Closed Bug 1601736 Opened 5 years ago Closed 5 years ago

Use preemptible instances in GCP

Categories

(Release Engineering :: Firefox-CI Administration, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: coop, Assigned: coop)

References

Details

Attachments

(6 files, 1 obsolete file)

We've been running on on-demand instances in GCP rather than preemptible. We need to change this before we migrate any further workloads or we'll end up paying too much.

It's a pretty simple addition to the ci-configs, the equivalent of:

"scheduling": {
"onHostMaintenance": "TERMINATE",
"automaticRestart": false,
"preemptible": true
},

Patch incoming.

Depends on D56045

Deployed.

Status: ASSIGNED → RESOLVED
Closed: 5 years ago
Resolution: --- → FIXED

We should not set these options per-instance, but rather globally.

Status: RESOLVED → REOPENED
Resolution: FIXED → ---

We initially filtered out some worker-pools from worker-manager providers
in the legacy cluster that didn't exist in the firefox-ci cluster. Now
that those have been removed, we can stop filtering them out.

These worker pools have never been used in the firefox-ci cluster. They were
test workers in the legacy cluster with provider ids that are not in the new
cluster.

These worker pools have never been used in the firefox-ci cluster. They were
test workers in the legacy cluster with provider ids that are not in the new
cluster.

These are hard-coded in ci-admin (and if we ever need non-prememtible workers
we can add a setting to control that there).

Backed out changeset 45c9502ffe23 (bug 1601736)
Backed out changeset cf2dffcf28c6 (bug 1601736)

Attachment #9114293 - Attachment is obsolete: true
Status: REOPENED → RESOLVED
Closed: 5 years ago5 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: