it I had wondered if it
was a vector that could cause data corruption by reporting larger files
than existed and giving out uninitialized memory.)
[14:37:26] <kaduk@jabber.openafs.org/barnowl> No bug report
[14:38:00] <kaduk@jabber.openafs.org/barnowl> I was just trying to figure out the semantics of the 'chunkBytes' field
by looking at all sites where it was used, and that in turn made me more
confused, since the math was in fact wrong
[14:38:35] <kaduk@jabber.openafs.org/barnowl> Luckily Anders was able to help sort things out.
[14:38:44] <meffie> ah, ok. nice catch. i think there are anecdotal reports of the cache manager getting the partition size wrong.
[14:39:16] <kaduk@jabber.openafs.org/barnowl> And I guess I didn't mention it in the commit message, but I did
experimentally verify that (e.g., my local system has) fragment sizes of
4k are out there in the wild
[14:41:09] <kaduk@jabber.openafs.org/barnowl> Okay.  That was a lot of stuff; hopefully I remembered (or gerrit did)
everything I had wanted to talk about.
[14:41:44] <meffie> ok, thank you!
[14:41:59] <meffie> some quick buildbot updates?
[14:42:59] <kaduk@jabber.openafs.org/barnowl> go for it
[14:43:52] <meffie> ok, the restart worked :)
[14:44:29] <meffie> i think i did a reload that broke the gerrit connection..  i'll avoid reloads from now on.
[14:45:09] <meffie> cheyenne added a gentoo builder this week with clang (i dont recall the version number, but he can change it if we need)
[14:45:37] <meffie> and we bumped up the memory on the sna solaris 11 builder so hopefully it will be doing better.
[14:46:16] <meffie> i'll be ugraded to buildbot 1.8.1, hopefully soon.
[14:46:24] <meffie> upgrading
[14:46:49] <meffie> that's about it. do i need to send email about our ssl issues?
[14:47:36] <kaduk@jabber.openafs.org/barnowl> No harm in doing so; it  would give me a reminder  to go build the
websocket-proxying module for the apache we're running
[14:47:52] <meffie> ok, thanks i will do.
[14:47:57] <kaduk@jabber.openafs.org/barnowl> And I'm also supposed to look into getting a newer python version for
you, which may or may not be easy
[14:49:04] <meffie> no, problem, it seems we are blocked on an older version but it is still getting security fixes.
[14:49:56] <kaduk@jabber.openafs.org/barnowl> On that note, I should probably upgrade the gerrit version, too...
[14:50:15] <wiesand> which OS is running on the server?
[14:50:25] <kaduk@jabber.openafs.org/barnowl> It's RHEL6 IIRC
[14:50:29] <meffie> rhel6 has an end of life this year, no? will that box get upgraded to rhel7 or later?
[14:50:47] <meffie> yes it is RHEL6 (i just checked yesterday)
[14:50:47] <wiesand> el6 eol is november 2020
[14:50:54] <meffie> ah, ok.
[14:51:29] <kaduk@jabber.openafs.org/barnowl> I mostly expect us to be able to skip 7 and go directly to RHEL8 when
it's time to upgrade
[14:51:50] <meffie> ok.
[14:52:08] <wiesand> but it's already in the last phase of the lifecycle where even security holes may not be fixed if RH doesn't consider them seious
[14:52:30] <meffie> well at least we are running python 3 (for buildbot)
[14:52:40] <wiesand> which in the past has affected things like ntpd…
[14:53:02] <wiesand> have I mentioned singularity containers here?
[14:53:10] <meffie> yes
[14:53:24] <kaduk@jabber.openafs.org/barnowl> singularity containers?  I'm not sure I remember this
[14:54:18] <wiesand> https://www.sylabs.io/guides/2.6/admin-guide/index.html
[14:55:17] <mvita> Ben, before we hit the top of the hour I wanted to mention that I put some review comments in rxgk-phase1
[14:55:24] <mvita> I have some more to make
[14:55:30] <kaduk@jabber.openafs.org/barnowl> I saw at least one of those, yes.
Thank you!
[14:55:39] <mvita> and I need to repull my stack to pick up the new changes
[14:55:55] <meffie> have you been able to run with rxgk-phase1 ?
[14:56:01] <mvita> kind of
[14:56:17] <mvita> I can get the ptservers and vlservers to talk to each other in rxgk
[14:56:23] <meffie> yay!
[14:56:30] <mvita> as long as they have the fix I put in the review comments
[14:56:52] <mvita> I'm still working through problems in the clients
[14:57:13] <meffie> clients? i though phase1 was just for the server to server?
[14:57:20] <mvita> vos and pts
[14:57:23] <mvita> not CM
[14:57:35] <meffie> ah ok.
[14:58:01] <meffie> thanks for the progress.
[14:58:16] <mvita> the pts auth code in particular was …. suboptimal … when we got it
[14:58:52] <kaduk@jabber.openafs.org/barnowl> yes, thank you
[14:58:53] <mvita> vos is much more straightforward, I believe I've done vos commands w/ rxgk iirc
[14:59:07] <meffie> great news
[14:59:34] <kaduk@jabber.openafs.org/barnowl> (I did do a brief bit of CM testing long ago, basically by hardcoding a
lot of things.)
[14:59:47] <mvita> I've not messed with that at all
[14:59:58] <kaduk@jabber.openafs.org/barnowl> Probably for the best
[15:00:02] <mvita> aye
[15:00:09] <meffie> top of the hour...
[15:00:17] <mvita> k
[15:00:54] <mvita> Ben, sol vnodes built for me but the vm is giving me aggro, so I can't verify yet
[15:01:12] <kaduk@jabber.openafs.org/barnowl> Shall we adjourn given top of the hour?
[15:01:25] <kaduk@jabber.openafs.org/barnowl> mvita: no rush; thank you for looking into it
[15:01:39] <meffie> oh, one more question... do we really want to kill bozo smail?
[15:02:01] <kaduk@jabber.openafs.org/barnowl> shoot
[15:02:05] <kaduk@jabber.openafs.org/barnowl> I knew I was forgetting something!
[15:02:11] <meffie> heh.
[15:02:19] <mvita> it's forgettable
[15:02:28] <mvita> but I hesitate to kill it
[15:02:32] <kaduk@jabber.openafs.org/barnowl> #define SENDMAIL
"/afs/cellname/fs/dev/localtools/dest/bin/rcs-sendmail"
[15:03:30] <kaduk@jabber.openafs.org/barnowl> Oh jeeze, it hardcodes /tmp/snote.<PID>
[15:03:40] <kaduk@jabber.openafs.org/barnowl> This is like actively harmful to have running
[15:05:00] <kaduk@jabber.openafs.org/barnowl> (debian does not install/package it, btw)
[15:05:51] <kaduk@jabber.openafs.org/barnowl> So my conclusion is that "yes, we really want to kill it"
[15:05:52] <meffie> it is optional? i thought it was baked into bozo
[15:06:24] <kaduk@jabber.openafs.org/barnowl> I'm pretty sure it's optional
[15:06:38] <meffie> (i'm all for clearing out the cobwebs)
[15:07:01] <meffie> ok, lets deorbit it. thanks.
[15:07:42] <kaduk@jabber.openafs.org/barnowl> Hmm, though I guess bos_create does talk about a "notifier program"
[15:07:56] <mvita> yes
[15:08:01] <kaduk@jabber.openafs.org/barnowl> But that seems to be something else, since:
[15:08:09] <kaduk@jabber.openafs.org/barnowl> The AFS distribution does
not include any notifier programs, but this argument is available for
administrator use. See L</NOTES>.
[15:08:32] <meffie> so, the think to get rid of is "smail-notifier", right?
[15:08:40] <meffie> s/think/thing/
[15:08:53] <kaduk@jabber.openafs.org/barnowl> yes
[15:09:44] <meffie> oh, ok, yeah it's not documented. let us remove it.
[15:10:01] <kaduk@jabber.openafs.org/barnowl> Sounds like a plan!
[15:10:07] <meffie> if someone needs it, they can build a replacement bin.
[15:10:19] <meffie> out of tree :)
[15:10:46] <meffie> ok thanks.
[15:10:51] <meffie> that's all i have.
[15:11:36] <kaduk@jabber.openafs.org/barnowl> thanks
[15:11:40] <kaduk@jabber.openafs.org/barnowl> motion to adjourn?
[15:11:58] <mvita> so moved
[15:12:13] <wiesand> Thanks everybody!
[15:12:23] wiesand leaves the room
[15:12:25] <mvita> Have a good weekend!
[15:14:04] <meffie> have a good weekend
[15:14:08] <kaduk@jabber.openafs.org/barnowl> thanks; you too
[16:40:45] mvita leaves the room
[18:44:22] meffie leaves the room
[18:54:48] meffie joins the room
[19:33:34] meffie leaves the room
[20:25:05] meffie joins the room
[23:28:41] meffie leaves the room