v0.42.0

[machine-controller-manager]

✨ New Features

• [USER] End User can now delete the backing machine object of the node instantly by annotating the desired node with 'node.machine.sapcloud.io/trigger-deletion-by-mcm="true"` (gardener/machine-controller-manager#648, @AxiomSamarth)

v0.41.0

[machine-controller-manager]

✨ New Features

• [USER] Added *expectedNodeDetails field to the MachineClass API (#644, @AxiomSamarth)

🐛 Bug Fixes

• [OPERATOR] A bug has been fixed in the pre-delivered CRD manifests for MCM (/kubernetes/crds). It caused data to be pruned from MCM related resources and led to reconciliation issues. (#641, @timuthy)

📖 Documentation

• [DEVELOPER] make generate now generates v1 version of CRDs by default instead of v1beta1. (#640, @himanshu-kun)

🏃 Others

• [USER] Update Kubernetes dependency versions to v1.20.6 (#601, @AxiomSamarth)

v0.40.0

[machine-controller-manager]

⚠️ Breaking Changes

• [OPERATOR] Draining of pods with PVs (Persistent Volume) now waits for re-attachment of PV on a different node when volumeAttachments support is enabled on the cluster. Else it falls back to the default PV reattachment timeout value configured. The default value is 90s and this can be overwritten via the machine-pv-reattach-timeout flag. Please enable permissions to allow listing of volumeAttachments resource while importing these changes. (#608, @prashanth26)

✨ New Features

• [USER] Increase default concurrent object syncs to 50 to allow more concurrent reconciles to occur. (#629, @prashanth26)
• [USER] Machine rollouts are now more as desired with the number of replicas always maintained to desired + maxSurge. Earlier machines in termination were left out of this calculation but now is considered with this change. (#627, @prashanth26)
• [OPERATOR] Finalizers will be added to the MachineClass which is used by at least one machine. Machines whose backing MachineClass does not have finalizers shall not be reconciled. (#593, @AxiomSamarth)
• [DEVELOPER] Replace integration test with unit test to test the functionality to freeze MachineSet (#620, @AxiomSamarth)

🐛 Bug Fixes

• [OPERATOR] Avoids blocking of drain call when the buffer is full for the volumeAttachmentHandlers. (#627, @prashanth26)
• [DEVELOPER] Test framework now fetches secrets from the correct (control) APIServer while running tests. (#617, @himanshu-kun)

🏃 Others

• [OPERATOR] Nodes attached to the cluster without MCM support are now annotated with "node.machine.sapcloud.io/notManagedByMCM": "1". This is then ignored by the MCM for further processing. (#612, @himanshu-kun)

v0.39.0

[machine-controller-manager]

✨ New Features

• [USER] Skip node drain on ReadOnlyFileSystem condition (#605, @himanshu-kun)
• [OPERATOR] Improved log details to include node name and provider-ID in addition to existing machine name (#607, @himanshu-kun)

🐛 Bug Fixes

• [OPERATOR] Fix panic when machineClass secretRef isn't found. (#609, @jsravn)
• [DEVELOPER] Adds finalizers on machines that are adopted by the machine controller. Without this change, it causes issues while migrating machine objects between clusters. (#611, @prashanth26)

v0.38.0

[machine-controller-manager]

🐛 Bug Fixes

• [DEVELOPER] Azure: Improved NIC creation and deletion logic to handle NIC creation and deletions more gracefully. (#594, @prashanth26)

v0.37.0

[machine-controller-manager]

⚠️ Breaking Changes

• [DEVELOPER] machine-controller-manager now checks for misconfigured PodDisruptionBudgets (ones that require zero voluntary evictions and make impossible the graceful Node drain) and sets better Machine .status.lastOperation.description for such Machines. This change is breaking as out-of-tree providers need new RBAC permissions - list and watch access for PodDisruptionBudgets in the target cluster. (#591, @ialidzhikov)

🏃 Others

• [OPERATOR] Avoid the deletion of the machines in CrashLoopBackoff state by the safety controller (#589, @AxiomSamarth)

v0.36.0

[machine-controller-manager]

✨ New Features

• [OPERATOR] All machine classes do now support an optional .{spec.}credentialsSecretRef field in addition to today's .{spec.}secretRef field. If .{spec.}credentialsSecretRef is non-nil then the provider credentials will be read out of this secret. The user-data for the machine bring-up is still required to be part of the secret referenced by .{spec.}secretRef. (#578, @rfranzke)
• [OPERATOR] Some machine class secrets are now supporting alternative data keys: (#578, @rfranzke)
  • The machine class secret for Alicloud machines does now also accept the data keys accessKeyID and accessKeySecret as alternatives for today's keys.
  • The machine class secret for AWS machines does now also accept the data keys accessKeyID and secretAccessKey as alternatives for today's keys.
  • The machine class secret for Azure machines does now also accept the data keys clientID, clientSecret, subscriptionID and tenantID as alternatives for today's keys.
  • The machine class secret for GCP machines does now also accept the data key serviceaccount.json as alternatives for today's key.

🏃 Others

• [OPERATOR] Bumped AWS SDK version to v1.23.13 (#580, @zjj2wry)
• [OPERATOR] An issue causing panic when the encoded machine template hash length is less than expect limit is now fixed. (#575, @ialidzhikov)
• [OPERATOR] MCM will delete Azure machines even if the underlying resource group is already deleted. (#566, @dkistner)
• [OPERATOR] Set Machine Phase to Terminating before draining. (#564, @prashanth26)
• [DEVELOPER] Update docker images to use gcr copy (#574, @prashanth26)
• [DEVELOPER] Update docker image versions to golang:1.15.5 & alpine:3.12.1 (#574, @prashanth26)

📰 Noteworthy

• [OPERATOR] Machine force deletion computation is based on deletionTimestamp instead of LastUpdatedTimestamp. (#564, @prashanth26)

v0.35.2

[machine-controller-manager]

Improvements

• [OPERATOR] An issue causing panic when the encoded machine template hash length is less than expect limit is now fixed. (#575, @AxiomSamarth)

v0.35.1

[machine-controller-manager]

Most notable changes

• [OPERATOR] Machine force deletion computation is based on deletionTimestamp instead of LastUpdatedTimestamp.. (#564, @prashanth26)

Improvements

• [OPERATOR] Set Machine Phase to Terminating before draining. (#564, @prashanth26)

v0.34.4

[machine-controller-manager]

Most notable changes

• [OPERATOR] Machine force deletion computation is based on deletionTimestamp instead of LastUpdatedTimestamp. (#564, @prashanth26)

Improvements

• [OPERATOR] Set Machine Phase to Terminating before draining. (#564, @prashanth26)