Use automatic seeding to initialize an Always On availability group

Applies to: SQL Server

SQL Server 2016 introduced automatic seeding of availability groups. When you create an availability group with automatic seeding, SQL Server automatically creates the secondary replicas for every database in the group. You no longer have to manually back up and restore secondary replicas. To enable automatic seeding, create the availability group with T-SQL or use the latest version of SQL Server Management Studio.

For background information, see Automatic seeding for secondary replicas.

Prerequisites

In SQL Server 2016, automatic seeding requires that the data and log file path is the same on every SQL Server instance participating in the availability group. In SQL Server 2017, you can use different paths, however Microsoft recommends using the same paths when all replicas are hosted on the same platform (for example either Windows or Linux). Cross-platform availability groups have different paths for the replicas. For details, see Disk layout.

Availability group seeding communicates over the database mirroring endpoint. Open inbound firewall rules to the mirroring endpoint port on each server.

Databases in an availability group must be in full recovery model. The database needs to have a current full backup and transaction log backup. These backup files are not used for automatic seeding, but they are required before including the database in an availability group.

Create availability group with automatic seeding

To create an availability group with automatic seeding, set SEEDING_MODE=AUTOMATIC.

The following example creates an availability group on a two-node Windows Server failover cluster. Before running the scripts, update the values for your environment.

  1. Create the endpoints. Each server needs an endpoint. The following script creates an endpoint that uses TCP port 5022 for the listener. Set <endpoint_name> and LISTENER_PORT to match your environment and run the script on both servers:

    CREATE ENDPOINT [<endpoint_name>] 
        STATE=STARTED
        AS TCP (LISTENER_PORT = 5022, LISTENER_IP = ALL)
        FOR DATA_MIRRORING (
            ROLE = ALL, 
            AUTHENTICATION = WINDOWS NEGOTIATE, 
            ENCRYPTION = REQUIRED ALGORITHM AES
            )
    GO
    
  2. Create the availability group. The following script creates the availability group. Update the values in angle brackets <> for the group name, server names, and domain names, and run it on the primary instance of SQL Server.

    CREATE AVAILABILITY GROUP [<availability_group_name>]
        FOR DATABASE db1
        REPLICA ON'<*primary_server*>'
        WITH (ENDPOINT_URL = N'TCP://<primary_server>.<fully_qualified_domain_name>:5022', 
            FAILOVER_MODE = AUTOMATIC, 
            AVAILABILITY_MODE = SYNCHRONOUS_COMMIT, 
            BACKUP_PRIORITY = 50, 
            SECONDARY_ROLE(ALLOW_CONNECTIONS = NO), 
            SEEDING_MODE = AUTOMATIC),
        N'<secondary_server>' WITH (ENDPOINT_URL = N'TCP://<secondary_server>.<fully_qualified_domain_name>:5022', 
            FAILOVER_MODE = AUTOMATIC, 
            AVAILABILITY_MODE = SYNCHRONOUS_COMMIT, 
            BACKUP_PRIORITY = 50, 
            SECONDARY_ROLE(ALLOW_CONNECTIONS = NO), 
            SEEDING_MODE = AUTOMATIC);
    GO
    
  3. Join the secondary server instance to the availability group and grant create database permission to the availability group. Update the following script, replace the values in angle brackets <> for your environment, and run it on the secondary replica instance of SQL Server:

    ALTER AVAILABILITY GROUP [<availability_group_name>] JOIN
    GO  
    ALTER AVAILABILITY GROUP [<availability_group_name>] GRANT CREATE ANY DATABASE
    GO
    

SQL Server automatically creates the database replica on the secondary server. If the database is large, it may take some time to complete synchronization of the database. If a database is in an availability group that is configured for automatic seeding, you can query sys.dm_hadr_automatic_seeding system view to monitor the seeding progress. The following query returns one row for every database that is in an availability group configured for automatic seeding.

SELECT start_time,
    ag.name,
    db.database_name,
    current_state,
    performed_seeding,
    failure_state,
    failure_state_desc
FROM sys.dm_hadr_automatic_seeding autos 
    JOIN sys.availability_databases_cluster db 
        ON autos.ag_db_id = db.group_database_id
    JOIN sys.availability_groups ag 
        ON autos.ag_id = ag.group_id

Prevent automatic seeding after an availability group

To temporarily prevent the primary replica from seeding more databases to the secondary replica, you can deny the availability group permission to create databases. Run the following query on the instance that hosts the secondary replica in order to deny the availability group permission to create replica databases.

ALTER AVAILABILITY GROUP [<availability_group_name>] 
    DENY CREATE ANY DATABASE
GO

Enable automatic seeding on an existing availability group

You can set automatic seeding on an existing database. The following command changes an availability group to use automatic seeding. Run the following command on the primary replica.

ALTER AVAILABILITY GROUP [<availability_group_name>] 
    MODIFY REPLICA ON '<secondary_node>' 
    WITH (SEEDING_MODE = AUTOMATIC)
GO

The preceding command forces a database to restart seeding if needed. For example, if seeding fails because of insufficient disk space on the secondary replica, run ALTER AVAILABILITY GROUP ... WITH (SEEDING_MODE=AUTOMATIC) to restart seeding after you have added free space.

Stop automatic seeding

To stop automatic seeding for an availability group, run the following script on the primary replica:

ALTER AVAILABILITY GROUP [<availability_group_name>] 
    MODIFY REPLICA ON '<secondary_node>'   
    WITH (SEEDING_MODE = MANUAL)
GO

The preceding script cancels any replicas that are currently seeding, and prevents SQL Server from automatically initializing any replicas in this availability group. It does not stop synchronization for any replicas that are already initialized.

Monitor automatic seeding availability group

Use system dynamic management views to monitor seeding

The following system views show the status of SQL Server automatic seeding.

sys.dm_hadr_automatic_seeding

On the primary replica, query sys.dm_hadr_automatic_seeding to check the status of the automatic seeding process. The view returns one row for each seeding process. For example:

SELECT start_time, 
    completion_time
    is_source,
    current_state,
    failure_state,
    failure_state_desc
FROM sys.dm_hadr_automatic_seeding

sys.dm_hadr_physical_seeding_stats

On the primary replica, query sys.dm_hadr_physical_seeding_stats DMV to see the physical statistics for each seeding process that is currently running. The following query returns rows when seeding is running:

SELECT * FROM sys.dm_hadr_physical_seeding_stats;

The two columns total_disk_io_wait_time_ms and the total_network_wait_time_ms can be used to determine performance bottleneck in the Automatic seeding process. The two columns are also present in the hadr_physical_seeding_progress extended event.

  • total_disk_io_wait_time_ms represents the time spent by the backup/restore thread while waiting on the disk. This value is cumulative since the start of the seeding operation. If the disks are not ready for reading or writing the backup stream, the backup/restore thread transitions into a sleep state and wakes up every one second to check if the disk is ready.

  • total_network_wait_time_ms is interpreted differently for the Primary and the Secondary replica. At the primary replica this counter represents the network flow control time. On the secondary replica this represents the time the restore thread is waiting for a message to be available for writing to the disk.

Diagnose database initialization using automatic seeding in the error log

When you add a database to an availability group configured for automatic seeding, SQL Server performs a VDI backup over the availability group endpoint. Review the SQL Server error log for information on when the backup completed and the secondary was synchronized.

Diagnose database level health with extended events

Automatic seeding has new extended events for tracking state change, failures, and performance statistics during initialization.

For example, this script creates an extended events session that captures events related to automatic seeding:

CREATE EVENT SESSION [AlwaysOn_autoseed] ON SERVER 
    ADD EVENT sqlserver.hadr_automatic_seeding_state_transition,
    ADD EVENT sqlserver.hadr_automatic_seeding_timeout,
    ADD EVENT sqlserver.hadr_db_manager_seeding_request_msg,
    ADD EVENT sqlserver.hadr_physical_seeding_backup_state_change,
    ADD EVENT sqlserver.hadr_physical_seeding_failure,
    ADD EVENT sqlserver.hadr_physical_seeding_forwarder_state_change,
    ADD EVENT sqlserver.hadr_physical_seeding_forwarder_target_state_change,
    ADD EVENT sqlserver.hadr_physical_seeding_progress,
    ADD EVENT sqlserver.hadr_physical_seeding_restore_state_change,
    ADD EVENT sqlserver.hadr_physical_seeding_submit_callback
    ADD TARGET package0.event_file(
        SET filename=N'autoseed.xel',
            max_file_size=(5),
            max_rollover_files=(4)
        )
WITH (
    MAX_MEMORY=4096 KB,
    EVENT_RETENTION_MODE=ALLOW_SINGLE_EVENT_LOSS,
    MAX_DISPATCH_LATENCY=30 SECONDS,
    MAX_EVENT_SIZE=0 KB,
    MEMORY_PARTITION_MODE=NONE,
    TRACK_CAUSALITY=OFF,
    STARTUP_STATE=ON
    )
GO 

ALTER EVENT SESSION AlwaysOn_autoseed ON SERVER STATE=START
GO 

The following table lists extended events related to automatic seeding:

Name Description
hadr_db_manager_seeding_request_msg Seeding request message.
hadr_physical_seeding_backup_state_change Physical seeding backup side state change.
hadr_physical_seeding_restore_state_change Physical seeding restore side state change.
hadr_physical_seeding_forwarder_state_change Physical seeding forwarder side state change.
hadr_physical_seeding_forwarder_target_state_change Physical seeding forwarder target side state change.
hadr_physical_seeding_submit_callback Physical seeding submit callback event.
hadr_physical_seeding_failure Physical seeding failure event.
hadr_physical_seeding_progress Physical seeding progress event.
hadr_physical_seeding_schedule_long_task_failure physical seeding schedule long task failure event.
hadr_automatic_seeding_start Occurs when an automatic seeding operation is submitted.
hadr_automatic_seeding_state_transition Occurs when an automatic seeding operation changes state.
hadr_automatic_seeding_success Occurs when an automatic seeding operation succeeds.
hadr_automatic_seeding_failure Occurs when an automatic seeding operation fails.
hadr_automatic_seeding_timeout Occurs when an automatic seeding operation times out.

Other troubleshooting considerations

Monitor when automatic seeding

Query sys.dm_hadr_physical_seeding_stats for currently running automatic seeding processes. The view returns one row for each database. For example:

SELECT local_database_name, 
    role_desc, 
    internal_state_desc, 
    transfer_rate_bytes_per_second, 
    transferred_size_bytes, 
    database_size_bytes, 
    start_time_utc, 
    end_time_utc, estimate_time_complete_utc, 
    total_disk_io_wait_time_ms, 
    total_network_wait_time_ms, 
    is_compression_enabled 
FROM sys.dm_hadr_physical_seeding_stats

Troubleshoot why a database fails to appear in an availability group configured for automatic seeding

When a database fails to appear as part of an availability group with automatic seeding enabled, the automatic seeding likely failed. This prevents addition of the database to the availability group on either the primary and secondary replica. Query sys.dm_hadr_automatic_seeding on both the primary and secondary replicas. For example, run the following query to identify failure state of automatic seeding.

SELECT start_time, 
    completion_time, 
    is_source, 
    current_state, 
    failure_state, 
    failure_state_desc, 
    error_code 
FROM sys.dm_hadr_automatic_seeding

Automatic seeding and performance considerations

SQL Server uses a fixed number of threads for automatic seeding. On the primary instance, SQL Server uses one thread per LUN to read changes. On the secondary instance SQL Server uses one thread per LUN to initialize database.

Set trace flag 9567 on the primary replica to enable compression of the data stream during automatic seeding. This can significantly reduce the transfer time of automatic seeding, however it also increases the CPU usage. For more information, see Tune compression for availability group.

When not to use automatic seeding

In some scenarios automatic seeding may not be optimal for initializing a secondary replica. During automatic seeding, SQL Server performs a backup over the network for initialization. This process can be slow if databases are very large, or the secondary replica is remote. The transaction log for these databases cannot be truncated during the backup process, so a prolonged initialization process on a busy database can result in significant transaction log growth. Before adding a database to an availability group with automatic seeding, evaluate the database size, load, and site distance between replicas.

See also