MCK 1.14.0 Release Notes

New Features

MongoDBOpsManager:

• Added fields spec.backup.s3Stores[].objectRetentionDays and spec.backup.s3Stores[].objectRetentionMode for configuring the S3 Object Lock retention period and mode (GOVERNANCE or COMPLIANCE) on backup snapshot stores. Both fields require objectLockEnabled and Ops Manager >= 8.0.27 (the operator omits them from the Ops Manager API on older versions, where they are rejected); objectLockEnabled itself requires Ops Manager >= 8.0.19.
• Added support for objectLockEnabled, objectRetentionDays and objectRetentionMode on S3 OpLog stores (spec.backup.s3OpLogStores), which were previously rejected by validation. S3 Object Lock on OpLog stores requires Ops Manager >= 8.0.28 (unlike snapshot stores, which support it since 8.0.19); objectRetentionDays and objectRetentionMode must be specified together and require objectLockEnabled.

Bug Fixes

• Ops Manager requests: Fixed an issue where Ops Manager API requests kept running to their own timeouts after a reconcile was cancelled or the Operator shut down. Requests are now cancelled together with the reconcile, freeing the worker to handle other resources.
• MongoDBOpsManager: The Application Database agent containers now follow the registry.pullPolicy Helm setting instead of always pulling their images. Setting it to IfNotPresent avoids re-downloading these images every time an Application Database pod restarts.
• Sharded migration scaling: Fixed a bug in the MongoDB sharded cluster resource where scaling Kubernetes members up from zero during a VM migration (while spec.externalMembers are present) added all new voting members in a single reconfiguration, which Ops Manager rejects. The operator now adds the members one at a time.
• Sharded cluster: Fixed a bug where scaling down spec.shardCount left the hosts of removed shards registered in Ops Manager, keeping orphaned processes visible in the deployment view.
• MongoDBSearch:
  • The mongot container now follows the registry.pullPolicy Helm setting instead of always pulling its image. Setting it to IfNotPresent avoids re-downloading the image every time a search pod restarts.
  • The operator-managed Envoy load balancer now restarts automatically when its TLS server or client certificate Secret changes. Existing TLS-enabled Envoy Deployments restart once on upgrade; non-TLS Deployments are unchanged.