Bigger Public Runners on GiHub

[Jarek Potiuk <ja...@potiuk.com>]

Hello here,

We've discussed recently on the possibility of enabling bigger public
runners for GitHub Actions. I recall that it was supposed to be raised
at a meeting with GitHub. We would very much like to start using
bigger runners - we started to experience much more frequent
intermittent failures and all traces lead to OOM (it's difficult to
track sometimes because sometimes OOM leads to no logs from the
instance - but this is most likely culprit).

I plan to look for more optimisations (which I would like to do
regardless, but this might be difficult as I think we pretty much
squeezed everything out of those instances) . I wanted to see if there
is a possibility we could have bigger instances?

As mentioned before - this will not really lead to bigger costs for
GitHub (at least in our case) - we have optimisations that allow us to
utilise more CPUS and memory (very well tested in our self-hosted
runners) and this will lead to bigger instances needed for less time
and greater stability (i.e. less reruns which we now have to do often)
so overall it might lead to decrease in cost that GitHub has to incur
sponsoring the ASF.

Looking forward to trying it out when we can :). The moment we will be
able to do it, I am able to switch the gears for all our builds pretty
much immediately - our builds self-adjust parallelism to the size of
the machine they run on :)

J.