- Friday, January 29, 2021

release-team@conference.openafs.org

Friday, January 29, 2021< ^ >

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

Room Configuration

Room Occupants

GMT+0
[12:06:49] mbarbosa joins the room
[12:55:40] mbarbosa leaves the room
[12:56:00] mbarbosa joins the room
[13:15:16] meffie leaves the room
[16:15:41] wiesand joins the room
[16:15:52] cwills joins the room
[17:00:03] yadayada joins the room
[17:00:08] <wiesand> Howdy
[17:00:13] <yadayada> Hi All
[17:00:23] <cwills> Hello :)
[17:00:32] <kaduk@jabber.openafs.org/barnowl> greetings
[17:01:43] <wiesand> Merged away the aklog changes.
[17:02:22] meffie joins the room
[17:02:27] <yadayada> Thanks Wiesand I did try some testing and all is fine
[17:02:28] <meffie> good day
[17:03:24] <wiesand> Pulled up the rx-restart-hang ones Mike brought up a while ago, plus a few others Andrwe mentioned in a comment on 14264. Stack is 14052..07. Review review review...
[17:03:28] <wiesand> Hello Miek
[17:03:31] <wiesand> Mike
[17:03:51] <meffie> thank you wiesand
[17:05:02] <wiesand> To-Do-List comprises FBSD, 10831, Linux 5.11 and Big Sur.
[17:06:07] <meffie> (aside, hmm, i wonder of the topic being discussed today on #openafs is related to the rc-restart issue.)
[17:06:16] <yadayada> For BigSur I reveiwed new patch for 14430. Looks fine to me, but want to do one round of testing. I think early next week it will be done
[17:06:31] <meffie> thanks yadayada
[17:06:35] <cwills> The 5.11 patch should fit cleanly on 1.8.x (once it's been merged into master)
[17:06:52] <kaduk@jabber.openafs.org/barnowl> What is "the rc-restart issue"?  (I read the #openafs thread, but you
seem to be referring to something else.)
[17:06:59] <wiesand> Good, thanks a lot.
[17:07:20] <wiesand> What's the rc-restart issue?
[17:07:46] <meffie> er, i mean the rx-restart-hang patches.
[17:08:32] <meffie> just a wild guess on my part, maybe not related at all without looking closer.
[17:09:37] <kaduk@jabber.openafs.org/barnowl> And
https://gerrit.openafs.org/#/q/topic:rx/restart-hang+(status:open+OR+status:merged)
seems to say that the rx-restart-hang patches are already merged to
master
[17:10:36] <meffie> thanks ben
[17:12:17] <wiesand> they are, and the 1.8.x pullups are 14505..14507
[17:12:59] <wiesand> the earlier changes in that stack will need to be reviewed too though, due to path conflicts
[17:14:12] <wiesand> 14508 makes sense to me
[17:14:51] mbarbosa leaves the room
[17:15:11] mbarbosa joins the room
[17:15:27] <wiesand> That's about what I have on 1.8.x today.
[17:15:39] <kaduk@jabber.openafs.org/barnowl> Oh, right, I was going to merge 14508 unless we thought we could redo
it in configure quickly.
[17:15:59] <kaduk@jabber.openafs.org/barnowl> cwills, do you want me to hold off on 14508 so we can investigate
further?
[17:16:17] <cwills> let me check
[17:17:00] <cwills> Go ahead with andrews's fix.  I think a more generic approach can be added later
[17:17:24] <kaduk@jabber.openafs.org/barnowl> Sounds good; that was basically what I was thinking.
[17:17:36] <kaduk@jabber.openafs.org/barnowl> merged 14508
[17:18:01] <wiesand> Relying on the advertised kernel version should be the very last resort anyway. In particular, RHEL kernels tend to be much more similar to last week's mainline kernel than what the vesion number suggests.
[17:18:06] <cwills> master with the 14499 patch still builds with 5.11
[17:18:11] <cwills> (latest rc)
[17:19:14] <kaduk@jabber.openafs.org/barnowl> That segues nicely into 14500, where we are adding an
afs/sysincludes.h inclusion from LINUX/osi_machdep.h
[17:20:12] <kaduk@jabber.openafs.org/barnowl> I kind of thought that the general expectation was that the
OS/osi_machdep.h got included from afsincludes.h (or afs_osi.h), and
that anything using either of those had to include sysincludes.h as
the first thing.
[17:20:29] <kaduk@jabber.openafs.org/barnowl> But I also don't know what kind of a mess you (cwills) got into
without the sysincludes where it is in 14500.
[17:21:29] <cwills> Oh .. that.. it was that I had to start adding an include for sysincludes in a bunch of files
[17:22:05] <kaduk@jabber.openafs.org/barnowl> Hmm, okay.  "a bunch" being like "more than three"?
[17:22:23] <cwills> I stopped at 3
[17:22:31] <wiesand> ;-)
[17:22:51] <kaduk@jabber.openafs.org/barnowl> Like a normal reasonable person would, yes
[17:23:10] <cwills> one, two, many
[17:24:22] <meffie> i heard the numbers are; zero, one, many
[17:25:03] <kaduk@jabber.openafs.org/barnowl> Anyway, not too much else from my end this week.  I'm still plugging
away with what I think is needed for a debian buster update (it's slow
because I have to recreate my VM infrastructure on my "new" laptop),
and it's been kind of a tiring week in general.
[17:25:12] <cwills> real programmers, the first index is 1 :)
[17:25:47] <wiesand> ah those fortran guys…
[17:26:07] <cwills> :)
[17:26:40] <kaduk@jabber.openafs.org/barnowl> So I had to cause problems last week for the pending IETF HTTP/3
specification -- it uses a header compression scheme called QPACK, and
there's a dynamic table of header field state, which in the protocol
is 0-indexed.  But the examples showed a table that was 1-indexed :)
[17:27:14] <meffie> oof
[17:27:24] <wiesand> oh my, 50 years after...
[17:27:28] <meffie> nice catch.
[17:28:51] <kaduk@jabber.openafs.org/barnowl> As the previous thread indicates, I'm looking at the linux 5.11 stuff,
and the bigsur is next on my slate.
[17:29:25] <kaduk@jabber.openafs.org/barnowl> It sounds like the discussion on 14430 w.r.t. the versioning scheme
Apple is using has concluded that the current state of the change is
correct; is that right?
[17:29:47] <yadayada> that's right, but I want to do some testing on it.
[17:30:01] <yadayada> by early next week will let you know the results
[17:30:10] <kaduk@jabber.openafs.org/barnowl> okay, thank you!
[17:30:24] <meffie> excellent, thank you yadayada
[17:32:04] <kaduk@jabber.openafs.org/barnowl> I don't think I had anything else for today.  As I said earlier, it
was kind of a slow week for openafs here.
[17:32:57] <meffie> thanks for getting the changes for buster in salsa.debian.org
[17:34:18] <wiesand> Anything else to discuss today>
[17:34:20] <wiesand> ?
[17:34:46] <cwills> i've been reworking some of the static-analysis patches
[17:35:10] <yadayada> One query I had on openafs-server service file for redHat
[17:36:17] <meffie> query?
[17:36:19] <yadayada> I our code I see the we have various timeouts for server as 30 sec .
[17:36:26] <yadayada> /* max time to wait for fileserver shutdown */
#define FSSDTIME        (30 * 60)       /* seconds */
[17:37:12] <kaduk@jabber.openafs.org/barnowl> 30 minutes, that looks like
[17:37:39] <yadayada> But in our service file we do not have systemd timeout specified, due to this sometimes systemd process will kill the process since it just wait for 90 sec (default timeout)
[17:38:29] <yadayada> This is just heads up on this will share some more info in coming days. But any input from your side will be helpful
[17:38:54] <kaduk@jabber.openafs.org/barnowl> It looks like we don't change the default timeout in the debian
service file, either.  We probably should.
[17:39:22] <wiesand> Well, systemd is so carefully optimized for the developers' laptops, it will typically shut down the network immediately….
[17:40:03] <wiesand> Maybe it's better on debian
[17:40:05] <kaduk@jabber.openafs.org/barnowl> Yeah, so we have to use a large hammer and say:
Wants=network-online.target
After=network.target network-online.target local-fs.target
sysinit.target
[17:40:54] <yadayada> I typically saw systemd killing bos server before it can cleanly shutdown. Struggling getting good timout number
[17:42:54] <meffie> there's no good value
[17:43:19] <yadayada> right, I will create a gerrit for it with some initial timeouts and then we can discuss there
[17:43:34] <wiesand> I'd never ever dare to simply reboot an AFS server and rely on systemd getting the order right. I always stop services manually before rebooting.
[17:43:38] <meffie> we need a way to tell the fileserver to gracefully let go
[17:44:24] <cwills> what kill -9 isn't good enough O_o
[17:44:47] <wiesand> That's not a problem with systemd as such, but rather the way it's configured on the distributions I know.
[17:45:44] <meffie> thanks for digging into this yadayada.
[17:45:58] <yadayada> sure will discuss more on it in coming days. Thanks
[17:46:29] <kaduk@jabber.openafs.org/barnowl> thanks for looking into it
[17:46:30] <wiesand> *Please* prove me wrong and find a real solution ;-)
[17:47:33] <meffie> kaduk are you able to release volumes on grand.central.org? if so could you release www.workshop <http://www.workshop> ? (i've been waiting on jhutz)
[17:48:05] <wiesand> Ben is. So am I.
[17:48:07] <kaduk@jabber.openafs.org/barnowl> I think wiesand and I can both do those releases
[17:48:09] <meffie> the foundation asked me to update the 2021 site.
[17:49:11] <meffie> ah! thanks, i did not know. can one of you vos release www.workshop <http://www.workshop> -c grand.central.org?
[17:49:20] <kaduk@jabber.openafs.org/barnowl> bjk@pleonasm:~$ vos release www.workshop -c grand.central.org
Released volume www.workshop successfully
[17:49:36] <meffie> they are planning on having a "virtual workshop" this summer.
[17:49:38] <wiesand> beat me
[17:50:19] <meffie> thank you guys. it is great to know you have access to release that volume :)
[17:51:05] <wiesand> IMHO, those "real" workshops having folks fly around the globe are mostly a thing of the past.
[17:52:45] <meffie> yes, so at least it will be easier to attend but not nearly as fun imho
[17:54:31] <meffie> thanks, that's all i have.
[17:54:41] <kaduk@jabber.openafs.org/barnowl> cwills: I left a note with one more thing for you to do on 14500
[17:54:51] <cwills> Okay.. will take a quick look
[17:54:55] <kaduk@jabber.openafs.org/barnowl> thanks!
[17:55:29] <wiesand> Alternative ways of team building will have to evolve. But they will.
[17:56:10] <wiesand> Now waiting for the snow to come. There's an official "severe weather" warning.
[17:56:40] <cwills> I think we are supposed to have a storm coming through as well
[17:56:57] <wiesand> Luckily, if it takes 4 more minutes to start falling, I'm not obliged to clear my sidewalk before tomorrow morning :)
[17:57:33] <kaduk@jabber.openafs.org/barnowl> :)
[17:58:33] <wiesand> Let's adjourn. Thanks a lot everybody!
[17:58:40] <yadayada> Thanks all
[17:58:56] <kaduk@jabber.openafs.org/barnowl> Thanks everyone; have a good weekend!
[17:59:01] <cwills> have a good week everyone
[17:59:42] wiesand leaves the room
[18:42:55] meffie leaves the room
[19:23:20] yadayada joins the room
[19:33:32] yadayada joins the room
[19:33:44] yadayada leaves the room
[19:39:26] yadayada leaves the room
[19:50:11] yadayada leaves the room
[22:02:29] mbarbosa leaves the room