Danger Zone settings
Operations in the Danger Zone section of the Settings tab will immediately affect the availability of a cluster. The CLI and API also provide mechanisms for performing each of these actions.
Restart Postgres service
This will immediately restart the Postgres service on your cluster. Once the service has restarted the Postgres shared buffers cache will be empty or "cold", which may result in period of slow query execution times for queries with read I/O requests for data that is not in the file system's cache.
Restart cluster
This will immediately restart the underlying server instance. Once the instance and service are back online, both the Postgres service's shared buffers cache and the instance's file system cache will be empty or "cold". This will cause all read I/O requests for queries to be satisfied by disk reads, resulting in a definite period of slow query execution times.
Info
Restarting a Postgres service or cluster will rarely improve performance due to the post-restart caches being cold. Please reach out to support if you would like guidance about database-related performance issues.
Suspend cluster
This action suspends a cluster, removing the underlying server but retaining its disk image in storage so it can later be resumed. There is no charge for a server instance while a cluster is suspended, but storage costs will continue to accrue as the data and backups are still stored on disk.
Resume cluster
Resumes normal operation of a currently suspended cluster by recreating the server instance and restarting the Postgres service. Normal billing, including instance charges, resumes as well.
Cluster protection
You can prevent a cluster from being destroyed or suspended by enabling the Protected switch on the Cluster Overview tab. More details on cluster protection can be found in the Cluster settings documentation.
Destroy cluster
This will stop and completely remove the cluster and its data from your account. Backups for destroyed clusters are maintained for 5 days after they have been destroyed, as long as the cluster has been running for at least 5 days. If you need to obtain a backup for a destroyed cluster, or restore a destroyed cluster during this 5-day period, please reach out to support.