Elemental Lf Benji Versions Save

Benji Backup for Kubernetes

Benji Backup for Kubernetes

Benji Backup for Kubernetes

Benji Backup for Kubernetes

• Support fsfreeze feature on Kubernetes 1.24+ (requires the updated Helm chart)
• Support for storage classes in S3
• Small change to get Benji to work with MySQL/MariaDB
• k8s-tools: Switch logging to structlog and JSON output
• k8s-tools: Pass through JSON logging from benji unaltered
• k8s-tools: Create PVC with default storage class if nothing is specified
• k8s-tools: Put an upper bound on the time we wait for PV creation

• Fix typo in PersistentVolume key name (#140)

• ACTION REQUIRED: The Helm chart has seen some changes to default values which might break some setups. Especially the default Ceph credential volume mounts are no longer there. In addition, the dependant charts have been updated and the PostgreSQL chart has changed significantly and some keys and values will need adjustment if PostgreSQL is deployed via Benji Helm chart.

• Support for RADOS/RBD namespaces has been added to both the rbd and the rbdaio io modules and the Kubernetes integration. The changes are based on code contributed by @jpsn123.

• The fsfreeze feature works with Ceph CSI provisioned RBD volumes (includes volumes provisioned by Rook).

• There is a new filtering operator like which mimics its SQL counterpart.

• Pushgateway grouping keys based on the Kubernetes namespace and PVC name are used to separate the backup status of different PVCs. Code contributed by @allenporter.

• Deep-scrubbing with source comparison also checks sparse blocks. This will detect blocks that are sparse in the backup and not in the source.

• A bug preventing restores of large images to succeed due using excessive amounts of RAM when using the rbdaio io module has been fixed.

• Container images are based on AlmaLinux 8 and Ceph Pacific.

• Helm chart: Migrate Helm chart to version 2 of the chart format (only compatible with Helm 3)
• Helm chart: Add option to set activeDeadlineSeconds for cronjobs (benji.cronJob.activeDeadlineSeconds) (#108)
• Update container images to fix CVE-2021-20288 in the Ceph client code
• Revalidate invalid blocks during deep-scrubbing (#105)

• Fixed wrong container image repository reference in Helm chart

• ACTION REQUIRED: New container images are only published to the GitHub container registry. The new image locations are:

  • ghcr.io/elemental-lf/benji
  • ghcr.io/elemental-lf/benji-k8s

  The image tag latest now points to the latest released version and not the current master branch. To get the container image for the master branch use the tag master as with all other branches.

• Add options to specify number of retries and timeouts to the S3 storage module (#100, #102, #103)

• Switch to improved retry logic of the boto3 library in the S3 storage module (#103, fixes #101)

• Fix a bug where Benji would not error out if the base version's storage did not match the storage that would've been taken if no base version was specified (fixes #105)

• Use proper fadvise flags on a Ceph RBD read (don't cache us but cache other clients)

• Migrate b2 storage module to b2sdk (closes #46)

• Kubernetes integration:

  • Add support for Ceph CSI in both backup and restore scripts (#106, #104, #107)
  • Add option to restore to a specific storage class (#104, #107)
  • Add option to template Benji restore URLs (#104, #107)
  • Refactor: Move K8s helper functions to benji-k8s-tools package to facilitate code sharing

• Development workflow:

  • Migrate to GitHub Actions
  • Start migration to pytest
  • Automate releases to PyPI