Skip to content

scale up our ci-runner-x86-03 and possibly use ci-runner-x86-14

we seem to be under-using our runners dramatically. just look at those memory usage graphs:

image

https://grafana.torproject.org/d/amgrk2Qnk/memory-usage?orgId=1&from=now-24h&to=now&timezone=browser&var-class=%24__all&var-node=ci-runner-x86-02.torproject.org&var-node=ci-runner-x86-03.torproject.org&var-node=ci-runner-x86-14.torproject.org&refresh=auto

at least -02 has the decency of almost (but not entirely!) using up its RAM for caching (the red area), but on -03, that barely makes a dent in the amazingly large 1.5TB capacity.

ci-runner-x86-14 is just completely unused at the moment as well, look at this CPU usage graph:

image

https://grafana.torproject.org/d/gex9eLcWz/cpu-usage?orgId=1&from=now-24h&to=now&timezone=browser&var-class=%24__all&var-node=ci-runner-x86-02.torproject.org%3A9100&var-node=ci-runner-x86-03.torproject.org%3A9100&var-node=ci-runner-x86-14.torproject.org%3A9100&var-show_cpu_count=or&refresh=auto

clearly, someone needs to use -14 for something. @morgan would the apps team be open to sharing that fancy new server we setup for you, or do you plan on using it soon? :)

and perhaps we could raise concurrency a bit on -02? i guess that depends on whether that would break shadow sims... @jnewsome, do you think we could run more jobs on -03?

memory pressure is decent on all three:

image

... although it is quite busy on -02 there, and it spiked above our threshold on -03 once today. but nothing alarming.

finally, note that disk I/O is getting close to being an issue on -03:

image

https://grafana.torproject.org/d/zbCoGRjnz/disk-usage?from=now-24h&to=now&timezone=America%2FToronto&var-class=%24__all&var-node=ci-runner-x86-02.torproject.org%3A9100&var-node=ci-runner-x86-03.torproject.org%3A9100&var-node=ci-runner-x86-14.torproject.org%3A9100&var-show_cpu_count=or&orgId=1&var-instance=ci-runner-x86-02.torproject.org&var-instance=ci-runner-x86-03.torproject.org&var-instance=ci-runner-x86-14.torproject.org&var-Filters&refresh=auto

we're getting close to the SATA 6gbps limit here, which will influence how/if we replace the drives on that machine (#42274 ).

i estimate wee could double concurrency on -03 without ill effects.

Edited by anarcat
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information