Merge lp://staging/~sinzui/juju-ci-tools/delete-s3-control-buckets into lp://staging/juju-ci-tools
| Status: | Merged |
|---|---|
| Merged at revision: | 1135 |
| Proposed branch: | lp://staging/~sinzui/juju-ci-tools/delete-s3-control-buckets |
| Merge into: | lp://staging/juju-ci-tools |
| Diff against target: |
32 lines (+7/-7) 1 file modified
delete-old-s3-control-buckets.bash (+7/-7) |
| To merge this branch: | bzr merge lp://staging/~sinzui/juju-ci-tools/delete-s3-control-buckets |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Martin Packman (community) | Approve | ||
|
Review via email:
|
|||
Description of the change
Delete s3 control-buckets that are about 2 hours old.
This branch is a quick change to the current deletion script. It changes the rule from about a day ago to about 2 hours ago. The changes include the hours in the s3 creation time and current time. There was a small matter of escaping the space between the day and the hour.
While testing many jujus, I discovered that healthy jujus are very good at deleting the control-bucket. I think many of the control-buckets are from bad jujus, some that fail to bootstrap, so we cannot ask the running env which control bucket needs to be removed. Since no test job runs more than 1 hour, and 1.21+ doesn't need a control bucket after bootstrap, I choose to delete buckets that are about 2 hours old. I have verified that 1.18 and 1.20 will recreate their control-bucket if a long running job did loose its control-bucket.
This script really should be using boto, but I don't have time to write a replacement script in my remaining hours this week.
Small enough change to the bash script that we can live without botoing it. I can't eyeball verify it works but trust manual runs have confirmed some amount of sanity.