<borneoa___>
zapb_: what happened with patches 8863 and 8907? They are in conflict
jfsimon has joined #openocd
tsal has joined #openocd
<borneoa___>
zapb_: I'm reworking it
<PaulFertser>
borneoa___: hey, I hope the system feels OKish now. Looks like Jenkins used to slowly grow in memory size (not VM heap which is limited) before and that was leading to "swap thrashing". It looks like it's stable in how much RAM it takes now, and we also get all the latest-and-greatest security fixes with this current LTS release, so it all looks good from my side.
<borneoa___>
PaulFertser: I actually still got some slowness and server not responding. But I can go ahead
<PaulFertser>
Still we might want to ask the VPS provider for one downsized VM to host Gerrit individually because apparently I'm not cool enough to make it responsible 100 % of the time.
<PaulFertser>
I thought these days with cgroups and systemd providing all those automatic slicing etc it should be really straightforward but somehow I do not see many real life stories :/
<PaulFertser>
Esp. given that the "i/o latency controller" is only about data i/o and doesn't seem to take swap into account.
<borneoa___>
PaulFertser: right now it's not responding
<PaulFertser>
borneoa___: yeah, it was i/o wait again. I'm adjusting some parameters for swap now, guess if gerrit isn't swapped out too much then it'll be "fine".
<PaulFertser>
It's respondiing again and I didn't restart anything.
<zapb_>
borneoa___, thanks!
<zapb_>
borneoa___, what about the other patches in the nordic-config topic?
<PaulFertser>
Are e-mails about Jenkins success/failure still working as expected?
<PaulFertser>
At least they get delivered to me.
<borneoa___>
PaulFertser: email looks ok
<borneoa___>
zapb_: yes I'm merging them. Now I wait for the rebuild of 8907
<borneoa___>
PaulFertser: again not responding. It's the usual uncomfortable situation during the merge
<zapb_>
Yes, mails are delivered but really we have to think about the infrastructure. While Jenkins runs (?) Gerrit is not usable and also sometimes the website as well
<borneoa___>
PaulFertser: works again. Anyway, I'm already used at taking some pause during the merge
<zapb_>
If you have only a couple of minutes during the day for review or something similar and it's not working then this is quite annoying
<PaulFertser>
borneoa___: the service wasn't restarted but my changes to /sys/fs/cgroup/system.slice/gerrit.service/memory.swap.max were reset, I can't understand what could have happened yet.
<PaulFertser>
zapb_: indeed. We should just migrate to using a separate VM for Gerrit it looks like.
<zapb_>
PaulFertser, yep, sounds reasonable
<PaulFertser>
I see, "systemctl daemon-reload" adjusts/resets all the parameters even for running units, OK, good to know.
d_olex has quit [Ping timeout: 252 seconds]
<zapb_>
borneoa___, thanks for merging!
<zapb_>
PaulFertser, do we have an additional VM for Gerrit? Or how could we proceed here?
<PaulFertser>
zapb_: I think Spencer can ask the hoster for another one.
tsal has quit [Ping timeout: 252 seconds]
<PaulFertser>
The current machine is semi-sponsored afaik.
<PaulFertser>
Probably if they can just give is 2x RAM for the one we have it should be fine already.
<PaulFertser>
2 GiB is really low by these days' standards :/
<zapb_>
Especially if you run Java applications :D
<zapb_>
PaulFertser, but is there money that allows us to rent an additional or more powerful VM? I have no clue about how pays for the infrastructure
<PaulFertser>
zapb_: this VM is paid from donation money but the hoster made it twice as powerful or something.
<PaulFertser>
Whatever needs to be done needs to be done anyway. If the hoster is not "cooperative" there should be a way to use some other free software friendly resource for CI.
<PaulFertser>
I do not know what exactly the state of OpenOCD financing these days.
tsal has joined #openocd
<zapb_>
Hm, okay. Not the best situation :D
d_olex has joined #openocd
balrog has quit [Ping timeout: 265 seconds]
balrog_ has joined #openocd
tsal has quit [Ping timeout: 265 seconds]
tsal has joined #openocd
<PaulFertser>
I retriggered the rc1 build to see how it behaves under load now.
<PaulFertser>
(I added more swap as I noticed it started to be really close to filling up the existing one fully which is guaranteed to lead to "swap thrashing" and also enabled zswap to trade some i/o wait for sys% CPU cycles which should improve latency when using swap)
<PaulFertser>
(I didn't restart any daemons so what was claimed is still claimed and yet it seems to behave reasonably for now)
nerozero has quit [Ping timeout: 252 seconds]
nerozero has joined #openocd
<borneoa___>
PaulFertser: I was waiting for the build openocd-clang... But Jenkins is again preparing for shutdown and no new builds can be started
<borneoa___>
PaulFertser: Humm, the banner disappeared. I manually triggered the clang build!
<PaulFertser>
borneoa___: I didn't make it prepare for shutdown, hm
<PaulFertser>
Can't see anything of relevance in the logs.