Checking Latency on SQL Server 2000 - Page 2

Steps involved in checking latency in a replication environment

1. Create the heartbeat table and heartbeat stored procedure on the publishing database
2. Create the heartbeat job on the publishing server
3. Create the heartbeat table, heartbeat stored procedure, stored procedure to write file and latency check stored procedure in the subscribing database
4. Create the Latency check job on the subscribing server
5. Add the heartbeat stored procedure as an article to the publisher so that the execution of that procedure can be replicated to the subscriber.

Steps explained

Step 1

Create the heartbeat table and heartbeat stored procedure on the publishing database using the script below.

Download Heartbeat.sql.

Click for full code

Step 2

Create the heartbeat job on the publishing server using the script below.

Download HeartbeatJob.sql

Click for full code

This will create the job shown in the Fig 1.5 and 1.6

[Fig 1.5]

[Fig 1.6]

Step 3

Create the heartbeat table, heartbeat stored procedure, stored procedure to write file and latency check stored procedure in the subscribing database using the script below.

Download Latency.sql

Click for full code

Step 4

Create the Latency check job on the subscribing server.

Download LatencyCheckjob.sql

Click for full code

This will create a job shown in Fig 1.7 and 1.8.

[Fig 1.7]

[Fig 1.8]

Step 5

Add usp_heartbeat to the existing publication using the script below.

Download Addarticle.sql

use Employee			--change to your publishing databasename
go
sp_addarticle 
@publication ='Employee',		--change to your publication 
@pre_creation_cmd='none',
@article =  'usp_heartbeat' ,
@destination_table =  'usp_heartbeat',
@type ='proc exec',
@schema_option=0x01,
@destination_owner ='dbo',
@source_owner ='dbo',
@source_object='usp_heartbeat',
@force_invalidate_snapshot =0,
@status=0
go
exec sp_addsubscription 
@publication = N'Employee', 	--change to your publication
@article = N'usp_heartbeat', 
@subscriber = N'MAK', 		-- Change to your subscriber name
@destination_db = N'Employee',  -- Change to your subscription database name
@sync_type = N'none', 
@update_mode = N'read only' ,
@subscription_type = 'PULL' --change this to push or PULL depending on your environment

go

sp_refreshsubscriptions 'Employee'	--change to your publisher name
go

This will add the execution of usp_heartbeat to the publication and subscription databases. [Refer Fig 1.9, 2.0]

[Fig 1.9]

[Fig 2.0]

Soon after executing the above script, execution of the procedure usp_heartbeat will start replicating.

Now let's check the heartbeat table on the Publishing database

use employee
go
select * from repl_heartbeat
go

Result

Sourceserver name	SourceDatabase Name	TargetServer Name	TargetDatabase Name	SourceTimeStamp	TargetTimeStamp
EBONY	Employee	NULL	NULL	"2004-11-05 18:55:00.373"	NULL

Now let us check heartbeat table in the Subscribing database

use employee
go
select * from repl_heartbeat
go

Sourceserver name	SourceDatabase Name	TargetServer Name	TargetDatabase Name	SourceTimeStamp	TargetTimeStamp
EBONY	Employee	MAK	Employee	"2004-11-05 18:55:00.373"	"2004-11-05 18:55:26.570"

The job "Job_Checklatency" creates a log file "Latency.log" on the distributor as shown in Fig 2.1.

[Fig 2.1]

The contents of the log file, "Latency.log," are shown below:

Replication from EBONY..Employee to MAK..Employee with latency  
  26 secs and threshold 120secs :Date =Nov  5 2004  7:05:00:013P
Replication from EBONY..Employee to MAK..Employee with latency  
  22 secs and threshold 120secs :Date =Nov  5 2004  7:06:00:013P
Replication from EBONY..Employee to MAK..Employee with latency  
  10 secs and threshold 120secs :Date =Nov  5 2004  7:07:00:980P
Replication from EBONY..Employee to MAK..Employee with latency  
  26 secs and threshold 120secs :Date =Nov  5 2004  7:08:00:247P
Replication from EBONY..Employee to MAK..Employee with latency  
  26 secs and threshold 120secs :Date =Nov  5 2004  7:09:00:123P

When the latency exceeds the threshold of 120secs, it create another log file "Error.log" as shown in Fig 2.2.

Click for larger image

[Fig 2.2]

The content of the file, "Error.log," is shown below.

Latency Warning :Replication from EBONY..Employee to MAK..Employee with latency  229 secs 
  exeeded the threshold 120 secs  :Date =Nov  5 2004  7:10:18:907P

The stpred procedure, "usp_checklatency," can be tweaked a little to alert the Database Administrators using xp_sendmail when the latency exceeds the threshold.

if @latency>@latencythreshold 
begin
xp_sendmail ..............
end

Conclusion

As mentioned earlier, the main intent of this article is to find and monitor the latency in a transactional replication environment.

» See All Articles by Columnist MAK