Share via


sys.dm_hadr_database_replica_states (Transact-SQL)

Returns a row for each database that is participating in an AlwaysOn availability group for which the local instance of SQL Server is hosting an availability replica. This dynamic management view exposes state information on both the primary and secondary replicas. On a secondary replica, this view returns a row for every secondary database on the server instance. On the primary replica, this view returns a row for each primary database and an additional row for the corresponding secondary database.

Important

Depending on the action and higher-level states, database-state information may be unavailable or out of date. Furthermore, the values have only local relevance. For example, on the primary replica, the value of the last_hardened_lsn column reflects the information about a given secondary database that is currently available to the primary replica, not the actual hardened LSN value that the secondary replica might have currently.

Column name

Data type

Description (on primary replica)

database_id

int

Identifier of the database, unique within an instance of SQL Server. This is the same value as displayed in the sys.databases catalog view.

group_id

uniqueidentifier

Identifier of the availability group to which the database belongs.

replica_id

uniqueidentifier

Identifier of the availability replica within the availability group.

group_database_id

uniqueidentifier

Identifier of the database within the availability group. This identifier is identical on every replica to which this database is joined.

is_local

bit

Whether the availability database is local, one of:

0 = The database is not local to the SQL Server instance.

1 = The database is local to the server instance.

synchronization_state

tinyint

Data-movement state, one of:

Value

Description

0

Not synchronizing.

  • For a primary database, indicates that the database is not ready to synchronize its transaction log with the corresponding secondary databases.

  • For a secondary database, indicates that the database has not started log synchronization because of a connection issue, is being suspended, or is going through transition states during startup or a role switch.

1

Synchronizing.

  • For a primary database, indicates that the database is ready to accept a scan request from a secondary database.

  • For a secondary database, indicates that active data movement is occurring for the database.

2

Synchronized.

  • A primary database shows SYNCHRONIZED in place of SYNCHRONIZING.

  • A synchronous-commit secondary database shows synchronized when the local cache says the database is failover ready and is synchronizing.

3

Reverting. Indicates the phase in the undo process when a secondary database is actively getting pages from the primary database.

Warning

When a database on a secondary replica is in the REVERTING state, forcing failover to the secondary replica leaves the database in a state in which it cannot be started as a primary database. Either the database will need to reconnect as a secondary database, or you will need to apply new log records from a log backup.

4

Initializing. Indicates the phase of undo when the transaction log required for a secondary database to catch up to the undo LSN is being shipped and hardened on a secondary replica.

Warning

When a database on a secondary replica is in the INITIALIZING state, forcing failover to the secondary replica leaves the database in a state in which it be started as a primary database. Either the database will need to reconnect as a secondary database, or you will need to apply new log records from a log backup.

synchronization_state_desc

nvarchar(60)

Description of the data-movement state, one of:

NOT SYNCHRONIZING

SYNCHRONIZING

SYNCHRONIZED

REVERTING

INITIALIZING

is_commit_participant

bit

0 = Transaction commit is not synchronized with respect to this database.

1 = Transaction commit is synchronized with respect to this database.

For a database on an asynchronous-commit availability replica, this value is always 0.

For a database on a synchronous-commit availability replica, this value is accurate only on the primary database.

synchronization_health

tinyint

Reflects the intersection of the synchronization state of a database that is joined to the availability group on the availability replica and the availability mode of the availability replica (synchronous-commit or asynchronous-commit mode), one of:

Value

Description

0

Not healthy. The synchronization_state of the database is 0 (NOT SYNCHRONIZING).

1

Partially healthy. A database on a synchronous-commit availability replica is considered partially healthy if synchronization_state is 1 (SYNCHRONIZING).

2

Healthy. A database on an synchronous-commit availability replica is considered healthy if synchronization_state is 2 (SYNCHRONIZED), and a database on an asynchronous-commit availability replica is considered healthy if synchronization_state is 1 (SYNCHRONIZING).

synchronization_health_desc

nvarchar(60)

Description of the synchronization_health of the availability database.

NOT_HEALTHY

PARTIALLY_HEALTHY

HEALTHY

database_state

tinyint

0 = Online

1 = Restoring

2 = Recovering

3 = Recovery pending

4 = Suspect

5 = Emergency

6 = Offline

Note

Same as state column in sys.databases.

database_state_desc

nvarchar(60)

Description of the database_state of the availability replica.

ONLINE

RESTORING

RECOVERING

RECOVERY_PENDING

SUSPECT

EMERGENCY

OFFLINE

Note

Same as state column in sys.databases.

is_suspended

bit

Database state, one of:

0 = Resumed

1 = Suspended

suspend_reason

tinyint

If the database is suspended, the reason for the suspended state, one of:

0 = User action

1 = Suspend from partner

2 = Redo

3 = Capture

4 = Apply

5 = Restart

6 = Undo

7 = Revalidation

8 = Error in the calculation of the secondary-replica synchronization point

suspend_reason_desc

nvarchar(60)

Description of the database suspended state reason, one of:

SUSPEND_FROM_USER = A user manually suspended data movement

SUSPEND_FROM_PARTNER = The database replica is suspended after a forced failover

SUSPEND_FROM_REDO = An error occurred during the redo phase

SUSPEND_FROM_APPLY = An error occurred when writing the log to file (see error log)

SUSPEND_FROM_CAPTURE = An error occurred while capturing log on the primary replica

SUSPEND_FROM_RESTART = The database replica was suspended before the database was restarted (see error log)

SUSPEND_FROM_UNDO = An error occurred during the undo phase (see error log)

SUSPEND_FROM_REVALIDATION = Log change mismatch is detected on reconnection (see error log)

SUSPEND_FROM_XRF_UPDATE = Unable to find the common log point (see error log)

recovery_lsn

numeric(25,0)

On the primary replica, the end of the transaction log before the primary database writes any new log records after recovery or failover. For a given secondary database, if this value is less than the current hardened LSN (last_hardened_lsn), recovery_lsn is the value to which this secondary database would need to resynchronize (that is, to revert to and reinitialize to). If this value is greater than or equal to the current hardened LSN, resynchronization would be unnecessary and would not occur.

recovery_lsn reflects a log-block ID padded with zeroes. It is not an actual log sequence number (LSN). For information about how this value is derived, see Understanding the LSN Column Values, later in this topic.

truncation_lsn

numeric(25,0)

On the primary replica, for the primary database, reflects the minimum log truncation LSN across all the corresponding secondary databases. If local log truncation is blocked (such as by a backup operation), this LSN might be higher than the local truncation LSN.

For a given secondary database, reflects the truncation point of that database.

truncation_lsn reflects a log-block ID padded with zeroes. It is not an actual log sequence number.

last_sent_lsn

numeric(25,0)

The log block identifier that indicates the point up to which all log blocks have been sent by the primary. This is the ID of the next log block that will be sent, rather than the ID of the most recently sent log block.

last_sent_lsn reflects a log-block ID padded with zeroes, It is not an actual log sequence number.

last_sent_time

datetime

Time when the last log block was sent.

last_received_lsn

numeric(25,0)

Log block ID identifying the point up to which all log blocks have been received by the secondary replica that hosts this secondary database.

last_received_lsn reflects a log-block ID padded with zeroes. It is not an actual log sequence number.

last_received_time

datetime

Time when the log block ID in last message received was read on the secondary replica.

last_hardened_lsn

numeric(25,0)

Start of the Log Block containing the log records of last hardened LSN on a secondary database.

On an asynchronous-commit primary database or on a synchronous-commit database whose current policy is "delay", the value is NULL. For other synchronous-commit primary databases, last_hardened_lsn indicates the minimum of the hardened LSN across all the secondary databases.

Note

last_hardened_lsn reflects a log-block ID padded with zeroes. It is not an actual log sequence number. For more information, see Understanding the LSN Column Values, later in this topic.

last_hardened_time

datetime

On a secondary database, time of the log-block identifier for the last hardened LSN (last_hardened_lsn). On a primary database, reflects the time corresponding to minimum hardened LSN.

last_redone_lsn

numeric(25,0)

Actual log sequence number of the last log record that was redone on the secondary database. last_redone_lsn is always less than last_hardened_lsn.

last_redone_time

datetime

Time when the last log record was redone on the secondary database.

log_send_queue_size

bigint

Amount of log records of the primary database that has not been sent to the secondary databases, in kilobytes (KB).

log_send_rate

bigint

Rate at which log records are being sent to the secondary databases, in kilobytes (KB)/second.

redo_queue_size

bigint

Amount of log records in the log files of the secondary replica that has not yet been redone, in kilobytes (KB).

redo_rate

bigint

Rate at which the log records are being redone on a given secondary database, in kilobytes (KB)/second.

filestream_send_rate

bigint

The rate at which the FILESTREAM files are shipped to the secondary replica, in kilobytes (KB)/second.

end_of_log_lsn

numeric(25,0)

Local end of log LSN. Actual LSN corresponding to the last log record in the log cache on the primary and secondary databases. On the primary replica, the secondary rows reflect the end of log LSN from the latest progress messages that the secondary replicas have sent to the primary replica.

end_of_log_lsn reflects a log-block ID padded with zeroes. It is not an actual log sequence number. For more information, see Understanding the LSN Column Values, later in this topic.

last_commit_lsn

Numeric(25,0)

Actual log sequence number corresponding to the last commit record in the transaction log.

On the primary database, this corresponds to the last commit record processed. Rows for secondary databases show the log sequence number that the secondary replica has sent to the primary replica.

On the secondary replica, this is the last commit record that was redone.

last_commit_time

datetime

Time corresponding to the last commit record.

On the secondary database, this time is the same as on the primary database.

On the primary replica, each secondary database row displays the time that the secondary replica that hosts that secondary database has reported back to the primary replica. The difference in time between the primary-database row and a given secondary-database row represents approximately the recovery time objective (RPO), assuming that the redo process is caught up and that the progress has been reported back to the primary replica by the secondary replica.

low_water_mark_for_ghosts

bigint

A monotonically increasing number for the database indicating a low water mark used by ghost cleanup on the primary database. If this number is not increasing over time, it implies that ghost cleanup might not happen. To decide which ghost rows to clean up, the primary replica uses the minimum value of this column for this database across all availability replicas (including the primary replica).

Understanding the LSN Column Values

The values of the end_of_log_lsn, last_hardened_lsn, last_received_lsn, last_sent_lsn, recovery_lsn, and truncation_lsn columns are not actual log sequence numbers (LSNs). Rather each of these values reflects a log-block ID padded with zeroes.

end_of_log_lsn, last_hardened_lsn, and recovery_lsn are flush LSNs. For example, last_hardened_lsn indicates the start of the next block past the blocks that are already on disk. So any LSN < the value of last_hardened_lsn is on disk. LSN that are >= to this value are not flushed.

Of the LSN values returned by sys.dm_hadr_database_replica_states, only last_redone_lsn is a real LSN.

Security

Permissions

Requires VIEW SERVER STATE permission on the server.

See Also

Concepts

AlwaysOn Availability Groups (SQL Server)

Monitor Availability Groups (Transact-SQL)