Configure Extended Events for availability groups

Статия
03/03/2023

Applies to: SQL Server

SQL Server defines Extended Events that are specific to availability groups. You can monitor these Extended Events in a session to help with root-cause diagnosis when you troubleshoot an availability group. You can view the availability group Extended Events using the following query:

SELECT * FROM sys.dm_xe_objects WHERE name LIKE '%hadr%';

The alwayson_health session

The alwayson_health Extended Events session is created automatically when you create the availability group, and captures a subset of the availability group related events. This session is preconfigured as a useful and convenient tool to help you get started quickly while troubleshooting an availability group. The Create Availability Group Wizard automatically starts the session on every participating availability replica configured in the wizard.

Important

If you did not create the availability group using the New Availability Group Wizard, the alwayson_health session may not be automatically started. If the session is not started, it cannot capture event data when an unexpected issue occurs. You should manually start the session and configure the session to start automatically by configuring the session properties.

To view the definition of the alwayson_health session:

In the Object Explorer, expand Management, Extended Events, and then Sessions.
Right-click Alwayson_health, then point to Script Session as, then point to CREATE To, and then select New Query Editor Window.

Extended Events for debugging

In addition to the Extended Events covered by the Alwayson_health session, SQL Server defines an extensive set of debug events for availability groups. To harness these additional Extended Events in a session, follow the procedures below:

In the Object Explorer, expand Management, Extended Events, and then Sessions.
Right-click Sessions and select New Session. Or, right-click Alwayson_health and select Properties.
In the Select a page pane, select Events.
In the event library, in the Category column, select alwayson and clear all other categories.
In the Channel column, select Debug. All the availability group related events that aren't already selected are now shown in the event library.
Highlight an event in the event library, then select the > button to select it for the session.
When finished with the session, select OK to close it. Make sure that the session is started so that it captures the events that you selected.

availability_replica_state_change

Occurs when the state of an availability replica has changed. The creation of an availability group or joining an availability replica can trigger this event. It is useful for the diagnostics of failed automatic failover. It can also be used to trace the failover steps.

Event information

Column	Description
Name	availability_replica_state_change
Category	always on
Channel	Operational

Event fields

Name	Type_name	Description
availability_group_id	guid	The ID of the Availability Group.
availability_group_name	unicode_string	The name of the Availability Group.
availability_replica_id	guid	The ID of the Availability Replica.
previous_state	availability_replica_state	The role of the replica before the change. Possible values are: Primary_Normal Secondary_Normal Resolving_Pending_Failover Resolving_Normal Primary_Pending Not_Available
current_state	availability_replica_state	The role of the replica after the change. Possible values are: Primary_Normal Secondary_Normal Resolving_Pending_Failover Resolving_Normal Primary_Pending Not_Available

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT availability_replica_state_change
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

availability_group_lease_expired

Occurs when the cluster and availability group has a connectivity issue and the lease is expired. This event indicates that connectivity between the availability group and the underlying WSFC cluster is broken. If the connectivity issue occurs on the primary replica, the event may cause an automatic failover or cause the availability group to be offline.

Event information

Column	Description
Name	availability_group_lease_expired
Category	always on
Channel	Operational

Event fields

Name	Type_name	Description
availability_group_id	guid	The ID of availability group.
availability_group_name	unicode_string	The name of availability group.

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT availability_group_lease_expired
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

availability_replica_automatic_failover_validation

Occurs when the automatic failover validates the readiness of an availability replica as a primary replica, and it shows whether the target availability replica is ready to be the new primary replica. For example, the failover validation returns false when not all databases are synchronized or not joined. This event is design to provide a failure point during failovers. This information is of interest to the database administrator especially for automatic failovers because an automatic failover is an unattended operation. The database administrator can review the event to see why an automatic failover failed.

Event information

Name	Description
availability_replica_automatic _failover_validation
Category	always on
Channel	Analytic

Event fields

Name	Type_name	Description
availability_group_id	guid	The ID of the availability group.
availability_group_name	unicode_string	The name of the availability group.
availability_replica_id	guid	The ID of the availability replica.
forced_quorum	validation_result_type	If the value is TRUE, then the automatic failover is invalidated on this availability replica. TRUE FALSE
joined_and_synchronized	validation_result_type	If the value is FALSE, then the automatic failover is invalidated on this availability replica. TRUE FALSE
previous_primary_or_automatic_failover_target	validation_result_type	If the value is FALSE, then the automatic failover is invalidated on this availability replica. TRUE FALSE

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT availability_replica_automatic_failover_validation (
    WHERE (
        [forced_quorum] = (TRUE)
        OR [joined_and_synchronized] = (FALSE)
        OR [previous_primary_or_automatic_failover_target] = (TRUE)
        )
    ) ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

error_reported (multiple error numbers): for transport or connection issues

Each filtered event indicates that a connectivity issue occurred in the transport or database mirroring endpoint that availability group depends on.

Column	Description
Name	error_reported numbers to filter: 35201, 35202, 35206, 35204, 35207, 35217, 9642, 9666, 9691, 9692, 9693, 28034, 28036, 28080, 28091, 33309
Category	errors
Channel	Admin

Error numbers to filter

Error Number	Description
35201	A connection timeout has occurred while attempting to establish a connection to availability replica '%ls'.
35202	A connection for availability group '%ls' from availability replica '%ls' with ID [%ls] to '%ls' with ID [%ls] has been successfully established. This is an informational message only. No user action is required.
35206	A connection timeout has occurred on a previously established connection to availability replica '%ls'.
35204	Connection between instance '%ls' and '%ls' has been disabled due to endpoint shutdown.
Timeout + connected
35207	Connection attempt on availability group ID '%ls' from replica ID '%ls' to replica ID '%ls' failed because of error %d severity %d state %d. severity %d state %d. (this may not have a good DBA use. Check and remove later in that case)
35217	(Starting with SQL Server 2019 CU15 (15.0.4198.2)) The thread pool for Always On Availability Groups was unable to start a new worker thread because there are not enough available worker threads. This may degrade Always On Availability Groups performance. Use the "max worker threads" configuration option to increase number of allowable threads.
9642	An error occurred in a Service Broker/Database Mirroring transport connection endpoint Error: %i State: %i. (Near endpoint role: %S_MSG far endpoint address: '%.hs') Error: %i State: %i. (Near endpoint role: %S_MSG far endpoint address: '%.hs')
9666	The %S_MSG endpoint is in disabled or stopped state.
9691	The %S_MSG endpoint has stopped listening for connections.
9692	The %S_MSG endpoint cannot listen on port %d because it is in use by another process.
9693	The %S_MSG endpoint cannot listen for connections due to the following error: '%.*ls'.
28034	Connection handshake failed. The login '%.*ls' does not have CONNECT permission on the endpoint. State %d.
28036	Connection handshake failed. The certificate used by this endpoint was not found: %S_MSG. Use DBCC CHECKDB in `master` database to verify the metadata integrity of the endpoints. State %d.
28080	Connection handshake failed. The %S_MSG endpoint is not configured. State %d.
28091	Starting endpoint for %S_MSG with no authentication is not supported.
33309	Cannot start cluster endpoint because the default %S_MSG endpoint configuration has not been loaded yet.

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT sqlserver.error_reported (
    WHERE (
        --Connectivity Error Messages
        [error_number] = (35201)
        OR [error_number] = (35202)
        OR [error_number] = (35204)
        OR [error_number] = (35206)
        OR [error_number] = (35207)
        OR [error_number] = (35217)
        OR [error_number] = (9642)
        --OR [error_number]=(9666)
        OR [error_number] = (9691)
        OR [error_number] = (9692)
        OR [error_number] = (9693)
        OR [error_number] = (28034)
        OR [error_number] = (28036)
        OR [error_number] = (28080)
        OR [error_number] = (28091)
        OR [error_number] = (33309)
        )
    ) ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

data_movement_suspend_resume

Occurs when the database movement of a database replica is suspended or resumed.

Event information

Column	Description
Name	data_movement_suspend_resume
Category	Always on
Channel	Operational

Event fields

Name	Type_name	Description
availability_group_id	guid	The ID of the availability group.
availability_group_name	unicode_string	The name of the availability group, if available.
availability_replica_id	guid	The ID of the availability replica.
database_replica_id	guid	The ID of the availability database.
database_replica_name	unicode_string	The name of the availability database.
database_id	uint32	The ID of the availability database.
suspend_status	suspend_status_type	The suspend status values. SUSPEND_NULL RESUMED SUSPENDED SUSPENDED_INVALID
suspend_source	suspend_source_type	The source of the suspend or resume action.
suspend_reason	unicode_string	The suspend reason captured in database replica manager.

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT data_movement_suspend_resume (
    WHERE (
        [suspend_status] = (SUSPENDED)
        OR [suspend_status] = (SUSPENDED_INVALID)
        OR [suspend_status] = (SUSPEND_NULL)
        )
    ) ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

alwayson_ddl_executed

Occurs when an availability group data definition language (DDL) statement is being executed, including CREATE, ALTER, or DROP. The main purpose of the event is to indicate an issue with a user action on an availability replica, or to indicate the starting point of an operational action, which is followed by a runtime issue such as a manual failover, a forced failover, suspended data movement, or resumed data movement.

Event information

Column	Description
Name	alwayson_ddl_execution
Category	always on
Channel	Analytic

Event fields

Name	Type_name	Description
availability_group_id	Guid	The ID of the availability group.
availability_group_name	unicode_string	The name of the availability group.
ddl_action	alwayson_ddl_action	Indicates the type of DDL action: CREATE, ALTER, or DROP.
ddl_phase	ddl_opcode	Indicates the phase of the DDL operation: BEGIN, COMMIT, or ROLLBACK.
Statement	unicode_string	The text of the statement that was executed.

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT alwayson_ddl_executed
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

availability_replica_manager_state

Occurs when the state of the availability replica manager is changed. This event indicates the heartbeat of availability replica manager. When the availability replica manager is not in healthy state, all availability replicas in the SQL Server instance will be down.

Event information

Column	Description
Name	availability_replica_manager_state_change
Category	always on
Channel	Operational

Event fields

Name	Type_name	Description
current_state	manager_state	The current state of the availability replica manager. Online Offline WaitingForClusterCommunication

Alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT availability_replica_manager_state (WHERE ([current_state] = (OFFLINE)))
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

error_reported (1480): Database Replica Role Change

This filtered error_reported event occurs asynchronously after an availability replica role change. It indicates which availability database fails to change its expected role during the failover process.

Event information

Column	Description
Name	error_reported Error Number 1480: The REPLICATION_TYPE_MSG database "DATABASE_NAME" is changing roles from "OLD_ROLE" to "NEW_ROLE" due to REASON_MSG
Category	errors
Channel	Admin

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT sqlserver.error_reported (
    WHERE (
        --database replica role change message
        OR [error_number] = (1480)
        --database replica runtime error messages
        OR [error_number] = (823)
        OR [error_number] = (824)
        OR [error_number] = (829)
        )
    ) ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

sqlserver.sp_server_diagnostics_component_result

Captures diagnostic data and health information about SQL Server to detect potential failures. The procedure runs in repeat mode and sends results periodically. This extended event session is available starting with SQL Server 2019 CU15 (15.0.4198.2).

Event information

Name	Description
Name	sp_server_diagnostics_component_result
Category	Server
Channel	Debug

Event fields

Name	Type_name	Description
component	UInt8	component name.
state	UInt8	Indicates the health status of the component.
data	XML	XML field containing extra information regarding the component.

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT sqlserver.sp_server_diagnostics_component_result (SET collect_data = (1) WHERE ([state] = (3)))
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

ucs.ucs_connection_setup

Dump connectivity or network related logs between primary and secondary replicas. This extended event session is available starting with SQL Server 2019 CU15 (15.0.4198.2).

Event information

Name	Description
Name	ucs_connection_setup
Category	Transport
Channel	Debug

Event fields

Name	Type_name	Description
setup_event	Int32	Connection setup event
obj_address	Pointer	Connection end point address
endpoint_type	Int32	Endpoint type
stream_status	Int32	Connection stream status
error_number	UInt32	Connection error code
connection_id	GUID	Connection id
error_message	UnicodeString	Connection error message
address	UnicodeString	Connection target address
circuit_id	UnicodeString	Connection circuit id

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT ucs.ucs_connection_setup
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

sqlserver.hadr_trace_message

Redirects the output of some DBCC commands and HADR log information to the extended event session (similar to trace flag 3605). This extended event session is available starting with SQL Server 2019 CU15 (15.0.4198.2).

Event information

Name	Description
Name	hadr_trace_message
Category	Always on
Channel	Debug

Event fields

Name	Type_name	Description
hadr_message	unicode_string	Redirects the output of some DBCC commands and HADR log information to the extended event session (similar to trace flag 3605).

alwayson_health session definition

CREATE EVENT SESSION [alwayson_health] ON SERVER
ADD EVENT sqlserver.hadr_trace_message
ADD TARGET package0.event_file (
    SET filename = N'alwayson_health.xel',
    max_file_size = (5),
    max_rollover_files = (4),
    metadatafile = N'alwayson_health.xem'
    )
    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

Next steps

View event session data

Споделяне чрез

Configure Extended Events for availability groups

The alwayson_health session

Extended Events for debugging

availability_replica_state_change

Event information

Event fields

alwayson_health session definition

availability_group_lease_expired

Event information

Event fields

alwayson_health session definition

availability_replica_automatic_failover_validation

Event information

Event fields

alwayson_health session definition

error_reported (multiple error numbers): for transport or connection issues

Error numbers to filter

alwayson_health session definition

data_movement_suspend_resume

Event information

Event fields

alwayson_health session definition

alwayson_ddl_executed

Event information

Event fields

alwayson_health session definition

availability_replica_manager_state

Event information

Event fields

Alwayson_health session definition

error_reported (1480): Database Replica Role Change

Event information

alwayson_health session definition

sqlserver.sp_server_diagnostics_component_result

Event information

Event fields

alwayson_health session definition

ucs.ucs_connection_setup

Event information

Event fields

alwayson_health session definition

sqlserver.hadr_trace_message

Event information

Event fields

alwayson_health session definition

Next steps

Обратна връзка

Допълнителни ресурси