- Friday, July 3, 2020

release-team@conference.openafs.org

Friday, July 3, 2020< ^ >

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

Room Configuration

Room Occupants

GMT+0
[14:47:29] yadayada joins the room
[15:54:57] cwills joins the room
[15:55:43] <cwills> Good morning Ben
[15:55:53] <kaduk@jabber.openafs.org/barnowl> greetings!
[15:59:23] wiesand joins the room
[15:59:49] <yadayada> hi all
[15:59:55] <wiesand> Hello
[16:00:32] <cwills> Hello
[16:00:53] <kaduk@jabber.openafs.org/barnowl> Hi Yadav, hi Stephan
[16:01:32] <wiesand> Ben, you sound better than the last two weeks?
[16:02:07] <kaduk@jabber.openafs.org/barnowl> Yeah, I've been feeling pretty close to normal productivity levels all
week.  "Now I just have to catch up on things that piled up while I
wasn't"
[16:02:42] <wiesand> Glad to hear that.
[16:02:43] <cwills> It's been a Monday all week this week
[16:03:50] <wiesand> Sorry…
[16:04:27] <wiesand> Ok, with 1.8.6 out the door, the next item on my agenda is obviously 1.8.7pre1
[16:04:50] <wiesand> I saw the last Linux 5.8 change was merged on master - thanks!
[16:05:03] <cwills> just a distracting week due to some outside events (thankfully things turned out better then they could have)
[16:05:09] <kaduk@jabber.openafs.org/barnowl> You're welcome!  Thanks to everyone that worked on it
[16:05:32] <kaduk@jabber.openafs.org/barnowl> I was pleasantly surprised when the final version was much shorter
than the first one I looked at :)
[16:06:24] <wiesand> So pulling that up and getting 1.8.x to build on linux-next is my priority.
[16:06:40] <cwills> The actual code fix was fairly trivial.. it was hunting through LInux's history that was difficult.
[16:07:04] <cwills> I'll get the 1.8.x version of the 5.8 stuff probably today
[16:07:32] <wiesand> Linux 5.8 is just a few weeks out. I hope for a much shorter cycle for 1.8.7 than we had for 1.8.6
[16:07:57] <cwills> I'll also work on the Clang 10 stuff probably today as well..
[16:08:02] meffie joins the room
[16:08:33] <kaduk@jabber.openafs.org/barnowl> Is it "open season" on 1.8.x cherry-picks now, or do you want to keep
some structure to it?
[16:09:22] <meffie> Greetings
[16:10:28] <wiesand> Please let me merge what's there already, + Linux 5.8, and then discuss what else should make 1.8.7.
[16:10:46] <meffie> Ok thanks
[16:11:33] <kaduk@jabber.openafs.org/barnowl> Seems reasonable, thanks for laying out the plan
[16:11:37] <wiesand> That's about what I have for today.
[16:12:50] <yadayada> I have dropped a mail on Nat Ping issue I talked about last week
[16:13:27] <yadayada> Thanks Ben for quick comments, Stephan I added you to mail just now, you will receive it shortly
[16:13:52] <yadayada> Still in middle of analysis and getting fix, but your feedback will really help
[16:14:30] <meffie> Thanks yadav
[16:14:32] <yadayada> Also wanted to know based on current analysis, do you think fileserver will also get impacted my it, mainly on Memory usage side ?
[16:14:42] <kaduk@jabber.openafs.org/barnowl> Thank you for sending the interim analysis; it makes sense on first
read.  I think I remember some level of subtlety in some of the event
handlers about what order I could drop references in, but didn't get
to check about this specific one
[16:15:03] <cwills> So for the clang stuff (fall through comments), I'm planning on creating a new 'afs_fallthrough' attribute and turning off the fallthrough switch for generated code (flex or external)
[16:15:14] <wiesand> Received it, thanks. I'll have to look at it though.
[16:16:06] <wiesand> Fixing NAT pings is obviously a major 1.8.7 objective.
[16:16:42] <kaduk@jabber.openafs.org/barnowl> Yadav: I think there's supposed to be a periodic event that goes and reaps
connection objects for which the only reference count is for keepalive
traffic, but would have to confirm that.  If my recollection is
correct, the impact on fileserver memory should not be too large
[16:17:30] <yadayada> ok, I saw one post https://lists.openafs.org/pipermail/openafs-info/2012-January/037449.html
[16:17:43] <yadayada> Here some user faced same NATPing flood
[16:17:57] <yadayada> and they talked about ref leak, below is snippet
[16:18:03] <yadayada> “””
As a potential cause, leaking connection reference counters
and subsequently failing garbage collection come to mind,
but we cannot point to anything in the code yet.
“””"
[16:18:54] <yadayada> In 1.8 code base mostly NatPing is disbaled
[16:19:19] <wiesand> Yes, it was a real problem for CERN, and for DESY too.,
[16:19:28] <kaduk@jabber.openafs.org/barnowl> Cheyenne: thanks, that sounds good
[16:21:13] <wiesand> I remember repurposing a former AFS fileserver, and it was being bombarded by NAT pings to the point where it was actually unusable…
[16:21:27] <kaduk@jabber.openafs.org/barnowl> yikes
[16:22:07] <yadayada> We saw some fileserver with huge RSS memory usage, just trying to link NatPing with That, I may be incorrect there
[16:22:54] <kaduk@jabber.openafs.org/barnowl> It's certainly worth checking
[16:23:14] <wiesand> The system I'm talking about was no longer running an AFS server, for weeks.
[16:24:05] <wiesand> This issue is so potentionally nasty because you have to restart all affected clients to get rid of it.
[16:26:47] <yadayada> right, I am playing with code by intrumenting the code paths, Ben in case you need more details or want to add some more intrumentation in code, do let me know
[16:26:58] <kaduk@jabber.openafs.org/barnowl> will do, thanks
[16:29:04] <yadayada> This is all I have for NatPing, will keep updated
[16:29:04] <wiesand> Mike: thanks for finding gerrit 14263/
[16:29:36] <meffie> You are welcome
[16:31:09] <wiesand> Is there more to discuss today?
[16:32:14] <meffie> Ben can we have the buildbot port opened on the new host?
[16:32:41] <kaduk@jabber.openafs.org/barnowl> I will send mail requesting the firewall change this weekend
[16:32:57] <meffie> Thanks
[16:33:06] <kaduk@jabber.openafs.org/barnowl> I don't have much noteworthy, just trying to plug away at the gerrits.
[16:33:26] <kaduk@jabber.openafs.org/barnowl> 1.8.6 is in debian (just a day after someone files a bug noting that
we don't build against the 5.7 kernel)
[16:33:39] <wiesand> Great
[16:34:44] <wiesand> If we're finished for today, let's adjourn. Enjoy Independence Day!
[16:35:06] <yadayada> Thanks All
[16:35:14] <kaduk@jabber.openafs.org/barnowl> I will happily be independent from my work inbox, but not from weeding
in the garden, taxes, or housework :)
[16:35:19] <kaduk@jabber.openafs.org/barnowl> Thanks everyone!
[16:35:20] <cwills> Thanks everyone
[16:35:27] <meffie> Thanks have a good weekend
[16:35:54] wiesand leaves the room
[16:49:32] cwills leaves the room
[17:20:59] yadayada joins the room
[17:33:41] meffie leaves the room
[17:33:42] meffie joins the room
[17:37:09] yadayada leaves the room
[18:50:06] meffie leaves the room
[18:50:06] meffie joins the room
[19:27:10] yadayada joins the room
[19:43:13] yadayada leaves the room
[20:15:32] yadayada leaves the room
[23:52:55] meffie leaves the room
[23:52:55] meffie joins the room