Releases: FoundationDB/fdb-kubernetes-operator
Releases · FoundationDB/fdb-kubernetes-operator
Release v0.42.1
patch version bump
Release v0.42.0
v0.42.0
- Adds a command to fix coordinator IPs when cluster is unavailable.
- Change the kubebuilder download path in the PR build.
- Allow exclusions when another process is missing.
- Replace pods with the PodFailing condition.
- Skip already upgraded processes.
- Printout pod names instead of a pod struct.
- Adds a log field reconciler.
- Minor refactoring.
Release v0.41.1
v0.41.1
- Restore a missing shim
Release v0.41.0
v0.41.0
- Fix the security context and ClusterRoleBindings in the helm chart.
- Fix some scenarios where we block reconciliation on pending pods.
- Add more flexible options for customizing images for built-in containers.
Release v0.40.0
v0.40.0
- Prevent doubl-encoding ipv6 addresses when translating between pod IPs and process addresses.
- Replace instances if PVC hash/name does not match.
- Using explicit listen address for all clusters.
- Revert the change for th eCRD to use coordinators as strings.
Release v0.39.2
v0.39.2
- Refactor common methods into more specific files in the internal package
- Expose the GetPodSpec method in the controller package
- Allow to only set the IP address without any ports and add more tests
Release v0.39.1
v0.39.1
- add condition for unreachable sidecar
- revert unexpected args checks
Release v0.39.0
v0.39.0
- Check for unexpected arguments.
- Remove sort from initial file generation.
- Add the number of desired pods in the message for initial cluster file generation.
- Enable log file rotation.
- Fix log message using wrong cluster key.
- Remove explicity gomega imports.
- Remove sidecar code.
- Add methods to get next version based on current version.
- Add additional logs for reconcile state.
- Wrap error with additional process group info.
- Make use of net.IP for address.
- Updaet the index page in user manual.
Release v0.38.0
v0.38.0
- Fix CI.
- Remove deprecated metrics.
- Don't skip update status after process groups are removed.
- Wait to set the HasPendingRemoval flag in the status until we’ve started terminating the resources.
- Deprecate the Resolver field in the process counts.
- Support defining custom match labels for identifying resources owned by the operator.
- Refactor reconciliations results to use a custom structure.
- Pod IP selection rules for dual-stack environments.
- Make health check in the database more narrowly-targeted.
- Enable automatic replacements by default.
- Improve user manual.
- Remove cache for the database.
- Merge affinities instead of replacing them.
- Update output files.
- Add a compact format to the FDB version.
- Allow defininig which processes should take part in coordinator selection.
- Fallback to pod IP rather than machine ID.
Release v0.37.0
v0.37.0
- Show the availability instead of the cluster health in the kubectl output.
- Ensure that coordinator selection is deterministic.
- Ensure that we never remove an active coordinator.
- Print messages about pending removals in the
kubectl fdb analyze
command. - Only reconcile on generation or annotation change.
- Allow reconciliation to move forward when PVCs are stuck in terminating.
- Add an option for the maximum concurrent replacements when doing safety checks
for automatic replacements. - Add an option for the maximum concurrent reconciliations.
- Clean up the helm chart.