Remove site_metahost_scheduler logic, it is broken.
The site_metahost_scheduler does more harm than it does good.
It creates issues where hosts are magically left at READY and Autotest
never tries to schdule on them, even if they come back after not being
pingable.
If Autotest cannot try to schedule on every host you end up with a weird
situation. The scheduler thinks those hosts are available and the frontend
shows them as available but they are broken. The only way to figure out they
are bad is by way of a third party "health dashboard" which isn't very useful
for the scheduler.
TEST=Ran jobs on test server with a few hosts some pingable some not. Ensured
that Autotest still tried to schedule on downed hosts and eventually marked
them as repair failed.
BUG=chromium-os:27794
Change-Id: If4abafc2a29c15ad09a4d4f61a985d349df36e12
Reviewed-on: https://gerrit.chromium.org/gerrit/18700
Reviewed-by: Chris Masone <cmasone@chromium.org>
Commit-Ready: Scott Zawalski <scottz@chromium.org>
Reviewed-by: Scott Zawalski <scottz@chromium.org>
Tested-by: Scott Zawalski <scottz@chromium.org>
2 files changed