Edit

Scheduled maintenance in Azure Cosmos DB for PostgreSQL

  • Article

APPLIES TO: Azure Cosmos DB for PostgreSQL (powered by the Citus database extension to PostgreSQL)

Azure Cosmos DB for PostgreSQL does periodic maintenance to keep your managed database secure, stable, and up-to-date. During maintenance, all nodes in the cluster get new features, updates, and patches.

The key features of scheduled maintenance for Azure Cosmos DB for PostgreSQL are:

  • Updates are applied at the same time on all nodes in the cluster
  • Notifications about upcoming maintenance are posted to Azure Service Health five days in advance
  • Usually there are at least 30 days between successful maintenance events for a cluster
  • Preferred day of the week and time window within that day for maintenance start can be defined for each cluster individually

Selecting a maintenance window and notification about upcoming maintenance

You can schedule maintenance during a specific day of the week and a time window within that day. Or you can let the system pick a day and a time window for you automatically. Either way, the system will alert you five days before running any maintenance. The system will also let you know when maintenance is started, and when it's successfully completed.

Notifications about upcoming scheduled maintenance are posted to Azure Service Health and can be:

  • Emailed to a specific address
  • Emailed to an Azure Resource Manager Role
  • Sent in a text message (SMS) to mobile devices
  • Pushed as a notification to an Azure app
  • Delivered as a voice message

When specifying preferences for the maintenance schedule, you can pick a day of the week and a time window. If you don't specify, the system will pick times between 11pm and 7am in your cluster's region time. You can define different schedules for each cluster in your Azure subscription.

Important

Normally there are at least 30 days between successful scheduled maintenance events for a cluster.

However, in case of a critical emergency update such as a severe vulnerability, the notification window could be shorter than five days. The critical update may be applied to your server even if a successful scheduled maintenance was performed in the last 30 days.

You can update scheduling settings at any time. If there's maintenance scheduled for your cluster and you update the schedule, the pre-existing events will be rescheduled.

If maintenance fails or gets canceled, the system will create a notification. It will try maintenance again according to current scheduling settings, and notify you five days before the next maintenance event.

Next steps