We now have 50 machines in our shared spread pool.
@fgimenez Can you please bump all systems by one more worker so we can test the waters with 50% more systems? If all goes well, we can add some more machines to the pool and yet another worker to every system again, doubling the current workers per system.
Sure, this is the PR https://github.com/snapcore/snapd/pull/3293, we should see improvements in that build.
Unfortunately right now master is broken. I think @zyga-snapd knows what this is about (something about a branch from @pstolowski landing at an inopportune time).
Thanks @chipaca! Master is fine again and the PR above (3293) got merged this morning.
Thanks Federico. Per discussion elsewhere, it looks like the builds have dropped to about 28 minutes. I’ve bumped it up to 70 machines now, so let’s please go ahead and add the extra worker on all systems and observe again.
Related errors or flakiness that needs fixing?
Not related, it seems that a new ubuntu-image version is breaking the tests https://travis-ci.org/snapcore/snapd/builds/231151352#L1235
I see… please drop a note here once the problem is sorted and we have a more clear idea of the actual run times.
The ubuntu-image problem is sorted now. The ubuntu-image snap moved from regular confinement to classic confinement. This caused our spread tests to stop working because
snap install --devmode ubuntu-image no longer works. We switched our tests to use
snap install --classic ubuntu-image and things are working again. The downside of the new confinement of course is that ubntu-image will no longer work on pure ubuntu core systems.