Frequently Asked Questions¶

Why do we need to follow “the Kubernetes way” when Kubernetes was never intended to run databases?¶

As it is well known, the Kubernetes approach is targeted at stateless applications but provides ways to store state (in Persistent Volumes, etc.) if the application needs it. Generally, a stateless mode of operation is supposed to provide better safety, sustainability, and scalability, it makes the already-deployed components interchangeable. You can find more about substantial benefits brought by Kubernetes to databases in this blog post .

The architecture of state-centric applications (like databases) should be composed in a right way to avoid crashes, data loss, or data inconsistencies during hardware failure. Percona Operator for MySQL provides out-of-the-box functionality to automate provisioning and management of highly available MySQL database clusters on Kubernetes.

How can I contact the developers?¶

The best place to discuss Percona Operator for MySQL based on Percona XtraDB Cluster with developers and other community members is the community forum .

If you would like to report a bug, use the Percona Operator for MySQL project in JIRA .

What is the difference between the Operator quickstart and advanced installation ways?¶

As you have noticed, the installation section of docs contains both quickstart and advanced installation guides.

The quickstart guide is simpler. It has fewer installation steps in favor of predefined default choices. Particularly, in advanced installation guides, you separately apply the Custom Resource Definition and Role-based Access Control configuration files with possible edits in them. At the same time, quickstart guides rely on the all-inclusive bundle configuration.

At another point, quickstart guides are related to specific platforms you are going to use (Minikube, Google Kubernetes Engine, etc.) and therefore include some additional steps needed for these platforms.

Generally, rely on the quickstart guide if you are a beginner user of the specific platform and/or you are new to the Percona Distribution for MySQL Operator as a whole.

Which versions of MySQL does the Percona Operator for MySQL support?¶

Percona Operator for MySQL based on Percona XtraDB Cluster provides a ready-to-use installation of the MySQL-based Percona XtraDB Cluster inside your Kubernetes installation. It works with both MySQL 8.0 and 5.7 branches, and the exact version is determined by the Docker image in use.

Percona-certified Docker images used by the Operator are listed here. As you can see, both Percona XtraDB Cluster 8.0 and 5.7 are supported with the following recommended versions: 8.0.39-30.1 and 5.7.44-31.65. Three major numbers in the XtraDB Cluster version refer to the version of Percona Server in use. More details on the exact Percona Server version can be found in the release notes (8.0 , 5.7 ).

How is HAProxy better than ProxySQL?¶

Percona Operator for MySQL based on Percona XtraDB Cluster supports both HAProxy and ProxySQL as a load balancer. HAProxy is turned on by default, but both solutions are similar in terms of their configuration and operation under the control of the Operator.

Still, they have technical differences. HAProxy is a general and widely used high availability, load balancing, and proxying solution for TCP and HTTP-based applications. ProxySQL provides similar functionality but is specific to MySQL clusters. As an SQL-aware solution, it is able to provide more tight internal integration with MySQL instances.

Both projects do a really good job with the Operator. The proxy choice should depend mostly on application-specific workload (including object-relational mapping), performance requirements, advanced routing and caching needs with one or another project, components already in use in the current infrastructure, and any other specific needs of the application.

How can I create a directory on the node to use it as a local storage¶

You can configure hostPath volume to mount some existing file or directory from the node’s filesystem into the Pod and use it as a local storage. The directory used for local storage should already exist in the node’s filesystem. You can create it through the shell access to the node, with mkdir command, as all other directories. Alternatively you can create a Pod which will do this job. Let’s suppose you are going to use /var/run/data-dir directory as your local storage, describing it in the deploy/cr.yaml configuration file as follows:

...
pxc:
  ...
  volumeSpec:
     hostPath:
       path: /var/run/data-dir
       type: Directory
  containerSecurityContext:
    privileged: false
  podSecurityContext:
    runAsUser: 1001
    runAsGroup: 1001
    supplementalGroups: [1001]
  nodeSelector:
    kubernetes.io/hostname: a.b.c

Create the yaml file (e.g. mypod.yaml), with the following contents:

apiVersion: v1
kind: Pod
metadata:
  name: hostpath-helper
spec:
  containers:
  - name: init
    image: busybox
    command: ["install", "-o", "1001", "-g", "1001", "-m", "755", "-d", "/mnt/data-dir"]
    volumeMounts:
    - name: root
      mountPath: /mnt
    securityContext:
      runAsUser: 0
  volumes:
  - name: root
    hostPath:
      path: /var/run
  restartPolicy: Never
  nodeSelector:
    kubernetes.io/hostname: a.b.c

Don’t forget to apply it as usual:

$ kubectl apply -f mypod.yaml

How can I add custom sidecar containers to my cluster?¶

The Operator allows you to deploy additional (so-called sidecar) containers to the Pod. You can use this feature to run debugging tools, some specific monitoring solutions, etc. Add such sidecar container to the deploy/cr.yaml configuration file, specifying its name and image, and possibly a command to run:

spec:
  pxc:
    ....
    sidecars:
    - image: busybox
      command: ["/bin/sh"]
      args: ["-c", "while true; do echo echo $(date -u) 'test' >> /dev/null; sleep 5; done"]
      name: my-sidecar-1
    ....

You can add sidecars subsection to pxc, haproxy, and proxysql sections.

Note

Custom sidecar containers can easily access other components of your cluster . Therefore they should be used carefully and by experienced users only.

Find more information on sidecar containers in the appropriate documentation page.

How to get core dumps in case of the Percona XtraDB Cluster crash¶

In the Percona XtraDB Cluster crash case, gathering all possible information for enhanced diagnostics to be shared with Percona Support helps to solve an issue faster. One of such helpful artifacts is core dump .

Percona XtraDB Cluster can create core dumps on crush using libcoredumper . The Operator has this feature turned on by default. Core dumps are saved to DATADIR (var/lib/mysql/). You can find appropriate core files in the following way (substitute some-name-pxc-1 with the name of your Pod):

$ kubectl exec some-name-pxc-1 -c pxc -it -- sh -c 'ls -alh /var/lib/mysql/ | grep core'
-rw------- 1 mysql mysql 1.3G Jan 15 09:30 core.20210015093005

When identified, the appropriate core dump can be downloaded as follows:

$ kubectl cp some-name-pxc-1:/var/lib/mysql/core.20210015093005  /tmp/core.20210015093005

Note

It is useful to provide Build ID and Server Version in addition to core dump when Creating a support ticket. Both can be found from logs:

$ kubectl logs some-name-pxc-1 -c logs

[1] init-deploy-949.some-name-pxc-1.mysqld-error.log: [1610702394.259356066, {"log"=>"09:19:54 UTC - mysqld got signal 11 ;"}]
[2] init-deploy-949.some-name-pxc-1.mysqld-error.log: [1610702394.259356829, {"log"=>"Most likely, you have hit a bug, but this error can also be caused by malfunctioning hardware."}]
[3] init-deploy-949.some-name-pxc-1.mysqld-error.log: [1610702394.259457282, {"log"=>"Build ID: 5a2199b1784b967a713a3bde8d996dc517c41adb"}]
[4] init-deploy-949.some-name-pxc-1.mysqld-error.log: [1610702394.259465692, {"log"=>"Server Version: 8.0.21-12.1 Percona XtraDB Cluster (GPL), Release rel12, Revision 4d973e2, WSREP version 26.4.3, wsrep_26.4.3"}]
.....

How to choose between HAProxy and ProxySQL when configuring the cluster?¶

You can configure the Operator to use one of two different proxies, HAProxy (the default choice) and ProxySQL. Both solutions are fully supported by the Operator, but they have some differences in the architecture, which can make one of them more suitable then the other one in some use cases.

The main difference is that HAProxy operates in TCP mode as an OSI level 4 proxy , while ProxySQL implements OSI level 7 proxy, and thus can provide some additional functionality like read/write split, firewalling and caching.

From the other side, utilizing HAProxy for the service is the easier way to go, and getting use of the ProxySQL level 7 specifics requires good understanding of Kubernetes and ProxySQL.

You can enable ProxySQL only at cluster creation time. Otherwise you will be able to use HAProxy only. The switch from HAProxy to ProxySQL is not possible, because ProxySQL does not yet support caching_sha2_password MySQL authentication plugin used by the Operator by default instead of the older mysql_native_password one.

See more detailed functionality and performance comparison of using the Operator with both solutions in this blog post .

Which additional access permissions are used by the Custom Resource validation webhook?¶

The spec.enableCRValidationWebhook key in the deploy/cr.yaml file enables or disables schema validation done by the Operator before applying cr.yaml file. This feature works only in cluster-wide mode due to access restrictions. It uses the following additional RBAC permissions :

- apiGroups:
  - admissionregistration.k8s.io
  resources:
  - validatingwebhookconfigurations
  verbs:
  - get
  - list
  - watch
  - create
  - update
  - patch
  - delete

Get expert help¶

If you need assistance, visit the community forum for comprehensive and free database knowledge, or contact our Percona Database Experts for professional support and services. Join K8S Squad to benefit from early access to features and “ask me anything” sessions with the Experts.

Community Forum Get a Percona Expert Join K8S Squad

Last update: 2024-12-14