Re: [PATCH v6 00/25] xl / libxl: named PCI pass-through devices

[Wei Liu <wl@xxxxxxx>]

On Wed, Dec 09, 2020 at 06:41:03PM -0800, Stefano Stabellini wrote:
> On Wed, 9 Dec 2020, Stefano Stabellini wrote:
> > On Wed, 9 Dec 2020, Wei Liu wrote:
> > > On Tue, Dec 08, 2020 at 05:02:50PM -0800, Stefano Stabellini wrote:
> > > > The pipeline failed because the "fedora-gcc-debug" build failed with a
> > > > timeout: 
> > > > 
> > > > ERROR: Job failed: execution took longer than 1h0m0s seconds
> > > > 
> > > > given that all the other jobs passed (including the other Fedora job), I
> > > > take this failed because the gitlab-ci x86 runners were overloaded?
> > > > 
> > > 
> > > The CI system is configured to auto-scale as the number of jobs grows.
> > > The limit is set to 10 (VMs) at the moment.
> > > 
> > > https://gitlab.com/xen-project/xen-gitlab-ci/-/commit/832bfd72ea3a227283bf3df88b418a9aae95a5a4
> > > 
> > > I haven't looked at the log, but the number of build jobs looks rather
> > > larger than when we get started. Maybe the limit of 10 is not good
> > > enough?
> > 
> > Interesting! That's only for the x86 runners, not the ARM runners (we
> > only have 1 ARM64 runner), is that right?
> > 
> > If we could increase the number of VMs for x86 I think that would be
> > helpful because we have very many x86 jobs.
> 
> I don't know what is going on but at the moment there seems to be only
> one x86 build active
> (https://gitlab.com/xen-project/patchew/xen/-/pipelines/227280736).
> Should there be at least 3 of them?

Not sure what you meant here. That pipeline is green.

It may take some time for the CI to scale up if it is "cold". By default
there is only 1 standby runner to reduce cost.

Wei.