Редактиране

Споделяне чрез


Use Alerts for Replication Agent Events

Applies to: SQL Server

SQL Server Management Studio and Microsoft SQL Server Agent provide a way to monitor events, such as replication agent events, using alerts. SQL Server Agent monitors the Windows application log for events that are associated with alerts. If such an event occurs, SQL Server Agent responds automatically, by executing a task that you have defined and/or sending e-mail or a pager message to a specified operator. SQL Server includes a set of predefined alerts for replication agents that you can configure to execute a task and/or notify an operator. For more information about defining a task to execute, see Automating a Response to an Alert.

The following alerts are installed when a computer is configured as a Distributor:

Message ID Predefined alert Condition causing the alert to fire Enters additional information in msdb..sysreplicationalerts
14150 Replication: agent success Agent shuts down successfully. Yes
14151 Replication: agent failure Agent shuts down with an error. Yes
14152 Replication: agent retry Agent shuts down after unsuccessfully retrying an operation (agent encounters an error such as server not available, deadlock, connection failure, or time-out failure). Yes
14157 Replication: expired subscription dropped Expired subscription was dropped. No
20572 Replication: Subscription reinitialized after validation failure Response job 'Reinitialize subscriptions on data validation failure' reinitializes a subscription successfully. No
20574 Replication: Subscriber has failed data validation Distribution or Merge Agent fails data validation. Yes
20575 Replication: Subscriber has passed data validation Distribution or Merge Agent passes data validation. Yes
20578 Replication: agent custom shutdown When data validation is invoked through sp_publication_validation and @shutdown_agent is set to 1, the distribution agent is shut down after validation completes. Yes
22815 Peer-to-peer conflict detection alert Distribution Agent detected a conflict when it tries to apply a change at a peer-to-peer node. Yes

In addition to these alerts, Replication Monitor provides a set of warnings and alerts related to status and performance. For more information, see Set Thresholds and Warnings in Replication Monitor. You can also define alerts for other replication events using the SQL Server alerts infrastructure. For more information, see Create a User-Defined Event.

To configure predefined replication alerts

View the Application Log directly

To view the Windows application log, use the Microsoft Windows Event Viewer. The application log contains SQL Server error messages as well as messages for many other activities on the computer. Unlike the SQL Server error log, a new application log is not created each time you start SQL Server (each SQL Server session writes new events to an existing application log); however, you can specify how long logged events will be retained. When you view the Windows application log, you can filter the log for specific events. For more information, see the Windows documentation.

Automate a response to an alert

Replication provides a response job for subscriptions that fail data validation, and also provides a framework for creating additional automated responses to alerts. The response job is titled Reinitialize subscriptions on data validation failure and is stored in the SQL Server Agent Jobs folder in SQL Server Management Studio. For information about enabling this response job, see Configure Predefined Replication Alerts (SQL Server Management Studio). If articles in a transactional publication fail validation, the response job reinitializes only those articles that failed. If articles in a merge publication fail validation, the response job reinitializes all articles in the publication.

Framework for automating responses

Usually, when an alert occurs, the only information you have to help you understand what caused the alert and the appropriate action to take is contained in the alert message itself. Parsing this information can be error-prone and time-consuming. Replication makes automating responses easier by providing additional information about the alert in the sysreplicationalerts system table; the information provided is already parsed in a form easily used by customized programs.

For example, if the data in the Sales.SalesOrderHeader table at Subscriber A fails validation, SQL Server can trigger message 20574, notifying you of that failure. The message you receive will be: "Subscriber 'A', subscription to article 'SalesOrderHeader' in publication 'MyPublication' failed data validation."

If you create a response based on the message, you must manually parse the Subscriber name, article name, publication name, and error from the message. However, because the Distribution Agent and Merge Agent write that same information to sysreplicationalerts (along with details such as the type of agent, time of the alert, publication database, Subscriber database, and type of publication) the response job can directly query the relevant information from the table. Although the exact row cannot be associated with a specific instance of the alert, the table has a status column, which can be used to keep track of serviced entries. The entries in this table are maintained for the history retention period.

For example, if you were to create a response job in Transact-SQL that services alert message 20574, you might use the following logic:

declare @publisher sysname, @publisher_db sysname, @publication sysname, @publication_type int, @article sysname, @subscriber sysname, @subscriber_db sysname, @alert_id int  
declare hc cursor local for select publisher, publisher_db, publication, publication_type, article, subscriber,   
  subscriber_db, alert_id from   
  msdb..sysreplicationalerts where  
  alert_error_code = 20574 and status = 0  
  for read only  
open hc  
fetch hc into  @publisher, @publisher_db, @publication, @publication_type, @article, @subscriber, @subscriber_db, @alert_id  
while (@@fetch_status <> -1)  
begin  
/* Do custom work  */  
/* Update status to 1, which means the alert has been serviced. This prevents subsequent runs of this job from doing this again */  
update msdb..sysreplicationalerts set status = 1 where alert_id = @alert_id  
 fetch hc into  @publisher, @publisher_db, @publication, @publication_type, @article, @subscriber, @subscriber_db, @alert_id  
end  
close hc  
deallocate hc