Skip to content

Conversation

@hawkw
Copy link
Member

@hawkw hawkw commented Jan 30, 2026

Presently there is a shortage of servers to run Buildomat workers for CI jobs. Fortunately, we work at a company that makes servers. So we can just make more servers. And then we can use those servers for CI. And then everyone's happy.

In order to do that, we must use the lab-2.0-gimlet Buildomat target rather than helios-2.0 or lab-2.0-opte targets. It's possible that some or all of Omicron's CI cannot easily do that due to special-cased behavior for the i86pc architecture to detect running in a dev environment. Luckily, no such behavior exists in Propolis, so (hopefully!) we can just switch our Buildomat jobs to the lab-2.0-gimlet target and (hopefully) it will Just Work. This way, we're not competing with Omicron builds for the relatively scarce non-Gimlet workers.

@hawkw hawkw requested review from iximeow and jclulow January 30, 2026 02:23
Presently there is a shortage of servers to run Buildomat workers for CI
jobs. Fortunately, we work at a company that makes servers. So we can
just make more servers. And then we can use those servers for CI. And
then everyone's happy.

In order to do that, we must use the `lab-2.0-gimlet` Buildomat target
rather than `helios-2.0` or `lab-2.0-opte` targets. It's possible that
some or all of Omicron's CI cannot easily do that due to special-cased
behavior for the `i86pc` architecture to detect running in a dev
environment. Luckily, no such behavior exists in Propolis, so
(hopefully!) we can just switch our Buildomat jobs to the
`lab-2.0-gimlet` target and (hopefully) it will Just Work. This way,
we're not competing with Omicron builds for the relatively scarce
non-Gimlet workers.
@hawkw hawkw force-pushed the eliza/use-our-computers branch from 89243be to a47c9b8 Compare January 30, 2026 02:27
@hawkw
Copy link
Member Author

hawkw commented Jan 30, 2026

Hmm, seems the phd-run tests do not, in fact, work out of the box on lab-2.0-gimlet targets. It seems that they are trying to create a zpool with hard-coded device names that don't exist on the Gimlet target:

34	2026-01-30T02:49:55.341Z	starting task 2: "build"
35	2026-01-30T02:49:55.367Z	+ indir=/input
36	2026-01-30T02:49:55.367Z	+ indir_suffix='phd-build/out/*.tar.gz'
37	2026-01-30T02:49:55.367Z	+ phddir=/work/phd-test
38	2026-01-30T02:49:55.367Z	+ pfexec zpool create -f phd-artifacts c1t1d0 c2t1d0
39	2026-01-30T02:49:55.393Z	cannot open 'c1t1d0': no such device in /dev/dsk
40	2026-01-30T02:49:55.393Z	must be a full path or shorthand device name
41	2026-01-30T02:49:55.393Z	process exited: duration 36 ms, exit code 1

This is probably fixable, but will need to figure out what the names of devs on the lab-2.0-gimlet target are.

Copy link
Member

@iximeow iximeow left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@hawkw hawkw merged commit 40dbfad into master Jan 30, 2026
12 checks passed
@hawkw hawkw deleted the eliza/use-our-computers branch January 30, 2026 20:15
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants