Re: sbuild race with sporadic build breakages

["'MOESSBAUER, Felix' via isar-users" <isar-users@googlegroups.com>]

On Mon, 2024-11-11 at 20:48 +0100, Jan Kiszka wrote:
> On 11.11.24 18:58, Moessbauer, Felix (FT RPD CED OES-DE) wrote:
> > On Mon, 2024-11-11 at 16:56 +0100, Jan Kiszka wrote:
> > > On 11.11.24 16:17, Anton Mikanovich wrote:
> > > > 11/11/2024 16:52, Jan Kiszka wrote:
> > > > > We could only lock known sessions against each other. But
> > > > > that
> > > > > would
> > > > > already help everything that runs in a CI runner, maybe in a
> > > > > kas-
> > > > > isar
> > > > > container. Not perfect but would probably address 99% of
> > > > > cases
> > > > > today.
> > > > > 
> > > > > Jan
> > > > > 
> > > > This will result in serialization of do_dpkg_build tasks and
> > > > will
> > > > cause
> > > > Isar
> > > > slowdown.
> > > > 
> > > 
> > > OK, that was the information I was looking for - and was afraid
> > > of.
> > > 
> > > But even if your PR would be merged tomorrow (which is completely
> > > unpredictable right now), it would still require backports to
> > > bookworm
> > > at least, and those would need to be deployed locally or
> > > integrated
> > > into
> > > future versions of kas. All that is not going to happen soon - if
> > > at
> > > all. It may even be that only trixes schroot will have all what
> > > we
> > > need.
> > 
> > We have similar problems with apt and proper rate-limit handling.
> > But
> > that's OT.
> > 
> > At a second thought I'm wondering if we should drop schroot at all
> > and
> > use mmdebstrap + unshare. By that, the sbuild also does no longer
> > require root access. I'm running sbuild with unshare backend for
> > some
> > time locally now (not with ISAR, though).
> > 
> 
> https://lists.debian.org/debian-devel/2024/06/msg00257.html - there
> is
> likely no simple answer for the backend yet.

For me that reads like unschroot is just a unshare backend on steroids.
But as we can't use that prior to trixie, it probably does not help
here. Still I see no reason that blocks us from using the unshare
backend.

> 
> BTW, would unshare work at all inside our build containers?

Yes, that works as namespaces can be stacked - that's how podman user
(non root) containers work.

Felix

> 
> Jan
> 

-- 
Siemens AG, Technology
Linux Expert Center


-- 
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/isar-users/e18b3a8ec7eeb5f894a368b1558884fe73092573.camel%40siemens.com.