Edit

Limits in Azure Database for PostgreSQL - Flexible Server

  • Article

APPLIES TO: Azure Database for PostgreSQL - Flexible Server

The following sections describe capacity and functional limits in Azure Database for PostgreSQL flexible server. If you want to learn about resource (compute, memory, or storage) tiers, see the Compute and storage article.

Maximum connections

The following table shows the default maximum number of connections for each pricing tier and vCore configuration. Azure Database for PostgreSQL flexible server reserves 15 connections for physical replication and monitoring of the Azure Database for PostgreSQL flexible server instance. Consequently, the value for maximum user connections listed in the table is reduced by 15 from the total maximum connections.

Product name vCores Memory size Maximum connections Maximum user connections
Burstable
B1ms 1 2 GiB 50 35
B2s 2 4 GiB 429 414
B2ms 2 8 GiB 859 844
B4ms 4 16 GiB 1,718 1,703
B8ms 8 32 GiB 3,437 3,422
B12ms 12 48 GiB 5,000 4,985
B16ms 16 64 GiB 5,000 4,985
B20ms 20 80 GiB 5,000 4,985
General Purpose
D2s_v3 / D2ds_v4 / D2ds_v5 / D2ads_v5 2 8 GiB 859 844
D4s_v3 / D4ds_v4 / D4ds_v5 / D4ads_v5 4 16 GiB 1,718 1,703
D8s_v3 / D8ds_V4 / D8ds_v5 / D8ads_v5 8 32 GiB 3,437 3,422
D16s_v3 / D16ds_v4 / D16ds_v5 / D16ads_v5 16 64 GiB 5,000 4,985
D32s_v3 / D32ds_v4 / D32ds_v5 / D32ads_v5 32 128 GiB 5,000 4,985
D48s_v3 / D48ds_v4 / D48ds_v5 / D48ads_v5 48 192 GiB 5,000 4,985
D64s_v3 / D64ds_v4 / D64ds_v5 / D64ads_v5 64 256 GiB 5,000 4,985
D96ds_v5 / D96ads_v5 96 384 GiB 5,000 4,985
Memory Optimized
E2s_v3 / E2ds_v4 / E2ds_v5 / E2ads_v5 2 16 GiB 1,718 1,703
E4s_v3 / E4ds_v4 / E4ds_v5 / E4ads_v5 4 32 GiB 3,437 3,422
E8s_v3 / E8ds_v4 / E8ds_v5 / E8ads_v5 8 64 GiB 5,000 4,985
E16s_v3 / E16ds_v4 / E16ds_v5 / E16ads_v5 16 128 GiB 5,000 4,985
E20ds_v4 / E20ds_v5 / E20ads_v5 20 160 GiB 5,000 4,985
E32s_v3 / E32ds_v4 / E32ds_v5 / E32ads_v5 32 256 GiB 5,000 4,985
E48s_v3 / E48ds_v4 / E48ds_v5 / E48ads_v5 48 384 GiB 5,000 4,985
E64s_v3 / E64ds_v4 / E64ds_v5 / E64ads_v5 64 432 GiB 5,000 4,985
E96ds_v5 / E96ads_v5 96 672 GiB 5,000 4,985

The reserved connection slots, presently at 15, could change. We advise regularly verifying the total reserved connections on the server. You calculate this number by summing the values of the reserved_connections and superuser_reserved_connections server parameters. The maximum number of available user connections is max_connections - (reserved_connections + superuser_reserved_connections).

The default value for the max_connections server parameter is calculated when you provision the instance of Azure Database for PostgreSQL flexible server, based on the product name that you select for its compute. Any subsequent changes of product selection to the compute that supports the flexible server won't have any effect on the default value for the max_connections server parameter of that instance. We recommend that whenever you change the product assigned to an instance, you also adjust the value for the max_connections parameter according to the values in the preceding table.

Changing the max_connections value

When you first set up your Azure Database for Postgres flexible server instance, it automatically decides the highest number of connections that it can handle concurrently. This number is based on your server's configuration and can't be changed.

However, you can use the max_connections setting to adjust how many connections are allowed at a particular time. After you change this setting, you need to restart your server for the new limit to start working.

Caution

Although it's possible to increase the value of max_connections beyond the default setting, we advise against it.

Instances might encounter difficulties when the workload expands and demands more memory. As the number of connections increases, memory usage also rises. Instances with limited memory might face issues such as crashes or high latency. Although a higher value for max_connections might be acceptable when most connections are idle, it can lead to significant performance problems after they become active.

If you need more connections, we suggest that you instead use PgBouncer, the built-in Azure solution for connection pool management. Use it in transaction mode. To start, we recommend that you use conservative values by multiplying the vCores within the range of 2 to 5. Afterward, carefully monitor resource utilization and application performance to ensure smooth operation. For detailed information on PgBouncer, see PgBouncer in Azure Database for PostgreSQL - Flexible Server.

When connections exceed the limit, you might receive the following error:

FATAL: sorry, too many clients already.

When you're using Azure Database for PostgreSQL flexible server for a busy database with a large number of concurrent connections, there might be a significant strain on resources. This strain can result in high CPU utilization, especially when many connections are established simultaneously and when connections have short durations (less than 60 seconds). These factors can negatively affect overall database performance by increasing the time spent on processing connections and disconnections.

Be aware that each connection in Azure Database for PostgreSQL flexible server, regardless of whether it's idle or active, consumes a significant amount of resources from your database. This consumption can lead to performance issues beyond high CPU utilization, such as disk and lock contention. The Number of Database Connections article on the PostgreSQL Wiki discusses this topic in more detail. To learn more, see Identify and solve connection performance in Azure Database for PostgreSQL flexible server.

Functional limitations

The following sections list considerations for what is and isn't supported in Azure Database for PostgreSQL flexible server.

Scale operations

  • At this time, scaling up the server storage requires a server restart.
  • You can scale server storage only in 2x increments. For details, see Compute and storage.

Storage

  • After you configure the storage size, you can't reduce it. You have to create a new server with the desired storage size, perform a manual dump and restore operation, and migrate your databases to the new server.
  • When the storage usage reaches 95% or if the available capacity is less than 5 GiB (whichever is more), the server is automatically switched to read-only mode to avoid errors associated with disk-full situations. In rare cases, if the rate of data growth outpaces the time it takes to switch to read-only mode, your server might still run out of storage. You can enable storage autogrow to avoid these issues and automatically scale your storage based on your workload demands.
  • We recommend setting alert rules for storage used or storage percent when they exceed certain thresholds so that you can proactively take action such as increasing the storage size. For example, you can set an alert if the storage percentage exceeds 80% usage.
  • If you're using logical replication, you must drop the logical replication slot in the primary server if the corresponding subscriber no longer exists. Otherwise, the write-ahead logging (WAL) files accumulate in the primary and fill up the storage. If the storage exceeds a certain threshold and if the logical replication slot isn't in use (because of an unavailable subscriber), Azure Database for PostgreSQL flexible server automatically drops that unused logical replication slot. That action releases accumulated WAL files and prevents your server from becoming unavailable because the storage is filled.
  • We don't support the creation of tablespaces. If you're creating a database, don't provide a tablespace name. Azure Database for PostgreSQL flexible server uses the default one that's inherited from the template database. It's unsafe to provide a tablespace like the temporary one, because we can't ensure that such objects will remain persistent after events like server restarts and high-availability (HA) failovers.

Networking

  • Moving in and out of a virtual network is currently not supported.
  • Combining public access with deployment in a virtual network is currently not supported.
  • Firewall rules aren't supported on virtual networks. You can use network security groups instead.
  • Public access database servers can connect to the public internet; for example, through postgres_fdw. You can't restrict this access. Servers in virtual networks can have restricted outbound access through network security groups.

High availability

Availability zones

  • Manually moving servers to a different availability zone is currently not supported. However, by using the preferred availability zone as the standby zone, you can turn on HA. After you establish the standby zone, you can fail over to it and then turn off HA.

Postgres engine, extensions, and PgBouncer

  • Postgres 10 and older versions aren't supported, because the open-source community retired them. If you must use one of these versions, you need to use the Azure Database for PostgreSQL single server option, which supports the older major versions 9.5, 9.6, and 10.
  • Azure Database for PostgreSQL flexible server supports all contrib extensions and more. For more information, see PostgreSQL extensions.
  • The built-in PgBouncer connection pooler is currently not available for Burstable servers.

Stop/start operations

  • After you stop the Azure Database for PostgreSQL flexible server instance, it automatically starts after 7 days.

Scheduled maintenance

  • You can change the custom maintenance window to any day/time of the week. However, any changes that you make after receiving the maintenance notification will have no impact on the next maintenance. Changes take effect only with the following monthly scheduled maintenance.

Server backups

  • The system manages backups. There's currently no way to run backups manually. We recommend using pg_dump instead.

  • The first snapshot is a full backup, and consecutive snapshots are differential backups. The differential backups back up only the changed data since the last snapshot backup.

    For example, if the size of your database is 40 GB and your provisioned storage is 64 GB, the first snapshot backup will be 40 GB. Now, if you change 4 GB of data, the next differential snapshot backup size will be only 4 GB. The transaction logs (write-ahead logs) are separate from the full and differential backups, and they're archived continuously.

Server restoration

  • When you're using the point-in-time restore (PITR) feature, the new server is created with the same compute and storage configurations as the server that it's based on.
  • Database servers in virtual networks are restored into the same virtual networks when you restore from a backup.
  • The new server created during a restore doesn't have the firewall rules that existed on the original server. You need to create firewall rules separately for the new server.
  • Restore to a different subscription isn't supported. As a workaround, you can restore the server within the same subscription and then migrate the restored server to a different subscription.

Next steps