Skip to main content

How to restart a cluster

This guide explains how to restart a PostgreSQL cluster from the Hikube console, for example when an instance no longer behaves normally.

Prerequisites​

  • A PostgreSQL cluster with the status Ready in your project
  • The right to edit this cluster

What happens during the restart​

The restart is rolling: the replicas restart first, then the primary.

  • Reads keep being served throughout the operation.
  • Writes are paused for about forty seconds while the primary restarts.
  • Data is kept.

On a cluster with 1 replica, no other instance takes over: the database is unavailable while it restarts.

warning

Plan the restart in a maintenance window or outside peak hours, and check that your applications can reconnect and retry a rejected write.

Steps​

1. Open the cluster page​

  1. Open DB & Messaging → PostgreSQL.
  2. In the PostgreSQL Clusters list, open the cluster.

2. Start the restart​

  1. Click Restart, next to the Edit and Delete buttons.
  2. The "Restart <name>?" window recalls the effect of the restart. Confirm with Restart.

The "Restart started" message confirms that the request was accepted. The restart continues in the background.

note

A cluster that is still being created has nothing to restart yet: the request is refused. Wait a few minutes after the cluster reaches the status Ready, while it finishes starting, then start the restart again.

Verification​

The cluster status does not change during a restart: it keeps showing Ready. To confirm the restart, check the start time of the primary:

psql "host=<host> port=5432 dbname=<database> user=<user> sslmode=require" \
  -c "SELECT pg_postmaster_start_time();"

The returned time matches the moment of the restart.

Further reading​