Opened 7 months ago

Last modified 6 months ago

#32920 new task

rebuild or replace scaleway boxes (scw-arm-*)

Reported by: anarcat Owned by: weasel
Priority: Medium Milestone:
Component: Internal Services/Service - jenkins Version:
Severity: Normal Keywords:
Cc: Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

the boxes we have at scaleway (scw-arm-ams-01 and scw-arm-par-01) are "old" according to scaleway. this leads to various problems, documented in their ticket tracker:

  • #WJVJ-9718-KIPG - machines do not reboot unattended
  • #GHYX-4885-QWLA - kernel not used

Their newer platform allows us to run our own kernels, and also handles reboots better. Right now scq-arm-ams-01 is down because I can't figure out how to reboot it properly, for example.

This would require reinstalling the machines and setting them up as build boxes again.

Child Tickets

TicketStatusOwnerSummaryComponent
#33001closedanarcatdecomission scw-arm-ams-01Internal Services/Service - jenkins

Change History (4)

comment:1 Changed 7 months ago by anarcat

Their newer platform allows us to run our own kernels, and also handles reboots better.

Actually, scw-arm-ams-01 *is* running their newer platform, so rebuilding one of those might not fix the problem at all.

It's currently down, still, and I'm dealing with scaleway to try to get that fixed. I've been talking with them all week and so far it's not going great.

comment:2 Changed 7 months ago by anarcat

this just in: scw-arm-ams-01 is still down and scaleway says they can't recover it, and suggest reinstalling. i'll just decom the machine for now so that we at least save the costs here, but we should consider another place to host those VMs from here on, because that's really just unacceptable.

comment:3 Changed 7 months ago by anarcat

i decomissioned scw-arm-ams-01 in #33001 and closed ticket WJVJ-9718-KIPG with scaleway. the other ticket is still open because scw-arm-par-01 still doesn't boot the right kernel, but it might be able to reboot unattended, unclear.

in any case, we might want to shift all this away from this scaleway platform... but that will mean dropping the 32-bit ARM infrastructure, as scw-arm-par-01 is a armv7l which we don't have an equivalent for elsewhere right now. weasel was looking for an excuse to drop those anyways, but I'll leave him with that decision.

comment:4 Changed 6 months ago by weasel

build-arm-10.torproject.org is now building arm64 jenkins stuff.

Note: See TracTickets for help on using tickets.