Percona Kubernetes Operator for Percona XtraDB Cluster 1.6.0¶
-
Date
October 9, 2020
-
Installation
Installing Percona Kubernetes Operator for Percona XtraDB Cluster
New Features¶
-
K8SPXC-394: Support of “cluster-wide” mode for Percona XtraDB Cluster Operator
-
K8SPXC-416: Support of the proxy-protocol in HAProxy (to use this feature, you should have a Percona XtraDB Cluster image version
8.0.21
or newer) -
K8SPXC-429: A possibility to restore backups to a new Kubernetes-based environment with no existing Percona XtraDB Cluster Custom Resource
-
K8SPXC-343: Helm chart officially provided with the Operator
Improvements¶
-
K8SPXC-398: New
crVersion key
indeploy/cr.yaml
to indicate the API version that the Custom Resource corresponds to (thanks to user mike.saah for contribution) -
K8SPXC-474: The init container now has the same resource requests as the main container of a correspondent Pod (thanks to user yann.leenhardt for contribution)
-
K8SPXC-372: Support new versions of cert-manager by the Operator (thanks to user rf_enigm for contribution)
-
K8SPXC-317: Possibility to configure the
imagePullPolicy
Operator option (thanks to user imranrazakhan for contribution) -
K8SPXC-462: Add readiness probe for HAProxy
-
K8SPXC-411: Extend cert-manager configuration to add additional domains (multiple SAN) to a certificate
-
K8SPXC-375: Improve HAProxy behavior in case of switching writer node to a new one and back
-
K8SPXC-368: Autoupdate system users by changing the appropriate Secret name
Known Issues and Limitations¶
-
OpenShift 3.11 requires additional configuration for the correct HAProxy operation: the feature gate
PodShareProcessNamespace
should be set totrue
. If getting it enabled is not possible, we recommend using ProxySQL instead of HAProxy with OpenShift 3.11. Other OpenShift and Kubernetes versions are not affected. -
K8SPXC-491: Compressed backups are not compatible with the Operator 1.6.0 (
percona/percona-xtradb-cluster-operator:1.5.0-pxc8.0-backup
orpercona/percona-xtradb-cluster-operator:1.5.0-pxc5.7-backup
image can be used as a workaround if needed).
Bugs Fixed¶
-
K8SPXC-431: HAProxy unable to start on OpenShift with the default
cr.yaml
file -
K8SPXC-408: Insufficient MAX_USER_CONNECTIONS=10 for ProxySQL monitor user (increased to 100)
-
K8SPXC-391: HAProxy and PMM cannot be enabled at the same time (thanks to user rf_enigm for reporting this issue)
-
K8SPXC-406: Second node (XXX-pxc-1) always selected as a donor (thanks to user pservit for reporting this issue)
-
K8SPXC-390: Crash on missing HAProxy PodDisruptionBudget
-
K8SPXC-355: Counterintuitive YYYY-DD-MM dates in the S3 backup folder names (thanks to user graham-web for contribution)
-
K8SPXC-305: ProxySQL not working in case of passwords with a
%
symbol in the Secrets object (thanks to user ben.wilson for reporting this issue) -
K8SPXC-278: ProxySQL never getting ready status in some environments after the cluster launch due to the
proxysql-monit
Pod crash (thanks to user lots0logs for contribution) -
K8SPXC-274: The 1.2.0 -> 1.3.0 -> 1.4.0 upgrade path not working (thanks to user martin.atroo for reporting this issue)
-
K8SPXC-476: SmartUpdate failing to fetch version from Version Service in case of incorrectly formatted Percona XtraDB Cluster patch version higher than the last known one
-
K8SPXC-454: After the cluster creation, pxc-0 Pod restarting due to Operator not waiting for cert-manager to issue requested certificates (thanks to user mike.saah for reporting this issue)
-
K8SPXC-450: TLS annotations causing unnecessary HAProxy Pod restarts
-
K8SPXC-443 and K8SPXC-456: The outdated version service endpoint URL (fix with preserving backward compatibility)
-
K8SPXC-435: MySQL root password visible through
kubectl logs
-
K8SPXC-426: mysqld recovery logs not logged to file and not available through
kubectl logs
-
K8SPXC-423: HAProxy not refreshing IP addresses even when the node gets a different address
-
K8SPXC-419: Percona XtraDB Cluster incremental state transfers not taken into account by readiness/liveness checks
-
K8SPXC-418: HAProxy not routing traffic for 1 donor, 2 joiners
-
K8SPXC-417: Cert-manager not compatible with Kubernetes versions below v1.15 due to unnecessarily high API version demand
-
K8SPXC-384: Debug images were not fully functional for the latest version of the Operator because of having no infinity loop
-
K8SPXC-383: DNS warnings in PXC Pods when using HAProxy
-
K8SPXC-364: Smart Updates showing empty “from” versions for non-PXC objects in logs
-
K8SPXC-379: The Operator user credentials not added into internal secrets when upgrading from 1.4.0 (thanks to user pservit for reporting this issue)