Previous attempt: https://github.com/getsentry/ops/pull/9218
We could try to revive that version here.
Pros:
- Automatic, prevent accidental rollback entirely
Cons:
- New services will be broken until a subsequent code deploy goes out. This would slow feedback loop on new services significantly (eg. if there's a separate issue which will cause crash loop, we might take an hour to realize it due to code deploy).
An alternative could be a little simpler / manual - add a new flag to sentry-kube apply --image 'us.gcr.io/internal-sentry/github-getsentry-github-deployment-gate:55f0132854a0c018c00a533da567379853d912b7'
This would mean that conflicts during GoCD deployments could be dealt with, but it would require a little more manual effort. On the positive side, new services would work as is.
Previous attempt: https://github.com/getsentry/ops/pull/9218
We could try to revive that version here.
Pros:
Cons:
An alternative could be a little simpler / manual - add a new flag to
sentry-kube apply --image 'us.gcr.io/internal-sentry/github-getsentry-github-deployment-gate:55f0132854a0c018c00a533da567379853d912b7'This would mean that conflicts during GoCD deployments could be dealt with, but it would require a little more manual effort. On the positive side, new services would work as is.