-- Extended event session definition to track AlwaysOn latency
-- These events can be very noisy and it is not recommended to capture these events continuously on production instances
-- This event would need to be started on all the Availability Group replicas
IF EXISTS (select * from sys.server_event_sessions 
	where name = N'AlwaysOn_Data_Movement_Tracing')
		DROP EVENT SESSION [AlwaysOn_Data_Movement_Tracing] ON SERVER 
GO
CREATE EVENT SESSION [AlwaysOn_Data_Movement_Tracing] ON SERVER 
ADD EVENT sqlserver.hadr_apply_log_block,
ADD EVENT sqlserver.hadr_capture_log_block,
ADD EVENT sqlserver.hadr_database_flow_control_action,
ADD EVENT sqlserver.hadr_db_commit_mgr_harden,
ADD EVENT sqlserver.hadr_log_block_send_complete,
ADD EVENT sqlserver.hadr_send_harden_lsn_message,
ADD EVENT sqlserver.hadr_transport_flow_control_action,
ADD EVENT sqlserver.log_flush_complete,
ADD EVENT sqlserver.log_flush_start,
ADD EVENT sqlserver.recovery_unit_harden_log_timestamps,
ADD EVENT sqlserver.log_block_pushed_to_logpool,
ADD EVENT sqlserver.hadr_transport_receive_log_block_message,
ADD EVENT sqlserver.hadr_receive_harden_lsn_message,
ADD EVENT sqlserver.hadr_log_block_group_commit,
ADD EVENT sqlserver.hadr_log_block_compression,
ADD EVENT sqlserver.hadr_log_block_decompression,
ADD EVENT sqlserver.hadr_lsn_send_complete,
ADD EVENT sqlserver.hadr_capture_filestream_wait,
ADD EVENT sqlserver.hadr_capture_vlfheader,
ADD EVENT ucs.ucs_connection_send_msg(
    WHERE ([correlation_id]<>(0))) 
ADD TARGET package0.event_file(SET filename=N'AlwaysOn_Data_Movement_Tracing.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