go/botmap shows that it is a bot on "chromium.gpu.fyi" waterfall: https://chromium.googlesource.com/chromium/tools/build/+/438f445e28e5dd6dd97f0ac7a3d7a39384f06fca/masters/master.chromium.gpu.fyi/builders.pyl#896

https://chrome-internal.googlesource.com/infradata/hosts/+/master/botmap.md

No idea why, probably a mistake. Passing to kbr@, the "cros-*-c2" slaves are generally reserved for ChromeOS project waterfalls, weird seeing this slave there. Any idea what's up?

It's not a mistake. It was added in  Issue 666077  in order to allow running ChromeOS tests on the GPU waterfall.

If there are bookkeeping issues with this bot then let's fix them.

Well, the cros-*-c2 bots are just generic GCE systems that are managed by the CrOS project. Is there a reason a standard ccompute bot wouldn't work?

More specifically, paid for by the CrOS project instead of the Chrome project.  That's really the only difference.

Well, it may also have a different ACL set.

Perhaps I'm confused about what this bot was intended to do. Frank, can you work with the infra team to clarify this?

I just found documentation of wimpy8 being used for the chromium.chromeos waterfall. I suspect the right thing is just to update those docs to chromium.gpu.fyi.

The following revision refers to this bug:
  https://chrome-internal.googlesource.com/infra/infra_internal.git/+/9ffa63a0815c22f00bdb6052cea5897458cb99e6

commit 9ffa63a0815c22f00bdb6052cea5897458cb99e6
Author: Don Garrett <dgarrett@google.com>
Date: Mon Jan 30 23:11:07 2017