#archlinux32 | Logs for 2024-08-09
Back
[08:58:01] -!- drathir_tor has quit [Ping timeout: 260 seconds]
[08:59:41] -!- drathir_tor has joined #archlinux32
[09:13:08] -!- Cthuutloops has quit [Remote host closed the connection]
[09:23:06] -!- drathir_tor has quit [Ping timeout: 260 seconds]
[09:25:17] -!- drathir_tor has joined #archlinux32
[10:41:28] -!- tippfehlr5 has quit [Quit: The Lounge - https://thelounge.chat]
[10:41:46] -!- tippfehlr5 has joined #archlinux32
[11:18:10] -!- abaumann has joined #archlinux32
[11:18:10] <buildmaster> Hi abaumann!
[11:18:10] <buildmaster> !rq abaumann
[11:18:11] <phrik> buildmaster: <abaumann> If it takes long.. it works..
[11:18:48] <abaumann> I simplified the hunspell patches (they didn't work in the build chroot, at least on my build machines). Also some download locations had to be fixed.
[11:19:22] <abaumann> Usually, if the buildmaster schedules packages over and over again and you think, something is wrong, actually - there is nothing else to build and it just schedules the errornous packages..
[11:19:44] <abaumann> I start to think that this is the situation currently.
[11:19:55] -!- abaumann has quit [Client Quit]
[11:51:16] -!- drathir_tor has quit [Ping timeout: 260 seconds]
[11:56:49] <KillerWasp> 1889MB for virtual memory in pulse audio, 996MB for virtual memory for xfce4-terminal, what happen???
[11:58:50] <KillerWasp> I don't understand anything, I have all the memory and swap occupied, and apparently the percentage of occupied memory does not appear in the top command unless I look at the virtual memories.
[11:59:47] <KillerWasp> What are virtual memories exactly? Are they just accumulations of memory leaks?
[11:59:56] <KillerWasp> how to make free it?
[12:00:23] <T`aZ> wat, no
[12:00:51] <T`aZ> tldr: look at rss + swapped size memory, virtual is, virtual, not really that amound is being used
[12:01:13] <T`aZ> amount*
[12:03:56] <KillerWasp> 145MB for xfce4-terminal, and 52MB for Xorg, and all the rest more less, but still i have filled like 2GB in memory, 1GB for ram and swap...
[12:04:16] <KillerWasp> i'm in the limit...
[12:08:13] <KitsuWhooa> abaumann: That is not the case, because when I delete those build assignments, it successfully builds other packages until it encounters another failure again
[12:08:17] <KitsuWhooa> that is the first thing I checked
[12:08:21] <KitsuWhooa> unless we somehow ran out of things
[12:08:44] <KitsuWhooa> and yes, virtual ram is virtual
[12:09:31] <KitsuWhooa> I think it's best to read up an article on it. I don't think any of us (active) here can explain it correctly
[12:11:06] <KillerWasp> then i guess its all leak memory and don't exist in all the process...
[12:11:09] -!- drathir_tor has joined #archlinux32
[12:11:22] <KillerWasp> i must restart the machine
[12:11:51] <T`aZ> it's not leaked, just allocaed or mapped, but not used, and since linux is allocating lazily, it doesnt actually consume the real memory unless needed
[12:11:57] <T`aZ> and most of it is never needed
[12:12:20] <T`aZ> and yes, it's tricky, i tried to find a link, but the first 2 on google did not satisfy me so, i'll let you continue :p
[12:26:17] <KillerWasp> "Reboot Your Linux PC "
[12:26:30] <KillerWasp> the best help that i can found in internet, lol
[12:27:48] <KillerWasp> or stress-ng, i dont understand how burst your memory it can help to free.
[12:29:12] <T`aZ> you cant free virtual memory
[12:32:30] <KillerWasp> T`aZ: i'm not ask by virtual memory, i'm ask by memory and swap.
[14:42:26] -!- abaumann has joined #archlinux32
[14:42:26] <buildmaster> Hi abaumann!
[14:42:26] <buildmaster> !rq abaumann
[14:42:27] <phrik> buildmaster: <abaumann> this is not a database.. this is pure stupidity..
[14:43:25] <abaumann> KillerWasp: virtual memory is completely irrelevant, if a program has a lot of threads, the value can be quite high. virtual memory is the virtual address space the process has allocated. A much better value is RES in top (residential memory)
[14:44:39] <abaumann> KitsuWhooa: quite weird. Some perl modules got built today. Those look ok. Still I have some modules erroring out. 486 is not better, I think, I saw garcon quite some times on the eurnuc-i486 vm.
[14:44:40] -!- abaumann has quit [Client Quit]
[14:56:30] <KillerWasp> abaumann: ok, i understand. About memory and swap I solved it in part with swapoff and swapon. It produced a massive read of more than 1GB on disk and transferred 400MB in memory, it was at the limit and does not seem to have activated OOM Killer. Now it has 1.9GB of ram occupied and it doesn't seem to want to use the swap as much as before, but at least now it's not lagged.
[14:56:55] <KitsuWhooa> you can set vm.swappiness to control that behaviour
[14:59:05] <KillerWasp> I still haven't found anything on the internet about the massive consumption of memory when there are no processes that use it or how to solve the problem.
[14:59:13] <KitsuWhooa> check tmpfs
[14:59:17] <KitsuWhooa> like /tmp
[14:59:29] <KitsuWhooa> but also, are you checking the "free" or "avail" memory
[14:59:45] <KitsuWhooa> because the linux kernel will fill up your ram with cached files to speed up the system
[14:59:51] <KitsuWhooa> and will drop them when the memory is needed for processes
[15:00:29] <KitsuWhooa> s/avail/available/
[15:00:58] <KitsuWhooa> !lucky
[15:00:58] <phrik> KitsuWhooa: You might be able to bribe me for an answer…
[15:01:05] <KitsuWhooa> https://www.linuxatemyram.com
[15:01:06] <phrik> Title: Help! Linux ate my RAM! (at www.linuxatemyram.com)
[15:01:15] <KitsuWhooa> I like that website :p
[15:01:43] <KillerWasp> KitsuWhooa: for caches i already send values "1", "2", and "3" in /proc/sys/vm/caches
[15:03:30] <KillerWasp> also with 'sync'
[15:12:08] <KillerWasp> why dont exist vm.wipemem?
[16:28:20] <KitsuWhooa> Hm, the build log links fail with " No logs found in the mysql database for this package."
[17:20:57] -!- hexagenic has joined #archlinux32
[17:52:28] -!- morriset has joined #archlinux32
[17:53:41] -!- Cthuutloops has joined #archlinux32
[18:14:16] -!- bdju has quit [Ping timeout: 272 seconds]
[18:24:23] -!- bdju has joined #archlinux32
[18:52:02] <girls> KitsuWhooa: then look into /srv/http/build-logs/error directly on the buildmaster
[18:52:16] <girls> as I saw before, the db seems to have some memory gaps
[18:52:21] <KitsuWhooa> Oh yeah I know how to access the logs
[18:52:29] <girls> and regarding the "nothing more to build": this can well be.
[18:52:32] <girls> ah, ok
[18:52:35] <KitsuWhooa> It's just that even the latest links in devops are broken
[18:52:39] <KitsuWhooa> Thanks though :)
[18:53:01] <KitsuWhooa> And I remember it working :p
[18:53:08] <girls> if there's a cycle, the buildmaster tries to break it, but if all packages fail, I think, it might not notice and still try to build all of them again and again
[18:53:39] <girls> if you remove the cycle packages, there may well be more to-be-built packages dependent on cycle packages (in the view of the buildmaster)
[18:53:48] <girls> those would then of course be scheduled and may even build
[18:54:28] <girls> regarding "some stuff got built" - that's probably on me: I force-scheduled *all* packages on Streusel-1, Streusel-2 and Streusel-3 to unblock stuff
[18:54:44] <girls> those packages probably need to be rebuilt some time later again to fix broken dependencies
[18:54:54] <girls> but the auto-seeder should(tm) notice this
[19:01:00] <girls> and again, the upload of the fail logs did not work - strange
[19:01:51] <girls> ah, btw, abaumann: thanks for fixing my PKGBUILD patches o:-)
[19:05:25] -!- hexagenic has quit [Ping timeout: 248 seconds]
[19:06:17] -!- hexagenic has joined #archlinux32
[19:12:04] <girls> somehow file I/O on the buildmaster is really slow
[19:13:46] <girls> listing 6k build logs of the last 5 years (which is a lot, but not "that much") takes almost a minute
[19:25:20] <T`aZ>
[19:25:32] <T`aZ> ww, sorry
[19:25:49] <girls> did you just post your password? ;)
[19:26:16] -!- drathir_tor has quit [Ping timeout: 260 seconds]
[19:26:24] <T`aZ> nah, my root password is root
[19:26:29] <girls> smart
[19:26:31] <T`aZ> so im fine
[19:26:37] <girls> phew, close one
[19:35:12] <girls> I think the mysql db is just garbage: I can insert rows into "failed_builds" table, but they never appear. I even get duplicate key violations, when I insert them twice, but selecting for them yields nothing. ... or - which is totally impossible - I'm doing something stupid. :-/
[19:36:58] <girls> how can I get "duplicates" if there is no unique key constraint on that table and the id is not provided but autogenerated?
[19:38:46] <girls> AHA! "ERROR 167 (22003): Out of range value for column 'id' at row 1"
[19:39:17] <girls> select max(id) from failed_builds;
[19:39:17] <girls> +---------+
[19:39:18] <girls> | max(id) |
[19:39:18] <girls> +---------+
[19:39:18] <girls> | 8388600 |
[19:39:19] <girls> +---------+
[19:39:27] <girls> doesn't look too 2^n-ish to me
[19:42:13] <girls> we now have a bigint as id - let's see, whether that magically makes things work :D
[19:47:32] <girls> (it was mediumint before)
[19:49:25] <girls> woohoo - we have fail logs in the db (again) :)
[19:56:08] <girls> Ah, now I see: it tries to insert 7 rows, which fails, because it would put the index to 2^23-1 and presumably the variable, where the "next to be inserted" index is stored to 2^23, thus to something, that is not storable in a 3 byte mediumint (c.f. https://dev.mysql.com)
[19:57:13] <girls> 2^23 ist just a too odd number for me to recognize (although it is as even as it could be)
[19:57:53] -!- drathir_tor has joined #archlinux32
[20:13:40] <girls> /builder/bin/build-packages: line 582: core-staging-with-build-support-pentium4-build: command not found
[20:13:48] <girls> does this no longer come with devtools32?
[20:14:09] <girls> devtools32 20221208-1 looks a bit outdated, too
[20:14:21] <girls> where do I find a recent one? apparently not in x86_64/releng :(
[20:47:09] <girls> we could use our builders to also update x86_64/releng packages, hmmm
[21:01:55] <girls> or I could just fix my scripts and continue being the single point of failure, here :D
[21:05:26] -!- drathir_tor has quit [Ping timeout: 260 seconds]
[21:15:59] -!- drathir_tor has joined #archlinux32
[21:24:12] -!- hexagenic has quit [Quit: WeeChat 4.2.2]
[21:58:49] <KitsuWhooa> Ooooh
[21:58:58] <KitsuWhooa> I thought it might be an int issue but I didn't think to check that table
[21:59:01] <KitsuWhooa> Nice catch!
[21:59:54] <girls> no idea, why we put mediumint on that table in the first place
[22:06:51] <KitsuWhooa> oh yeah the devtools package is long broken
[22:06:57] <KitsuWhooa> I've always installed from source
[22:07:05] <KitsuWhooa> plus or minus the hack I had to make to have the builders work inside lxc
[22:07:09] <KitsuWhooa> or was that in builder
[22:07:11] <KitsuWhooa> I can't remember
[22:07:52] <KitsuWhooa> but yeah I wonder if there are other ids that are not bigints that are causing issues