IRC channel logs

2021-03-08.log

back to list of logs

***rekado_ is now known as rekado
<zimoun>hi!
<zimoun>Looking for sysadmin advice. I have a machine (64 cores, 252GB RAM, 2.5TB space) currently running Ubuntu and Guix on the top, obviously. :-) Some people physically access to the machine to run graphical analysis tools. Other users remotely access via SSH. The issue is that time to time, I overload the machine by building / computing a lot of stuff (mainly Guix stuff ;-)), so the machine is lagging and people are complaining (French
<zimoun>people are always complaining ;-)). I am transfering the graphical tool to another machine. What is the best strategy here? Configure a front-end for only one node? Something else?
<civodul>hi there!
<civodul>hmm
<civodul>these could be two separate machines, maybe? :-)
<civodul>otherwise you could choose --max-jobs/--cores so as to always leave room for non-build activity
<zimoun>I am playing around with this -M/-c options but I would prefer soemthing like a container with fixed parameters
<rekado>zimoun: you can use cgroups to reserve/limit resources
<efraim>like running the daemon and builders inside a container/VM so they can only use up to X% of the machine?
<efraim>what rekado said sounds better
<rekado>zimoun: since you’re using Ubuntu you can configure cgroups with systemd: https://wiki.archlinux.org/index.php/Cgroups#With_systemd
<zimoun>cool! Thanks, that’s what I was looking for :-)