#archlinux32 | Logs for 2018-08-20

[00:00:59] <buildmaster> gst-editing-services is broken (says rechenknecht).
[00:09:12] -!- JayDubb has joined #archlinux32
[01:55:11] <buildmaster> python-faker is broken (says rechenknecht).
[02:25:37] -!- NoobAlice has quit [Ping timeout: 268 seconds]
[02:30:55] -!- NoobAlice has joined #archlinux32
[02:35:45] <buildmaster> linux is broken (says rechenknecht).
[02:42:36] <buildmaster> linux-zen is broken (says rechenknecht).
[04:18:18] -!- JayDubb has quit [Ping timeout: 276 seconds]
[05:10:31] <buildmaster> julia is broken (says nlopc43).
[05:19:57] -!- davor has quit [Ping timeout: 260 seconds]
[05:20:34] -!- davor has joined #archlinux32
[05:30:24] -!- davor_ has joined #archlinux32
[05:30:28] -!- davor has quit [Ping timeout: 272 seconds]
[05:30:30] davor_ is now known as davor
[07:01:08] -!- oaken-source has joined #archlinux32
[07:06:36] <buildmaster> spice-gtk is broken (says rechenknecht).
[07:38:59] -!- abaumann has joined #archlinux32
[07:39:00] <buildmaster> Hi abaumann!
[07:57:33] -!- oaken-source has quit [Ping timeout: 244 seconds]
[08:00:22] -!- NoobAlice has quit [Quit: Leaving.]
[08:00:31] <abaumann> tyzoid: mmh, on the buildmaster I see: Mem: 988 (1GB) and tmpfs 506136 454876 51260 90% /tmp
[08:00:48] <buildmaster> grafana is broken (says nlopc43).
[08:00:52] <abaumann> Is it just me, but I would put /tmp on real hard disks to avoid swapping and trashing the buildmaster..
[08:23:20] -!- eduardoeae_ has quit [Ping timeout: 276 seconds]
[09:24:46] * buildmaster failed to execute a mysql query - can you have a look at "tmp.mysql-functions.query.stdin.2018-08-20T09:24:43.xjqEBk"?.
[09:40:44] <abaumann> bad buildmaster :-)
[10:25:00] <buildmaster> virtualbox-modules-arch is broken (says nlopc43).
[10:54:20] <abaumann> mmh. git-clean and sanity check take a while..
[12:24:02] * buildmaster goes insane.
[12:52:30] <abaumann> *grrr*
[14:16:23] * buildmaster goes insane.
[14:20:34] <abaumann> yeah. sanity-checks and git-clean (gc) are not friends and should lock each other :-)
[14:26:34] * buildmaster goes insane.
[14:29:18] -!- abaumann has quit [Ping timeout: 252 seconds]
[14:37:17] -!- oaken-source has joined #archlinux32
[14:43:31] -!- eduardoeae_ has joined #archlinux32
[14:49:20] <tyzoid> abaumann: I don't run the buildmaster on my servers, IIRC it's deep42thought administering that on rewby's servers.
[15:30:28] -!- abaumannn has joined #archlinux32
[15:30:49] <abaumannn> tyzoid: yes, I'm aware of that. deep42thought gave me ssh access yesterday.
[15:31:10] <abaumannn> now the buildmaster is angry at me and is sulking because deep42thought is in holydays.
[15:31:17] <tyzoid> abaumannn: Then I must be misunderstanding your question
[15:31:28] <tyzoid> Are you asking what I would do?
[15:31:36] <abaumannn> Yes, basically. :-)
[15:31:48] <tyzoid> What's the current swap usage?
[15:31:51] <abaumannn> I would not put mysql to a /tmpfs, especially not, if I'm low on memory already.
[15:32:12] <tyzoid> I would recommend not putting mysql in tmpfs, esp. since it caches anyway
[15:32:30] <abaumannn> yes. and memory can also be used for file system caches again.
[15:32:43] <tyzoid> but what's the current swap usage on that system?
[15:32:58] <abaumannn> currently I have a git gc going hard with I/O on the machine.
[15:33:04] <tyzoid> ah
[15:33:05] <abaumannn> now zero.
[15:33:11] <tyzoid> Oh, interesting.
[15:33:18] <abaumannn> I had to stop mysql and restarted it with Tmpdir /var/tmp
[15:33:25] <tyzoid> then yeah, making /tmp operate on hdd would speed things up
[15:33:47] <tyzoid> IIRC it's a VM, so it likely has some disk caching there somewhere anyway
[15:33:55] <abaumannn> yes.
[15:35:08] <abaumannn> I get a feeling that also git gc in clean-git and the sanity-check script block each other, when they access the same git repository, that is
[15:35:41] <abaumannn> let's wait till this compacting is over. then I can restore the build jobs..
[16:34:40] <abaumannn> mmh. 99% I/O and < 5M/s I/O transfer rates.. what's exactly fishy here?
[16:35:51] -!- abaumannn has quit [Quit: Page closed]
[16:36:08] <tyzoid> abaumann: What's iotop got to say about it?
[16:36:42] <tyzoid> Could be other vms using the disk slowing things down
[16:38:38] -!- oaken-source has quit [Ping timeout: 272 seconds]
[16:55:58] -!- Alina-malina has quit [Ping timeout: 260 seconds]
[17:33:15] -!- abaumann has joined #archlinux32
[17:33:20] <buildmaster> Hi abaumann!
[17:33:29] <abaumann> tyzoid: those numbers are from iotop. :-)
[17:37:36] <tyzoid> ah
[17:37:48] <tyzoid> well then
[17:38:00] <tyzoid> Could be a lot of seeks
[17:38:12] <tyzoid> which would make sense if the disk cache is too small
[17:42:13] <abaumann> 17337 slave 20 0 579264 487348 32816 D 5.0 48.1 4:22.90 git
[17:42:21] <abaumann> I see hich SWAPIN.
[17:42:31] <abaumann> So I try to shrink tmpfs
[17:42:44] <abaumann> ..on the running system :->
[17:47:06] <abaumann> So. tmpfs is auto-created by systemd and consumes half of the available memory, sweet :-)
[17:47:30] <abaumann> now swapon is 0%
[17:47:34] <abaumann> *swapin
[17:47:51] <abaumann> ..and the system is actually responsive. :-)
[17:52:19] <tyzoid> nice
[17:52:29] <tyzoid> btw, pivot_root is a sweet thing
[17:52:32] <abaumann> how can you specify the size of tmpfiles.d?
[17:52:59] <abaumann> pivot_root, yes :-)
[17:53:17] <tyzoid> not sure
[17:54:24] <abaumann> https://unix.stackexchange.com
[17:54:26] <phrik> Title:disk - Systemd backed tmpfs | How to specify /tmp size manually - Unix & Linux Stack Exchange (at unix.stackexchange.com)
[17:54:27] <abaumann> oi!
[17:54:53] <zer0def> yeah, i was about to propose overriding tmp.mount options
[17:55:11] <abaumann> *abaumann sighs, why not /etc/fstab?
[17:56:32] <zer0def> i agree, although i remember having *two* overlaying tmpfs mounts on /tmp when using fstab, the fstab one being mounted later and taking preference.
[17:56:52] <abaumann> ah, thanks zer0def. :-)
[17:57:11] <zer0def> with that said, i still have those entries and `mount | grep /tmp` lists one record atm
[17:57:31] <zer0def> so /shrug
[17:58:10] <zer0def> given there's no escaping systemd, might as well have notes on both options for the future
[17:58:25] <abaumann> like the 'non escaping' part :-)
[17:58:50] <zer0def> i don't, it's herd mentality
[17:59:07] <abaumann> * abaumann mistipped the smiley -> :->
[17:59:44] <abaumann> "In general, configuring mount
[17:59:44] <abaumann> points through /etc/fstab is the preferred approach. See systemd-fstab-generator(8) for details
[17:59:47] <abaumann> about the conversion.
[17:59:49] <abaumann> "
[17:59:54] <abaumann> from the systemd.mount man page :-)
[18:00:17] <zer0def> time to view the systemd pkgbuild and see whose addition that one is
[18:00:27] <abaumann> So, maybe only things like /dev/shm should be moved there in order not to confuse users in /etc/fstab, /tmp belongs in /etc/fstab iIMHO
[18:01:42] <zer0def> ok, tmp.mount is upstream
[18:01:47] <abaumann> I start to think 'git gc' takes longer than a fresh surface clone
[18:02:32] <zer0def> in which case, why provide it unless someone insists on being schizophrenic
[18:02:39] <abaumann> :-)
[18:03:49] <zer0def> and i don't care enough to look for reasoning… yet; although i can imagine the excuse would be "we needed tmpfs before fstab for X situation"
[18:05:20] <zer0def> btw, figured i'll ask - any of you used or is using coreos' grub for tpm-based measured/trusted boot?
[18:05:48] <abaumann> nope
[18:11:58] <zer0def> looks like the closest thing to having a fully verified boot with encrypted /boot and /root
[18:12:10] <zer0def> erm, s/root$//
[18:13:42] <zer0def> that said, 32-bit efi wasn't a consideration in that fork
[18:55:52] -!- Alina-malina has joined #archlinux32
[19:16:20] -!- Cthulu201 has quit [Ping timeout: 272 seconds]
[19:22:06] -!- abaumann has quit [Quit: leaving]
[19:23:56] -!- woshty has quit [Ping timeout: 272 seconds]
[19:24:22] -!- woshty has joined #archlinux32
[19:39:18] -!- Cthulu201 has joined #archlinux32
[20:16:24] -!- abaumann has joined #archlinux32
[20:16:24] <buildmaster> Hi abaumann!
[20:53:43] * buildmaster resumes sanity.
[20:56:02] <abaumann> good boy :-)
[21:01:32] <abaumann> sorry, but this vm is just way out of I/O bandwith!
[21:02:26] <abaumann> but at least mysql is not filling up the tmp anymore now during sanity checks.
[21:02:51] <abaumann> not sure, if this all a good thing, it may be quite some slower now, but at least - so I hope - more stable.
[21:03:24] <abaumann> sanity-check running: mmh, again. How often is this triggered, I wonder?
[21:04:18] <buildmaster> linux is broken (says nlopc43).
[21:05:02] <abaumann> yea yea, I know.
[21:05:15] <abaumann> * abaumann starts to doubt his mental sanity talking to a irc robot.
[21:07:34] <buildmaster> linux-zen is broken (says rechenknecht).
[21:12:13] <tyzoid> lol
[21:12:48] <tyzoid> Only other idea is to offload things from that box
[21:12:58] <tyzoid> i.e. split off non-critical tasks
[21:13:18] <tyzoid> like perhaps moving the irc bot from buildmaster onto the slave server on my system?
[21:15:10] <abaumann> the irc bot is hardly a resource consuming component. :-)
[21:15:23] <abaumann> let's see, at the moment it's tucking along nicely.
[21:15:36] <abaumann> I thing, the main problem was the 'git gc' running in parallel with the buildmaster.
[21:16:18] <abaumann> *think
[21:16:36] <abaumann> git repos of that size are really a problem to handle.
[21:17:24] <abaumann> mmh. mysql is.. well mysql.
[21:17:41] <abaumann> some finds and rsyncs.
[21:18:23] <abaumann> I really don't want to tear things apart on my first day as buildmaster deputy. :-)
[21:18:35] <abaumann> *deputee
[21:23:59] <abaumann> so, sleep well (not you buildmaster, you work!) :-)
[21:24:01] -!- abaumann has quit [Quit: leaving]
[21:24:22] <tyzoid> lol, you spelled it right the first time.
[21:54:42] -!- rcf1 has joined #archlinux32
[21:55:29] -!- rcf has quit [Ping timeout: 256 seconds]
[22:08:58] -!- City-busz has quit [Quit: No Ping reply in 180 seconds.]
[22:10:04] -!- elibrokeit has quit [Ping timeout: 256 seconds]
[22:10:15] -!- elibrokeit has joined #archlinux32
[22:23:27] -!- oaken-source has joined #archlinux32
[23:00:59] <buildmaster> libffado is broken (says nlopc43).
[23:07:47] -!- oaken-source has quit [Ping timeout: 255 seconds]