You are looking at the documentation of a prior release. To read the documentation of the latest release, please visit here.

New to KubeDB? Please start here.

Postgres

What is Postgres

Postgres is a Kubernetes Custom Resource Definitions (CRD). It provides declarative configuration for PostgreSQL in a Kubernetes native way. You only need to describe the desired database configuration in a Postgres object, and the KubeDB operator will create Kubernetes objects in the desired state for you.

Postgres Spec

As with all other Kubernetes objects, a Postgres needs apiVersion, kind, and metadata fields. It also needs a .spec section.

Below is an example Postgres object.

apiVersion: kubedb.com/v1alpha2
kind: Postgres
metadata:
  name: p1
  namespace: demo
spec:
  version: "13.2"
  replicas: 2
  standbyMode: Hot
  streamingMode: Asynchronous
  leaderElection:
    leaseDurationSeconds: 15
    renewDeadlineSeconds: 10
    retryPeriodSeconds: 2
  authSecret:
    name: p1-auth
  storageType: "Durable"
  storage:
    storageClassName: standard
    accessModes:
    - ReadWriteOnce
    resources:
      requests:
        storage: 1Gi
  init:
    script:
      configMap:
        name: pg-init-script
  monitor:
    agent: prometheus.io/operator
    prometheus:
      serviceMonitor:
        labels:
          app: kubedb
        interval: 10s
  configSecret:
    name: pg-custom-config
  podTemplate:
    metadata:
      annotations:
        passMe: ToDatabasePod
    controller:
      annotations:
        passMe: ToStatefulSet
    spec:
      serviceAccountName: my-custom-sa
      schedulerName: my-scheduler
      nodeSelector:
        disktype: ssd
      imagePullSecrets:
      - name: myregistrykey
      env:
      - name: POSTGRES_DB
        value: pgdb
      resources:
        requests:
          memory: "64Mi"
          cpu: "250m"
        limits:
          memory: "128Mi"
          cpu: "500m"
  serviceTemplates:
  - alias: primary
    metadata:
      annotations:
        passMe: ToService
    spec:
      type: NodePort
      ports:
      - name:  http
        port:  5432
  - alias: standby
    metadata:
      annotations:
        passMe: ToReplicaService
    spec:
      type: NodePort
      ports:
      - name:  http
        port:  5432
  terminationPolicy: "Halt"

spec.version

spec.version is a required field that specifies the name of the PostgresVersion crd where the docker images are specified. Currently, when you install KubeDB, it creates the following PostgresVersion resources,

$ kubectl get pgversion
NAME       VERSION   DB_IMAGE                   DEPRECATED   AGE
10.2       10.2      kubedb/postgres:10.2       true         44m
10.2-v1    10.2      kubedb/postgres:10.2-v2    true         44m
10.2-v2    10.2      kubedb/postgres:10.2-v3                 44m
10.2-v3    10.2      kubedb/postgres:10.2-v4                 44m
10.2-v4    10.2      kubedb/postgres:10.2-v5                 44m
10.2-v5    10.2      kubedb/postgres:10.2-v6                 44m
10.6       10.6      kubedb/postgres:10.6                    44m
10.6-v1    10.6      kubedb/postgres:10.6-v1                 44m
10.6-v2    10.6      kubedb/postgres:10.6-v2                 44m
10.6-v3    10.6      kubedb/postgres:10.6-v3                 44m
11.1       11.1      kubedb/postgres:11.1                    44m
11.1-v1    11.1      kubedb/postgres:11.1-v1                 44m
11.1-v2    11.1      kubedb/postgres:11.1-v2                 44m
11.1-v3    11.1      kubedb/postgres:11.1-v3                 44m
11.2       11.2      kubedb/postgres:11.2                    44m
11.2-v1    11.2      kubedb/postgres:11.2-v1                 44m
9.6        9.6       kubedb/postgres:9.6        true         44m
9.6-v1     9.6       kubedb/postgres:9.6-v2     true         44m
9.6-v2     9.6       kubedb/postgres:9.6-v3                  44m
9.6-v3     9.6       kubedb/postgres:9.6-v4                  44m
9.6-v4     9.6       kubedb/postgres:9.6-v5                  44m
9.6-v5     9.6       kubedb/postgres:9.6-v6                  44m
9.6.7      9.6.7     kubedb/postgres:9.6.7      true         44m
9.6.7-v1   9.6.7     kubedb/postgres:9.6.7-v2   true         44m
9.6.7-v2   9.6.7     kubedb/postgres:9.6.7-v3                44m
9.6.7-v3   9.6.7     kubedb/postgres:9.6.7-v4                44m
9.6.7-v4   9.6.7     kubedb/postgres:9.6.7-v5                44m
9.6.7-v5   9.6.7     kubedb/postgres:9.6.7-v6                44m

spec.replicas

spec.replicas specifies the total number of primary and standby nodes in Postgres database cluster configuration. One pod is selected as Primary and others act as standby replicas. KubeDB uses PodDisruptionBudget to ensure that majority of the replicas are available during voluntary disruptions.

To learn more about how to setup a HA PostgreSQL cluster in KubeDB, please visit here.

spec.standbyMode

spec.standby is an optional field that specifies the standby mode (Warm / Hot) to use for standby replicas. In hot standby mode, standby replicas can accept connection and run read-only queries. In warm standby mode, standby replicas can’t accept connection and only used for replication purpose.

spec.streamingMode

spec.streamingMode is an optional field that specifies the streaming mode (Synchronous / Asynchronous) of the standby replicas. KubeDB currently supports only Asynchronous streaming mode.

spec.leaderElection

There are three fields under Postgres CRD’s spec.leaderElection. These values defines how fast the leader election can happen.

  • leaseDurationSeconds: This is the duration in seconds that non-leader candidates will wait to force acquire leadership. This is measured against time of last observed ack. Default 15 sec.
  • renewDeadlineSeconds: This is the duration in seconds that the acting master will retry refreshing leadership before giving up. Normally, LeaseDuration * 2 / 3. Default 10 sec.
  • retryPeriodSeconds: This is the duration in seconds the LeaderElector clients should wait between tries of actions. Normally, LeaseDuration / 3. Default 2 sec.

If the Cluster machine is powerful, user can reduce the times. But, Do not make it so little, in that case Postgres will restarts very often.

spec.authSecret

spec.authSecret is an optional field that points to a Secret used to hold credentials for postgres database. If not set, KubeDB operator creates a new Secret with name {postgres-name}-auth that hold username and password for postgres database.

If you want to use an existing or custom secret, please specify that when creating the Postgres object using spec.authSecret.name. This Secret should contain superuser username as POSTGRES_USER key and superuser password as POSTGRES_PASSWORD key. Secrets provided by users are not managed by KubeDB, and therefore, won’t be modified or garbage collected by the KubeDB operator (version >= 0.13.0).

Example:

$ kubectl create secret generic p1-auth -n demo \
--from-literal=POSTGRES_USER=not@user \
--from-literal=POSTGRES_PASSWORD=not@secret
secret "p1-auth" created
$ kubectl get secret -n demo p1-auth -o yaml
apiVersion: v1
data:
  POSTGRES_PASSWORD: bm90QHNlY3JldA==
  POSTGRES_USER: bm90QHVzZXI=
kind: Secret
metadata:
  creationTimestamp: 2018-09-03T11:25:39Z
  name: p1-auth
  namespace: demo
  resourceVersion: "1677"
  selfLink: /api/v1/namespaces/demo/secrets/p1-auth
  uid: 15b3e8a1-af6c-11e8-996d-0800270d7bae
type: Opaque

spec.storageType

spec.storageType is an optional field that specifies the type of storage to use for database. It can be either Durable or Ephemeral. The default value of this field is Durable. If Ephemeral is used then KubeDB will create Postgres database using emptyDir volume. In this case, you don’t have to specify spec.storage field.

spec.storage

If you don’t set spec.storageType: to Ephemeral then spec.storage field is required. This field specifies the StorageClass of PVCs dynamically allocated to store data for the database. This storage spec will be passed to the StatefulSet created by KubeDB operator to run database pods. You can specify any StorageClass available in your cluster with appropriate resource requests.

  • spec.storage.storageClassName is the name of the StorageClass used to provision PVCs. PVCs don’t necessarily have to request a class. A PVC with its storageClassName set equal to "" is always interpreted to be requesting a PV with no class, so it can only be bound to PVs with no class (no annotation or one set equal to “”). A PVC with no storageClassName is not quite the same and is treated differently by the cluster depending on whether the DefaultStorageClass admission plugin is turned on.
  • spec.storage.accessModes uses the same conventions as Kubernetes PVCs when requesting storage with specific access modes.
  • spec.storage.resources can be used to request specific quantities of storage. This follows the same resource model used by PVCs.

To learn how to configure spec.storage, please visit the links below:

spec.init

spec.init is an optional section that can be used to initialize a newly created Postgres database. PostgreSQL databases can be initialized from these three ways:

  1. Initialize from Script
  2. Initialize from Snapshot

Initialize via Script

To initialize a PostgreSQL database using a script (shell script, db migrator, etc.), set the spec.init.script section when creating a Postgres object. script must have the following information:

Below is an example showing how a script from a configMap can be used to initialize a PostgreSQL database.

apiVersion: kubedb.com/v1alpha2
kind: Postgres
metadata:
  name: postgres-db
  namespace: demo
spec:
  version: "13.2"
  init:
    script:
      configMap:
        name: pg-init-script

In the above example, Postgres will execute provided script once the database is running. For more details tutorial on how to initialize from script, please visit here.

spec.monitor

PostgreSQL managed by KubeDB can be monitored with builtin-Prometheus and Prometheus operator out-of-the-box. To learn more,

spec.configSecret

spec.configSecret is an optional field that allows users to provide custom configuration for PostgreSQL. This field accepts a VolumeSource. You can use any kubernetes supported volume source such as configMap, secret, azureDisk etc. To learn more about how to use a custom configuration file see here.

spec.podTemplate

KubeDB allows providing a template for database pod through spec.podTemplate. KubeDB operator will pass the information provided in spec.podTemplate to the StatefulSet created for Postgres database.

KubeDB accept following fields to set in spec.podTemplate:

  • metadata
    • annotations (pod’s annotation)
  • controller
    • annotations (statefulset’s annotation)
  • spec:
    • serviceAccountName
    • env
    • resources
    • initContainers
    • imagePullSecrets
    • nodeSelector
    • affinity
    • schedulerName
    • tolerations
    • priorityClassName
    • priority
    • securityContext
    • livenessProbe
    • readinessProbe
    • lifecycle

Uses of some field of spec.podTemplate is described below,

spec.podTemplate.spec.serviceAccountName

serviceAccountName is an optional field supported by KubeDB Operator (version 0.13.0 and higher) that can be used to specify a custom service account to fine tune role based access control.

If this field is left empty, the KubeDB operator will create a service account name matching Postgres crd name. Role and RoleBinding that provide necessary access permissions will also be generated automatically for this service account.

If a service account name is given, but there’s no existing service account by that name, the KubeDB operator will create one, and Role and RoleBinding that provide necessary access permissions will also be generated for this service account.

If a service account name is given, and there’s an existing service account by that name, the KubeDB operator will use that existing service account. Since this service account is not managed by KubeDB, users are responsible for providing necessary access permissions manually. Follow the guide here to grant necessary permissions in this scenario.

spec.podTemplate.spec.env

spec.podTemplate.spec.env is an optional field that specifies the environment variables to pass to the Postgres docker image. To know about supported environment variables, please visit here.

Note that, the KubeDB operator does not allow POSTGRES_USER and POSTGRES_PASSWORD environment variable to set in spec.podTemplate.spec.env. If you want to set the superuser username and password, please use spec.authSecret instead described earlier.

If you try to set POSTGRES_USER or POSTGRES_PASSWORD environment variable in Postgres crd, KubeDB operator will reject the request with following error,

Error from server (Forbidden): error when creating "./postgres.yaml": admission webhook "postgres.validators.kubedb.com" denied the request: environment variable POSTGRES_PASSWORD is forbidden to use in Postgres spec

Also, note that KubeDB does not allow to update the environment variables as updating them does not have any effect once the database is created. If you try to update environment variables, KubeDB operator will reject the request with following error,

Error from server (BadRequest): error when applying patch:
...
for: "./postgres.yaml": admission webhook "postgres.validators.kubedb.com" denied the request: precondition failed for:
...
At least one of the following was changed:
    apiVersion
    kind
    name
    namespace
    spec.standby
    spec.streaming
    spec.authSecret
    spec.storageType
    spec.storage
    spec.podTemplate.spec.nodeSelector
    spec.init

spec.podTemplate.spec.imagePullSecrets

spec.podTemplate.spec.imagePullSecrets is an optional field that points to secrets to be used for pulling docker image if you are using a private docker registry. For more details on how to use private docker registry, please visit here.

spec.podTemplate.spec.nodeSelector

spec.podTemplate.spec.nodeSelector is an optional field that specifies a map of key-value pairs. For the pod to be eligible to run on a node, the node must have each of the indicated key-value pairs as labels (it can have additional labels as well). To learn more, see here .

spec.podTemplate.spec.resources

spec.podTemplate.spec.resources is an optional field. This can be used to request compute resources required by the database pods. To learn more, visit here.

spec.serviceTemplate

KubeDB creates two different services for each Postgres instance. One of them is a master service named <postgres-name> and points to the Postgres Primary pod/node. Another one is a replica service named <postgres-name>-replicas and points to Postgres replica pods/nodes.

These master and replica services can be customized using spec.serviceTemplate and spec.replicaServiceTemplate respectively.

You can provide template for the master service using spec.serviceTemplate. This will allow you to set the type and other properties of the service. If spec.serviceTemplate is not provided, KubeDB will create a master service of type ClusterIP with minimal settings.

KubeDB allows following fields to set in spec.serviceTemplate:

  • metadata:
    • annotations
  • spec:
    • type
    • ports
    • clusterIP
    • externalIPs
    • loadBalancerIP
    • loadBalancerSourceRanges
    • externalTrafficPolicy
    • healthCheckNodePort
    • sessionAffinityConfig

See here to understand these fields in detail.

spec.replicaServiceTemplate

You can provide template for the replica service using spec.replicaServiceTemplate. If spec.replicaServiceTemplate is not provided, KubeDB will create a replica service of type ClusterIP with minimal settings.

The fileds of spec.replicaServiceTemplate is similar to spec.serviceTemplate, that is:

  • metadata:
    • annotations
  • spec:
    • type
    • ports
    • clusterIP
    • externalIPs
    • loadBalancerIP
    • loadBalancerSourceRanges
    • externalTrafficPolicy
    • healthCheckNodePort

See here to understand these fields in detail.

spec.terminationPolicy

terminationPolicy gives flexibility whether to nullify(reject) the delete operation of Postgres crd or which resources KubeDB should keep or delete when you delete Postgres crd. KubeDB provides following four termination policies:

  • DoNotTerminate
  • Halt
  • Delete (Default)
  • WipeOut

When terminationPolicy is DoNotTerminate, KubeDB takes advantage of ValidationWebhook feature in Kubernetes 1.9.0 or later clusters to provide safety from accidental deletion of database. If admission webhook is enabled, KubeDB prevents users from deleting the database as long as the spec.terminationPolicy is set to DoNotTerminate.

Following table show what KubeDB does when you delete Postgres crd for different termination policies,

BehaviorDoNotTerminateHaltDeleteWipeOut
1. Block Delete operation
2. Create Dormant Database
3. Delete StatefulSet
4. Delete Services
5. Delete PVCs
6. Delete Secrets
7. Delete Snapshots
8. Delete Snapshot data from bucket

If you don’t specify spec.terminationPolicy KubeDB uses Halt termination policy by default.

Next Steps