v0.17.0

Changed

• Enable ShareReconciler.

v0.16.0

Changed

• Update CAPI/CAPA/controller-runtime Go modules

Fixed

• Disable logger development mode to avoid panicking

v0.15.0

Changed

• Update capa to v2.3.0.
• Use ResourceID instead of ID for CAPA Subnets.

v0.14.6

Changed

• Add toleration for node.cluster.x-k8s.io/uninitialized taint.
• Add node affinity to prefer scheduling CAPI pods to control-plane nodes.

v0.14.5

Fixed

• Fix finding NS record of the DNS zone. This only happens in rare cases.

v0.14.4

Changed

• Avoid deletion reconciliation if finalizer is already gone

v0.14.3

Changed

• Further avoidance for Route53 rate limiting

  • Cache ListResourceRecordSets responses for NS record of a zone
  • Only upsert DNS records if not recently done

v0.14.2

Changed

• Cache hosted zone ID on creation
• List many hosted zones at once in one Route53 request and cache all returned zones. This further reduces the number of Route53 requests and therefore avoids rate limit (throttling) errors.

v0.14.1

Fixed

• Cache GetHostedZoneIdByName responses to avoid Route53 rate limit. The hosted zone IDs are unlikely to change so quickly. If reconciliation of an object (e.g. AWSCluster) permanently gets retriggered outside of this operator's control, it could previously lead to triggering the account-wide AWS Route53 rate limit of five requests per second.

v0.14.0

Changed

• Do not reconcile bastion machines anymore. They were removed in cluster-aws v0.53.0 in favor of Teleport. On such newer clusters without a bastion host, the controller would retry unnecessary Route53 calls every minute, leading to a rate limit – this is fixed by removing the whole logic.