Cleanup backup tar after copying to GCS #1152
Conversation
…bucket name if other options failed
| @@ -123,7 +123,11 @@ fi | |||
| # test and create bucket if necessary | |||
| gsutil ls gs://${gcs_bucket} &>/dev/null || { | |||
| echo "Bucket '${gcs_bucket}' was not found. Creating it.." | |||
| gsutil mb gs://"${gcs_bucket}" | |||
| gsutil mb gs://"${gcs_bucket}" || { | |||
There was a problem hiding this comment.
I'd suggest a slightly different ordering of the logic where we only try to run gsutil mb once:
Basically, we'd look at the "gcs_bucket" (without any default), and "project_id" environment variables to figure out what the bucket name should be.
If "gcs_bucket" is defined, then that is the bucket name. Otherwise we run gsutil ls gs://${project_id}.appspot.com to see if that bucket exists, and if so set the bucket name to it. Finally, if neither of the two previous tests passed, then we set the bucket name to match the project ID.
There was a problem hiding this comment.
Do you mean we shouldn't create the gs://${project_id}.appspot.com bucket if it doesn't exist? Other than this, I don't see much difference between the comment and the changes.
There was a problem hiding this comment.
In addition to not trying to create the "...appspot.com" bucket (which we know would always fail), the other difference between what I proposed and the current approach is that we would not try to create the project bucket if the user has explicitly specified a gcs_bucket, regardless of whether or not that gcs_bucket exists.
yebrahim
force-pushed
the
yebrahim/cleanup-gcsbackup-tar
branch
from
fc3811b
to
5685f70
Compare
yebrahim
force-pushed
the
yebrahim/cleanup-gcsbackup-tar
branch
from
5685f70
to
cb5de7c
Compare
| else | ||
| gsutil ls gs://"${gcs_bucket}" &>/dev/null || { | ||
| echo "Could not list bucket '${gcs_bucket}'. Will try to create it.." | ||
| gsutil mb gs://"${gcs_bucket}" || true |
There was a problem hiding this comment.
Should this have an "|| true"?
If this mb call fails, then the script should exit.
Also, try using project_id as bucket name if other options have failed.