Transactional replication is a critical component of many SQL Server environments, providing high availability, load balancing, and other essential benefits. However, managing replication latency, the delay between an action occurring on the publisher and it being reflected on the subscriber, is vital for ensuring system performance and data integrity. In this blog post, we’ll explore a proactive approach to monitor and alert on replication latency, helping database administrators (DBAs) maintain optimal system health.

The Issue:

Replication latency can sometimes go unnoticed until it impacts the system performance or data accuracy, leading to potential data loss or business disruptions. Traditional monitoring techniques may not provide real-time alerts or may require significant manual intervention, making them less effective for immediate latency identification and resolution.

The Script:

To address this challenge, we introduce a SQL script designed by Vivek Janakiraman from JBSWiki, specifically crafted to monitor transactional replication latency in SQL Server environments. This script efficiently posts tracer tokens to specified publications and measures the time taken for these tokens to move through the replication components, providing a clear picture of any latency present in the system.

/*
Author: Vivek Janakiraman
Company: JBSWiki
Description: This script is used to alert in case there is Transactional replication Log reader or distribution agent latency.
             It posts tracer tokens to specified publications and measures the latency to the distributor and subscriber.
*/

-- Switch to the publisher database to insert tracer tokens.
USE [Publisher_Database_Here]   -- Use your publisher database name here.
-- Insert tracer tokens into the specified publications.
EXEC sys.sp_posttracertoken @publication = 'Publication_Name' -- Change appropriate Publication that should be monitored.
EXEC sys.sp_posttracertoken @publication = 'Publication_Name1' -- Change appropriate Publication that should be monitored.
-- Wait for 5 minutes to allow the tokens to propagate.
WAITFOR DELAY '00:05:00'

-- Switch to the distribution database to query latency information.
USE distribution
;WITH LatestEntries AS (
       -- Select the latest entries for each publication and agent.
       SELECT publication_id, agent_id, MAX(publisher_commit) AS MaxDate       
       FROM MStracer_tokens t
       JOIN MStracer_history h ON t.tracer_id = h.parent_tracer_id
       GROUP BY publication_id, agent_id
)
-- Select latency information for the latest tokens.
SELECT c.name, t.publication_id, h.agent_id, t.publisher_commit,
       ISNULL(DATEDIFF(s,t.publisher_commit,t.distributor_commit), 299) as 'Time To Dist (sec)',
       ISNULL(DATEDIFF(s,t.distributor_commit,h.subscriber_commit), 299) as 'Time To Sub (sec)'
INTO #REPL_LATENCY
FROM MStracer_tokens t 
JOIN MStracer_history h ON t.tracer_id = h.parent_tracer_id
JOIN distribution.dbo.MSdistribution_agents c ON h.agent_id = c.id
JOIN LatestEntries le ON t.publication_id = le.publication_id AND h.agent_id = le.agent_id AND t.publisher_commit = le.MaxDate
ORDER BY t.publisher_commit DESC

-- Check if there is any latency beyond acceptable limits and select those records.
IF EXISTS (SELECT 1 FROM #REPL_LATENCY WHERE ([Time To Dist (sec)] > 30 OR [Time To Sub (sec)] > 30))
BEGIN
    SELECT name, publication_id, agent_id, publisher_commit, [Time To Dist (sec)], [Time To Sub (sec)]
    INTO #REPL_LATENCY_Email 
    FROM #REPL_LATENCY 
    WHERE ([Time To Dist (sec)] > 30 OR [Time To Sub (sec)] > 30)
END

-- Prepare the HTML body content for the email alert.
DECLARE @body_content NVARCHAR(MAX);
SET @body_content = N'
<style>
table.GeneratedTable {
  width: 100%;
  background-color: #D3D3D3;
  border-collapse: collapse;
  border-width: 2px;
  border-color: #A9A9A9;
  border-style: solid;
  color: #000000;
}
table.GeneratedTable td, table.GeneratedTable th {
  border-width: 2px;
  border-color: #A9A9A9;
  border-style: solid;
  padding: 3px;
}
table.GeneratedTable thead {
  background-color: #A9A9A9;
}
</style>
<table class="GeneratedTable">
  <thead>
    <tr>
         <th>name</th>
         <th>publication_id</th>
         <th>agent_id</th>
         <th>publisher_commit</th>
         <th>[Time To Dist (sec)]</th>
         <th>[Time To Sub (sec)]</th>
    </tr>
  </thead>
  <tbody>' +
CAST(
          (SELECT td = name, '',
                               td = publication_id, '',
                               td = agent_id, '',
                               td = publisher_commit, '',
                               td = [Time To Dist (sec)], '',
                               td = [Time To Sub (sec)], ''
        FROM [dbo].#REPL_LATENCY_Email
        FOR XML PATH('tr'), TYPE   
        ) AS NVARCHAR(MAX)
    ) +
  N'</tbody>
</table>';

-- Send an email alert if there is any latency issue found.
IF EXISTS (SELECT TOP 1 * FROM [dbo].#REPL_LATENCY_Email) 
BEGIN
    EXEC msdb.dbo.sp_send_dbmail @profile_name = 'JBSWIKI',
                                 @body = @body_content,
                                 @body_format = 'HTML',
                                 @recipients = 'jvivek2k1@yahoo.com',
                                 @subject = 'ALERT: Transactional Replication Latency Alert'; 
END

-- Cleanup temporary tables.
DROP TABLE #REPL_LATENCY
DROP TABLE #REPL_LATENCY_Email

The Solution:

The script works by first posting tracer tokens to the specified publications within the publisher database. It then waits for a predetermined amount of time (defaulted to 5 minutes in the script) to allow the tokens to propagate through the system. Following this, the script measures the latency to the distributor and subscriber, providing a detailed report of the time taken in each stage of the replication process.

This information is then used to generate an HTML-formatted email alert if the latency exceeds predefined thresholds (30 seconds in the provided script), allowing for immediate action to be taken. The use of HTML formatting in the email ensures that the information is presented in an easily digestible format, facilitating quick understanding and response by the DBA.

Conclusion:

Proactive monitoring and management of transactional replication latency are paramount for maintaining the health and performance of SQL Server environments. The script provided offers a straightforward and effective solution for DBAs to stay ahead of potential replication issues. By automating the process of latency detection and alerting, this approach not only saves valuable time but also helps in preventing the negative impact of replication latency on business operations.

Remember, while this script serves as a valuable tool in your monitoring arsenal, it’s also important to tailor the solution to your specific environment and requirements. Regularly reviewing and adjusting the latency thresholds and monitoring frequency will ensure you continue to get the most out of your replication setup.

Thank You,
Vivek Janakiraman

Disclaimer:
The views expressed on this blog are mine alone and do not reflect the views of my company or anyone else. All postings on this blog are provided “AS IS” with no warranties, and confers no rights.