This release fixed a long time bug: when creating new replicated volumes, sometimes one of the replica creation fails because of disk space, failed disk, or limited file handles, etc. Previously the other successfully created volume replicas are not purged. But these volumes are not in use since they are under replicated. So a lot of empty volumes are created quickly. This release purges these partially replicated volumes.
What's Changed
- Master
- Remove extra volume if creating one of the replicated volumes failed.
- avoid vacuum if not enough replica copies by @kmlebedev in #3924
- Volume Server
- add Content-MD5 header when UploadData to replication in ReplicatedWrite by @paochiang in #3881
- ReadNeedleMeta recovery panic occurred by @kmlebedev in #3883
- ensure monotonic n.AppendAtNs in each place by @kmlebedev in #3880
- Other
- Corrected the dashboard to use the new series name. by @egguy in #3887
- fix volume statefulset helm chart's tolerations by @theRealWardo in #3911
- S3
- Shell
- avoid fix missing of deleted needles by @kmlebedev in #3909
- [voluche.chek.disk] needles older than the cutoff time are not missing yet by @kmlebedev in #3922
- [volume.chek.disk] sync of deletions the fix by @kmlebedev in #3923
weed fix- add param for ignore error by @kmlebedev in #3918
New Contributors
- @egguy made their first contribution in #3887
- @theRealWardo made their first contribution in #3911
Full Changelog: 3.32...v3.33