Sunday, July 01, 2018
grimoire@muc.metronome.im
July
Mon Tue Wed Thu Fri Sat Sun
            1
2
3
4 5 6 7 8
9 10 11
12 13 14
15
16 17 18
19 20
21 22
23 24 25 26 27 28 29
30 31
         
Metronome IM "cook" book room | Support and Dev talk about the Metronome XMPP server.

Latest stable version build is: 4.0.3

For more information you can visit: https://metronome.im

Room logs can be found here: https://muc.metronome.im/logs/grimoire/

* Links *
<==========================================================>
- Building & Installation:
https://metronome.im/building
- Documentation:
https://metronome.im/documentation
- Issues Tracker:
https://github.com/maranda/metronome/issues
<==========================================================>

[09:12:39] <zinid> Maranda, what is the machine's configuration running metronome.im?
[09:25:24] <zinid> http://upload.zinid.ru/99bb0647172ef83f67aa892f2cd0f2fa9d04f206/4RsAAjasFztwnKLFYEdtxhTqSinYXWOgVtkT2USH/c2s-sessions.png
[09:25:34] <zinid> struggle!
[10:05:25] <Maranda> zinid: hmm?
[10:05:38] <Maranda> zinid: what do you mean?
[10:09:46] <Maranda> zinid, pay in mind that it's a VPS so cpu processes are shared unluckily, depending on the load it may have performance issues
[10:09:54] <Maranda> zinid, what did you test?
[10:13:23] <zinid> I'm testing on localhost, it's very slow machine here
[10:13:31] <zinid> I don't test your server 🙂
[10:13:59] <Maranda> zinid, you managed to install it easily?
[10:14:42] <zinid> I test ejabberd
[10:14:49] <zinid> no, I didn't install metronome
[10:15:09] <Maranda> Oh sorry I didn't pick up what the struggle was about XD
[10:15:13] <zinid> I just wanted to show how hard it's trying to establish connections 😀
[10:15:29] <zinid> it's failing, but it's trying harder!
[10:15:54] <zinid> anyway, ejabberd is not dead, just slow and spitting database timeouts from time to time
[10:16:15] <Maranda> zinid, I'm not sure maybe you should raise that timeout though 10 seconds may be a bit low on a heavy load.
[10:16:31] <Maranda> zinid, that's what caused most of the "cancer" me thinks
[10:17:15] <Maranda> because if the drones start the test as soon as they connect they will already drown the server of request
[10:17:24] <Maranda> but not sure if that's the purpose of the test
[10:18:35] <Maranda> zinid, normally a client wouldn't spam so much stuff on connect. Or if it does it's *once*.
[10:21:12] <zinid> Maranda: well, normally clients will actually create an avalanche, and the majority of clients are poorly written and don't have exponential back off and will retry every 30 seconds or so
[10:21:27] <zinid> So my scenario is pretty much optimistic
[10:23:12] <Maranda> zinid, avalanche effect sure, I'm not sure if they will go active and inactive multiple times after logining triggering multiple csi flushes
[10:24:14] <Maranda> I currently don't put queue/flush rate limits there so maybe that should be optimised