High availability via Galera

The mariadb-operator provides cloud native support for provisioning and operating multi-master MariaDB clusters using Galera. This setup enables the ability to perform both read and write operations on all nodes, enhancing availability and allowing scalability across multiple nodes.

In certain circumstances, it could be the case that all the nodes of your cluster go down at the same time, something that Galera is not able to recover by itself, and it requires manual action to bring the cluster up again, as documented in the Galera documentation. Luckly enough, mariadb-operator has you covered and it encapsulates this operational expertise in the MariaDB CRD. You just need to declaratively specify spec.galera, as explained in more detail later in this guide.

To accomplish this, after the MariaDB cluster has been provisioned, mariadb-operator will regularly monitor the cluster's status to make sure it is healthy. If any issues are detected, the operator will initiate the recovery process to restore the cluster to a healthy state. During this process, the operator will set status conditions in the MariaDB and emit Events so you have a better understanding of the recovery progress and the underlying activities being performed. For example, you may want to know which Pods were out of sync to further investigate infrastructure-related issues (i.e. networking, storage...) on the nodes where these Pods were scheduled.

Components

To be able to effectively provision and recover MariaDB Galera clusters, the following components were introduced to co-operate with mariadb-operator:

init: Init container that dynamically provisions the Galera configuration file before the MariaDB container starts. Guarantees ordered deployment of Pods even if spec.podManagementPolicy = Parallel is set on the MariaDB StatefulSet, something crucial for performing the Galera recovery, as the operator needs to restart Pods independently.
agent: Sidecar agent that exposes the Galera state (grastate.dat) via HTTP and allows the operator to remotely bootstrap and recover the Galera cluster. For security reasons, it has authentication based on Kubernetes service accounts; this way only the operator is able to call the agent.

All these components are available in the ghcr.io/mariadb-operator mariadb-operator image. More preciselly, they are subcommands of the CLI shipped as binary inside the image.

`MariaDB` configuration

The easiest way to get a MariaDB Galera cluster up and running is setting spec.galera.enabled = true, like in this example:

apiVersion: k8s.mariadb.com/v1alpha1
kind: MariaDB
metadata:
  name: mariadb-galera
spec:
...
  replicas: 3

  galera:
    enabled: true

This relies on sensible defaults set by the operator, which may not be suitable for your Kubernetes cluster. This can be solved by overriding the defaults, as in this other example, so you have fine-grained control over the Galera configuration.

Refer to the reference section to better understand the purpose of each field.

Storage

By default, mariadb-operator provisions two PVCs for running Galera:

Storage PVC: Used to back the MariaDB data directory, mounted at /var/lib/mysql.
Config PVC: Where the Galera config files are located, mounted at /etc/mysql/conf.d.

However, you are also able to use just one PVC for keeping both the data and the config files:

apiVersion: k8s.mariadb.com/v1alpha1
kind: MariaDB
metadata:
  name: mariadb-galera
spec:
  ...
  galera:
    enabled: true
    config:
      reuseStorageVolume: true

Galera cluster recovery

mariadb-operator is able to monitor the Galera cluster and act accordinly to recover it if needed. This feature is enabled by default, but you may tune it as you need:

apiVersion: k8s.mariadb.com/v1alpha1
kind: MariaDB
metadata:
  name: mariadb-galera
spec:
  ...
  galera:
    enabled: true
    recovery:
      enabled: true
      minClusterSize: 50%
      clusterMonitorInterval: 10s
      clusterHealthyTimeout: 30s
      clusterBootstrapTimeout: 10m
      podRecoveryTimeout: 3m
      podSyncTimeout: 3m

The minClusterSize field indicates the minimum cluster size (either absolut number or percentage) for the operator to consider the cluster healthy. If the cluster is unhealthy for more than the period defined in clusterHealthyTimeout, a cluster recovery process is initiated by the operator. The process is explained in the Galera documentation and consists of the following steps:

Recover the sequence number from the grastate.dat on each node.
Put the nodes in recovery mode and restart them to obtain the sequence number. This step is skipped if we have already obtained a valid sequence number in the previous step.
Mark the node with highest sequence (bootstrap node) as safe to bootstrap.
Bootstrap a new cluster in the bootstrap node.
Wait until the bootstrap node becomes ready.
Restart the rest of the nodes one by one so they can join the new cluster.

The operator monitors the Galera cluster health periodically and performs the cluster recovery described above if needed. You are able to tune the monitoring interval via the clusterMonitorInterval field.

Refer to the reference section to better understand the purpose of each field.

Wsrep provider

You are able to pass extra options to the Galera wsrep provider by using the galera.providerOptions field:

apiVersion: k8s.mariadb.com/v1alpha1
kind: MariaDB
metadata:
  name: mariadb-galera
spec:
  ...
  galera:
    providerOptions:
      gcs.fc_limit: '64'

It is important to note that, the ist.recv_addr cannot be set by the user, as it is automatically configured to the Pod IP by the operator, something that an user won't be able to know beforehand.

A list of the available options can be found in the MariaDB documentation.

IPv6 support

If you have a Kubernetes cluster running with IPv6, the operator will automatically detect the IPv6 addresses of your Pods and it will configure several wsrep provider options to ensure that the Galera protocol runs smoothly with IPv6.

Quickstart

First of all, install the following configuration manifests that will be referenced by the CRDs further:

kubectl apply -f examples/manifests/config

Next, you can proceed with the installation of a MariaDB instance with Galera support:

kubectl apply -f examples/manifests/mariadb_galera.yaml

kubectl get mariadbs
NAME             READY   STATUS    PRIMARY POD          AGE
mariadb-galera   True    Running   mariadb-galera-0     48m

kubectl get events --field-selector involvedObject.name=mariadb-galera --sort-by='.lastTimestamp'
LAST SEEN   TYPE     REASON                 OBJECT                               MESSAGE
...
45m         Normal   GaleraClusterHealthy   mariadb/mariadb-galera               Galera cluster is healthy

kubectl get mariadb mariadb-galera -o jsonpath="{.status.conditions[?(@.type=='GaleraReady')]}"
{"lastTransitionTime":"2023-07-13T18:22:31Z","message":"Galera ready","reason":"GaleraReady","status":"True","type":"GaleraReady"}

kubectl get mariadb mariadb-galera -o jsonpath="{.status.conditions[?(@.type=='GaleraConfigured')]}"
{"lastTransitionTime":"2023-07-13T18:22:31Z","message":"Galera configured","reason":"GaleraConfigured","status":"True","type":"GaleraConfigured"}

kubectl get statefulsets -o wide
NAME             READY   AGE   CONTAINERS      IMAGES
mariadb-galera   3/3     58m   mariadb,agent   mariadb:11.0.3,ghcr.io/mariadb-operator/mariadb-operator:v0.0.26

kubectl get pods -o wide
NAME                                        READY   STATUS    RESTARTS   AGE   IP           NODE          NOMINATED NODE   READINESS GATES
mariadb-galera-0                            2/2     Running   0          58m   10.244.2.4   mdb-worker3   <none>           <none>
mariadb-galera-1                            2/2     Running   0          58m   10.244.1.9   mdb-worker2   <none>           <none>
mariadb-galera-2                            2/2     Running   0          58m   10.244.5.4   mdb-worker4   <none>           <none>

Up and running. Let's now proceed with simulating a Galera cluster failure by deleting all the Pods at the same time:

kubectl delete pods -l app.kubernetes.io/instance=mariadb-galera
pod "mariadb-galera-0" deleted
pod "mariadb-galera-1" deleted
pod "mariadb-galera-2" deleted

After some time, we will see the MariaDB entering a non Ready state:

kubectl get mariadb mariadb-galera
NAME             READY   STATUS             PRIMARY POD             AGE
mariadb-galera   False   Galera not ready   mariadb-galera-0        67m

kubectl get events --field-selector involvedObject.name=mariadb-galera --sort-by='.lastTimestamp'
LAST SEEN   TYPE      REASON                    OBJECT                       MESSAGE
...
48s         Warning   GaleraClusterNotHealthy   mariadb/mariadb-galera       Galera cluster is not healthy

kubectl get mariadb mariadb-galera -o jsonpath="{.status.conditions[?(@.type=='GaleraReady')]}"
{"lastTransitionTime":"2023-07-13T19:25:17Z","message":"Galera not ready","reason":"GaleraNotReady","status":"False","type":"GaleraReady"}

Eventually, the operator will kick in and recover the Galera cluster:

kubectl get events --field-selector involvedObject.name=mariadb-galera --sort-by='.lastTimestamp'
LAST SEEN   TYPE      REASON                    OBJECT                       MESSAGE
...
16m         Warning   GaleraClusterNotHealthy   mariadb/mariadb-galera       Galera cluster is not healthy
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-2'
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-1'
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-0'
16m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-1'
16m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-2'
17m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-0'
17m         Normal    GaleraClusterBootstrap    mariadb/mariadb-galera       Bootstrapping Galera cluster in Pod 'mariadb-galera-2'
20m         Normal    GaleraClusterHealthy      mariadb/mariadb-galera       Galera cluster is healthy

kubectl get mariadb mariadb-galera -o jsonpath="{.status.galeraRecovery}"
{"bootstrap":{"pod":"mariadb-galera-2","time":"2023-07-13T19:25:28Z"},"recovered":{"mariadb-galera-0":{"seqno":3,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285"},"mariadb-galera-1":{"seqno":3,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285"},"mariadb-galera-2":{"seqno":3,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285"}},"state":{"mariadb-galera-0":{"safeToBootstrap":false,"seqno":-1,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285","version":"2.1"},"mariadb-galera-1":{"safeToBootstrap":false,"seqno":-1,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285","version":"2.1"},"mariadb-galera-2":{"safeToBootstrap":false,"seqno":-1,"uuid":"bf00b9c3-21a9-11ee-984f-9ba9ff0e9285","version":"2.1"}}}

Finally, the MariaDB resource will become Ready and your Galera cluster will be operational again:

kubectl get mariadb mariadb-galera -o jsonpath="{.status.conditions[?(@.type=='GaleraReady')]}"
{"lastTransitionTime":"2023-07-13T19:27:51Z","message":"Galera ready","reason":"GaleraReady","status":"True","type":"GaleraReady"}

kubectl get mariadb mariadb-galera
NAME             READY   STATUS    PRIMARY POD   AGE
mariadb-galera   True    Running   All           82m

Troubleshooting

The aim of this section is showing you how to diagnose your Galera cluster when something goes wrong. In this situations, observability is a key factor to understand the problem, so we recommend following these steps before jumping into debugging the problem.

Inspect MariaDB status conditions.

kubectl get mariadb mariadb-galera -o jsonpath="{.status}"
{"conditions":[{"lastTransitionTime":"2023-08-05T14:58:57Z","message":"Galera not ready","reason":"GaleraNotReady","status":"False","type":"Ready"},{"lastTransitionTime":"2023-08-05T14:58:57Z","message":"Galera not ready","reason":"GaleraNotReady","status":"False","type":"GaleraReady"},{"lastTransitionTime":"2023-08-03T19:21:16Z","message":"Galera configured","reason":"GaleraConfigured","status":"True","type":"GaleraConfigured"}],"currentPrimary":"All","galeraRecovery":{"bootstrap":{"pod":"mariadb-galera-1","time":"2023-08-05T14:59:18Z"},"recovered":{"mariadb-galera-0":{"seqno":17,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9"},"mariadb-galera-1":{"seqno":17,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9"},"mariadb-galera-2":{"seqno":16,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9"}},"state":{"mariadb-galera-0":{"safeToBootstrap":false,"seqno":-1,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9","version":"2.1"},"mariadb-galera-1":{"safeToBootstrap":false,"seqno":-1,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9","version":"2.1"},"mariadb-galera-2":{"safeToBootstrap":false,"seqno":-1,"uuid":"6ea235ec-3232-11ee-8152-4af03d2c43a9","version":"2.1"}}}}

Make sure network connectivity is fine by checking that you have an Endpoint per Pod in your Galera cluster.

kubectl get endpoints mariadb-galera-internal -o yaml
apiVersion: v1
kind: Endpoints
metadata:
  name: mariadb-internal
subsets:
- addresses:
  - hostname: mariadb-1
    ip: 10.255.140.181
    nodeName: k8s-worker-1
    targetRef:
      kind: Pod
      name: mariadb-1
      namespace: mariadb
  - hostname: mariadb-2
    ip: 10.255.20.156
    nodeName: k8s-worker-2
    targetRef:
      kind: Pod
      name: mariadb-2
      namespace: mariadb
  - hostname: mariadb-0
    ip: 10.255.214.164
    nodeName: k8s-worker-0
    targetRef:
      kind: Pod
      name: mariadb-0
      namespace: mariadb
  ports:
  - name: sst
    port: 4568
    protocol: TCP
  - name: ist
    port: 4567
    protocol: TCP
  - name: mariadb
    port: 3306
    protocol: TCP
  - name: agent
    port: 5555
    protocol: TCP
  - name: cluster
    port: 4444
    protocol: TCP

Check the events associated with the MariaDB object, as they provide significant insights for diagnosis, particularly within the context of cluster recovery.

kubectl get events --field-selector involvedObject.name=mariadb-galera --sort-by='.lastTimestamp'
LAST SEEN   TYPE      REASON                    OBJECT                       MESSAGE
...
16m         Warning   GaleraClusterNotHealthy   mariadb/mariadb-galera       Galera cluster is not healthy
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-2'
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-1'
16m         Normal    GaleraPodStateFetched     mariadb/mariadb-galera       Galera state fetched in Pod 'mariadb-galera-0'
16m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-1'
16m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-2'
17m         Normal    GaleraPodRecovered        mariadb/mariadb-galera       Recovered Galera sequence in Pod 'mariadb-galera-0'
17m         Normal    GaleraClusterBootstrap    mariadb/mariadb-galera       Bootstrapping Galera cluster in Pod 'mariadb-galera-2'
20m         Normal    GaleraClusterHealthy      mariadb/mariadb-galera       Galera cluster is healthy

Enable debug logs in mariadb-operator.

helm upgrade --install mariadb-operator mariadb-operator/mariadb-operator --set logLevel=debug
kubectl logs mariadb-operator-546c78f4f5-gq44k
{"level":"info","ts":1691090524.4911606,"logger":"galera.health","msg":"Checking Galera cluster health","controller":"statefulset","controllerGroup":"apps","controllerKind":"StatefulSet","statefulSet":{"name":"mariadb-galera","namespace":"default"},"namespace":"default","name":"mariadb-galera","reconcileID":"098620db-4486-45cc-966a-9f3fec0d165e"}
{"level":"debug","ts":1691090524.4911761,"logger":"galera.health","msg":"StatefulSet ready replicas","controller":"statefulset","controllerGroup":"apps","controllerKind":"StatefulSet","statefulSet":{"name":"mariadb-galera","namespace":"default"},"namespace":"default","name":"mariadb-galera","reconcileID":"098620db-4486-45cc-966a-9f3fec0d165e","replicas":1}

Get the logs of all the MariaDB Pod containers, not only of the main mariadb container but also the agent and init ones.

kubectl logs mariadb-galera-0 -c init
{"level":"info","ts":1691090778.5239124,"msg":"Starting init"}
{"level":"info","ts":1691090778.5305626,"msg":"Configuring Galera"}
{"level":"info","ts":1691090778.5307593,"msg":"Already initialized. Init done"}

kubectl logs mariadb-galera-0 -c agent
{"level":"info","ts":1691090779.3193653,"logger":"server","msg":"server listening","addr":":5555"}
2023/08/03 19:26:28 "POST http://mariadb-galera-0.mariadb-galera-internal.default.svc.cluster.local:5555/api/recovery HTTP/1.1" from 10.244.4.2:39162 - 200 58B in 4.112086ms
2023/08/03 19:26:28 "DELETE http://mariadb-galera-0.mariadb-galera-internal.default.svc.cluster.local:5555/api/recovery HTTP/1.1" from 10.244.4.2:39162 - 200 0B in 883.544µs

kubectl logs mariadb-galera-0 -c mariadb
2023-08-03 19:27:10 0 [Note] WSREP: Member 2.0 (mariadb-galera-0) synced with group.
2023-08-03 19:27:10 0 [Note] WSREP: Processing event queue:...100.0% (1/1 events) complete.
2023-08-03 19:27:10 0 [Note] WSREP: Shifting JOINED -> SYNCED (TO: 6)
2023-08-03 19:27:10 2 [Note] WSREP: Server mariadb-galera-0 synced with group
2023-08-03 19:27:10 2 [Note] WSREP: Server status change joined -> synced
2023-08-03 19:27:10 2 [Note] WSREP: Synchronized with group, ready for connections

Once you are done with these steps, you will have the context required to jump ahead to the Common errors section to see if any of them matches your case. If they don't, feel free to open an issue or even a PR updating this document if you managed to resolve it.

Common errors

Permission denied writing Galera configuration

 Error writing Galera config: open /etc/mysql/mariadb.conf.d/0-galera.cnf: permission denied

This error is returned by the init container when it is unable to write the configuration file in the filesystem backed by the PVC. In particular, this has been raised by users using longhorn and rook as a storage provider, which in some cases require a specific user to write in the filesystem:

longhorn/longhorn#3549

The remediation is matching the user expected by the storage provider to be able to write in the PVC:

apiVersion: k8s.mariadb.com/v1alpha1
kind: MariaDB
metadata:
  name: mariadb-galera
spec:
  podSecurityContext:
    runAsUser: 0
...

Unauthorized error disabling bootstrap

Error reconciling Galera: error disabling bootstrap in Pod 0: unauthorized

This situation occurs when the mariadb-operator credentials passed to the agent as authentication are either invalid or the agent is unable to verify them. To confirm this, ensure that both the mariadb-operator and the MariaDB ServiceAccounts are able to create TokenReview objects:

kubectl auth can-i --list --as=system:serviceaccount:default:mariadb-operator | grep tokenreview
tokenreviews.authentication.k8s.io              []                                    []               [create]

kubectl auth can-i --list --as=system:serviceaccount:default:mariadb-galera | grep tokenreview
tokenreviews.authentication.k8s.io              []                                    []               [create]

If that's not the case, check that the following ClusterRole and ClusterRoleBindings are available in your cluster:

kubectl get clusterrole system:auth-delegator
NAME                    CREATED AT
system:auth-delegator   2023-08-03T19:12:37Z

kubectl get clusterrolebinding | grep mariadb | grep auth-delegator
mariadb-galera:auth-delegator                     ClusterRole/system:auth-delegator                                                  108m
mariadb-operator:auth-delegator                        ClusterRole/system:auth-delegator                                                  112m

mariadb-operator:auth-delegator is the ClusterRoleBinding bound to the mariadb-operator ServiceAccount which is created by the helm chart, so you can re-install the helm release in order to recreate it:

 helm upgrade --install mariadb-operator mariadb-operator/mariadb-operator

mariadb-galera:auth-delegator is the ClusterRoleBinding bound to the mariadb-galera ServiceAccount which is created on the flight by the operator as part of the reconciliation logic. You may check the mariadb-operator logs to see if there are any issues reconciling it.

Bear in mind that ClusterRoleBindings are cluster-wide resources that are not garbage collected when the MariaDB owner object is deleted, which means that creating and deleting MariaDBs could leave leftovers in your cluster. These leftovers can lead to RBAC misconfigurations, as the ClusterRoleBinding might not be pointing to the right ServiceAccount. To overcome this, you can override the ClusterRoleBinding name setting the spec.galera.agent.kubernetesAuth.authDelegatorRoleName field.

Timeout waiting for Pod to be Synced

Timeout waiting for Pod 'mariadb-galera-2' to be Synced

This error appears in the mariadb-operator logs when a Pod is in non synced state for a duration exceeding the spec.galera.recovery.podRecoveryTimeout. Just after, the operator will restart the Pod.

Increase this timeout if you consider that your Pod may take longer to recover.

Galera cluster bootstrap timed out

Galera cluster bootstrap timed out. Resetting recovery status

This is error is returned by the mariadb-operator after exceeding the spec.galera.recovery.clusterBootstrapTimeout when recovering the cluster. At this point, the operator will reset the recovered sequence numbers and start again from a clean state.

Increase this timeout if you consider that your Galera cluster may take longer to recover.

GitHub Issues

Here it is a list of Galera-related issues reported by mariadb-operator users which might shed some light in your investigation:

https://github.com/mariadb-operator/mariadb-operator/issues?q=label%3Agalera+

Reference

API reference
Example suite

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

GALERA.md

GALERA.md

High availability via Galera

Table of contents

Components

`MariaDB` configuration

Storage

Galera cluster recovery

Wsrep provider

IPv6 support

Quickstart

Troubleshooting

Common errors

Permission denied writing Galera configuration

Unauthorized error disabling bootstrap

Timeout waiting for Pod to be Synced

Galera cluster bootstrap timed out

GitHub Issues

Reference

Files

GALERA.md

Latest commit

History

GALERA.md

File metadata and controls

High availability via Galera

Table of contents

Components

MariaDB configuration

Storage

Galera cluster recovery

Wsrep provider

IPv6 support

Quickstart

Troubleshooting

Common errors

Permission denied writing Galera configuration

Unauthorized error disabling bootstrap

Timeout waiting for Pod to be Synced

Galera cluster bootstrap timed out

GitHub Issues

Reference

`MariaDB` configuration