- Thursday, June 17, 2021

release-team@conference.openafs.org

Thursday, June 17, 2021< ^ >

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

Room Configuration

Room Occupants

GMT+0
[12:04:23] mbarbosa joins the room
[13:49:10] meffie joins the room
[13:58:46] Cheyenne joins the room
[13:59:48] <meffie> using a better browser for the call today.
[14:00:06] <meffie> (oh that was for mark)
[14:02:13] <mvita2> I'm there, but where's everyone else?
[14:02:58] <mvita2> what's going on w/ standup?
[14:03:06] <meffie> we are on the morgan call
[14:03:14] <mvita2> OHHHHHHHHH
[14:03:20] <mvita2> so sorry
[14:07:41] <meffie> did you join?
[14:11:45] <mvita2> yes
[14:16:42] meffie leaves the room
[14:39:13] meffie joins the room
[14:39:46] <meffie> (wrong window, sorry for the noise!)
[14:39:56] meffie leaves the room
[14:40:08] <mvita2> OH
[15:44:42] wiesand joins the room
[15:58:40] yadayada joins the room
[16:00:12] <wiesand> Hello
[16:00:22] <yadayada> Hi All
[16:01:09] <mvita2> present, hi everyone
[16:01:39] meffie joins the room
[16:01:48] <Cheyenne> Hello
[16:01:48] <meffie> hello
[16:02:13] <wiesand> No 1.8.8pre1 fedback yet. Which is both good and bad.
[16:03:35] <meffie> maybe everyone was busy at the workshop :)
[16:03:51] <mvita2> oh, there was a workshop?
[16:03:58] <wiesand> Probably
[16:04:34] <wiesand> I created 14641 and 14642. For now they assume no further changes before the release, but that can be modified of course.
[16:05:24] <wiesand> I think we should release 1.8.8 ~ at the same time Linux 5.13 is.
[16:06:26] <meffie> sound great. thank you
[16:06:37] <wiesand> I may still pull up some more FBSD changes before (including the removal of support for 8.x and 9.x)
[16:07:13] <meffie> sounds good.
[16:07:32] <kaduk@jabber.openafs.org/barnowl> (I am here but a bit distracted)
[16:09:01] <wiesand> That's about what I have on 1.8 for today.
[16:10:29] <Cheyenne> 1.8.x and master still build clean on latest linux-5.13 rc
[16:10:41] <meffie> thanks
[16:11:03] <wiesand> Thanks Cheyenne. Chances are they won't break us that late in their cycle.
[16:11:30] <kaduk@jabber.openafs.org/barnowl> I don't have specific topics for 1.9 to discuss, but am interested in
hearing what feedback others took from the workshop -- if there are
particular things in gerrit to go back to or prioritize, etc.
[16:11:41] <mvita2> weisand:  off-topic - do you know which site Kristen Nielsen was representing (if any)?
[16:12:19] <wiesand> I think Århus University
[16:12:25] <mvita2> kaduk:  I have been trying to chase down the problem reported with simultaneous access from "magic mount" and normal /afs
[16:13:02] <mvita2> but I don't have enough details - Billings just said that the client becomes unresponsive (to rx requests?)
[16:13:07] <mvita2> do you have any info?
[16:13:08] <kaduk@jabber.openafs.org/barnowl> I assumed that was the thing where linux insists on a given dentry
only having one path to it
[16:13:22] <kaduk@jabber.openafs.org/barnowl> but IIRC that situation is a panic, not an rx hang
[16:13:25] <mvita2> That's what I was thinking too but that wouldn't cause a hang
[16:13:31] <meffie> thanks ben. would it be possible to get andrew's ctap test patches merged soon? i've been basing changes on top of those in order to add more unit tests.
[16:13:57] <kaduk@jabber.openafs.org/barnowl> what's the gerrit number for the ctap patches?
[16:14:55] <meffie> 14632 and children
[16:15:25] <meffie> i need them for the tmp dir stuff, to add tests for BosConfig
[16:15:30] <wiesand> re workshop feedback: I'm a bit worried about Jeffrey's warning regarding the number of bind mounts on a linux system
[16:16:26] <kaduk@jabber.openafs.org/barnowl> Ah, that stack (14632).  I had made a pass over that and have some
comments staged on 14632 itself ... I think I wanted some responses
and/or updates to the earlier changes.
[16:16:36] <meffie> oh, ok.
[16:17:04] <meffie> i hope we can get them in, if not, i'll just push changes based on top of them.
[16:17:15] <meffie> since we hate merges :)
[16:18:12] <kaduk@jabber.openafs.org/barnowl> I don't think there are any fundamental issues that would prevent
getting them in.
[16:18:34] <kaduk@jabber.openafs.org/barnowl> I am just usually reluctant to push edits to others' stuff and merge
them right away, without getting feedback about the proposed/requested
changes
[16:18:49] <mvita2> re: bind mounts  I missed that warning
[16:18:50] <meffie> cheyenne and i have a set of changes to remove all the stack allocated strings in bosserver and bos, and some more unit tests (that's why i'd like to have those changes)
[16:19:44] <meffie> most of it is very straightforward and a cathartic cleanup
[16:20:18] <meffie> 1987 called, and they want their code back.
[16:21:15] <kaduk@jabber.openafs.org/barnowl> :)
[16:21:40] <Cheyenne> On an entirely unrelated topic.. Just found out that my nick(s) on Freenode are no longer registered
[16:21:53] <kaduk@jabber.openafs.org/barnowl> There's still a lot of code in the tree that has as its primary
motivation a belief that malloc is expensive and goes to some
contortions to avoid it at the cost of robustness and security risks
[16:21:59] <kaduk@jabber.openafs.org/barnowl> everyone's
[16:22:20] <kaduk@jabber.openafs.org/barnowl> https://www.devever.net/~hl/freenode_suicide is an article I've seen
linked a couple times
[16:22:31] <meffie> yes
[16:22:44] <meffie> i've removed all my freenode configs from irc clients.
[16:23:25] <meffie> re the workshop, i though participation was wonderful, and was happy to see so many good talks and discussions.
[16:23:36] <wiesand> mark: it was mostly in the chat during the last morgan stanley talk, and later during the bof
[16:23:40] <Cheyenne> the last channel hold out just moved to Libera.chat
[16:24:16] <meffie> i was happy to hear yadav is using the openafs-ansible collection in anger. maybe i should say something about on the lists again.
[16:25:20] <meffie> kaduk, wiesand, any feedback from andrew's talk?
[16:25:28] <meffie> about ubik-kv ?
[16:26:33] <kaduk@jabber.openafs.org/barnowl> It was interesting, and has some ideas that will take some time to
digest properly :)
[16:26:59] <wiesand> Er, not really, except that I'm kind of anxious touching ubik in general.
[16:27:33] <meffie> that was the last bullet, "change is scary"
[16:27:47] <wiesand> right ;-)
[16:28:11] <kaduk@jabber.openafs.org/barnowl> indeed
[16:28:11] <meffie> the idea is to be able to have a smooth, even online transition.
[16:28:27] <meffie> seems like a good way forward.
[16:29:26] <meffie> ben's talk included a topic, how can we help ben. i think he said we could review old gerrits. i can start that next week if that helps.
[16:29:55] <kaduk@jabber.openafs.org/barnowl> I am pretty sure there is some good stuff that we want to take, but
has been ignored for 3 years :-/
[16:30:27] <kaduk@jabber.openafs.org/barnowl> Whether or not we want that more urgently than we want, say, recvmmsg
is another matter, though
[16:30:31] <mvita2> Cheyenne rebased the cache manager negative volume lookup cache yesterday
[16:31:20] <wiesand> Adding a simple comment like "we may want this" will put an old change back onto the radar for a while.
[16:31:57] <meffie> yes, that can be done. some things can be abandoned i hope. others will need to be updated or redone, likely.
[16:32:07] <kaduk@jabber.openafs.org/barnowl> wiesand: yes, exactly
[16:32:52] <meffie> maybe we could add more gerrit labels to help organize?
[16:34:13] <kaduk@jabber.openafs.org/barnowl> worth looking into, though I don't remember a functionality that would
be directly applicable for this purpose
[16:34:31] <wiesand> Do we see those on the overview page, or is there a simple way to filter for those?
[16:35:00] <meffie> i was thinking they could be used in queries.
[16:36:43] <meffie> for now, comments and email to release-team is probably the short term way forward.
[16:36:53] <kaduk@jabber.openafs.org/barnowl> +1
[16:37:06] <wiesand> +1
[16:38:03] <meffie> also, i'm often tempted to ask questions in openafs-devel, for stuff i ask here when i have peoples attention. maybe we could use openafs-devel more?
[16:38:58] <wiesand> it's too quiet there, agreed
[16:39:42] <kaduk@jabber.openafs.org/barnowl> openafs-devel the mailing list or is there a jabber room by that name?
[16:39:53] <meffie> the mailing list i mean.
[16:40:03] <kaduk@jabber.openafs.org/barnowl> Ah, got it
[16:40:36] <meffie> sorry, wrong name.
[16:41:06] <meffie> so, i use it so little i spelled it wrong ;)
[16:41:08] <kaduk@jabber.openafs.org/barnowl> Going back to my workshop notes, I do have an entry about a claim
that openafs currently has some ubik logic to update the cacheVersion
prior to the completion of the quorum consistency protocol, and that
this behavior is dangerous in some sense.
[16:41:51] <mvita2> oh, I perked up on that one too
[16:42:06] <mvita2> but haven't checked it out yet
[16:42:13] <kaduk@jabber.openafs.org/barnowl> (I didn't have a chance to look into it yet, either)
[16:42:14] <meffie> maybe we can investigate and write some tests.
[16:42:35] <mvita2> did jeffrey say it was merged, or merely under review?
[16:42:59] <mvita2> because my first though was he was talking about read-during-write
[16:43:08] <mvita2> thought
[16:43:25] <kaduk@jabber.openafs.org/barnowl> I thought he said it was merged
[16:43:39] <kaduk@jabber.openafs.org/barnowl> I do know that he has concerns about read-during-write, at least in
some previous form.
[16:43:40] <wiesand> that's my understanding too
[16:43:43] <mvita2> pretty sure read-during-write is not merged… looking
[16:43:52] <kaduk@jabber.openafs.org/barnowl> read-during-write is not merged
[16:43:52] <meffie> correct.
[16:44:10] <meffie> but there are some other changes related to the version number.
[16:45:46] <kaduk@jabber.openafs.org/barnowl> So is Mark, Mike, or both going to investigate the ubik cacheVersion
potential issue? ;)
[16:45:59] <mvita2> yes, I am looking for it
[16:46:08] <mvita2> and will consult w/ Marcio if  needed
[16:46:30] <kaduk@jabber.openafs.org/barnowl> Excellent, thanks
[16:46:56] <meffie> yes, we should ask Marcio, he was last to look at that.
[16:47:35] <wiesand> yadav: will you make your slides available?
[16:48:04] <kaduk@jabber.openafs.org/barnowl> I will probably think about the rx congestion and flow control
algorithms at some point (not imminently) based on some of the points
Simon made.  I will want to consult with some of my IETF colleagues to
get a sense for what kind of experiments and testing I need to do
before deploying any changes on a broader scale, in order to be a
responsible internet citizen.
[16:48:27] <meffie> excellent. thank you.
[16:49:17] <meffie> now that the ansible stuff seems to be working well, i'll get back to the "clean up needed so we can start ivp6"
[16:49:39] <kaduk@jabber.openafs.org/barnowl> excellent
[16:49:40] <wiesand> nice
[16:49:40] <Cheyenne> Is there interest in refreshing the cleanup of the AFS_*LINUX*_ENV commits? (14389)
[16:50:11] <meffie> some of it is already in gerrit, some if it still in github (in a public fork)
[16:50:44] <kaduk@jabber.openafs.org/barnowl> I'd like to get back to the AFS_*LINUX*_ENV stack, yes ... it is
mostly easy to mechanically review
[16:51:25] <Cheyenne> Okay.. I'll do a rebase on the stack to bring it up to date
[16:52:26] <kaduk@jabber.openafs.org/barnowl> thanks
[16:53:31] <wiesand> that change looks quite attractive since it removes lots of cruft
[16:54:21] <meffie> in other news, we have added fedora 3x to our openafs rpm builder for vanilla 1.9.1.
[16:54:48] <meffie> i feel if you are running fedora 34, you can try 1.9.1 for us :)
[16:55:07] <wiesand> +1
[16:55:24] <wiesand> there's an open ticket re F33
[16:58:16] <meffie> do you know the number? i dont see it after a quick search
[16:58:48] <wiesand> 135221
[16:59:15] <meffie> thanks i'll take a look.
[16:59:25] <wiesand> May well be (hopefully is) solved with 1.8.8
[16:59:39] <Cheyenne> Yes it's fixed in 1.8.8
[16:59:52] <meffie> without looking, that's my guess. i know 1.8.7 will not work unpatched.
[17:00:08] <wiesand> Fine, another number to add to NEWS :)
[17:00:10] <meffie> (one of the reason we are building 1.9.x for fedora)
[17:00:26] <meffie> (actually the main reason)
[17:01:01] <Cheyenne> Once 1.8.8 is out it should be okay on bleeding edge linux kernels for intel
[17:02:05] <wiesand> Very good. Let's hope we'll get it out soon.
[17:02:49] <wiesand> Anything else to discuss today?
[17:03:08] <meffie> nothing more here. thank you.
[17:03:27] <kaduk@jabber.openafs.org/barnowl> nothing here
[17:03:34] <mvita2> nada
[17:03:47] <Cheyenne> (re 135221 -- I believe the gerrit is 14510)
[17:04:01] <Cheyenne> (14511 and stack)
[17:04:39] <wiesand> thanks
[17:05:08] <meffie> is this the part of the meeting when wiesand says "review review review" ?
[17:05:16] <mvita2> heh
[17:05:47] <wiesand> well, not much to review on 1.8.x right now z(but stau tuned ;-)
[17:06:00] <meffie> ok!
[17:08:46] <wiesand> OT: yadav: one of my colleagues is very interested in your CSI work. In particular, he'd like to know whether it's a way to share AFS workgroup space with containers, and how you're dealing with token renewal. Is there a short answer to those? If not, may I put you in touch?
[17:09:28] <wiesand> That aside: let's adjourn?
[17:09:56] <meffie> have a good week all.
[17:10:08] <kaduk@jabber.openafs.org/barnowl> thanks everyone, and have a good week
[17:10:41] <wiesand> Thanks a lot everyone!
[17:13:12] <Cheyenne> Have a good week everyone
[17:14:36] meffie leaves the room
[17:27:19] wiesand leaves the room
[18:05:01] mbarbosa leaves the room
[18:21:04] yadayada leaves the room
[19:10:16] mbarbosa joins the room
[21:28:45] mbarbosa leaves the room
[22:51:47] Cheyenne leaves the room