Excessive queries to fetch new messages when no messages are available to be processed

[DavidBoike]

Symptoms

An excessive amount of queries can occur when an endpoint is scaled out.

Who's affected

You are affected if you have SQL Transport endpoints scaled out to multiple instances. The problem is exacerbated when the database has snapshot isolation enabled, which is the default in SQL Azure instances.

Root cause

The SQL Transport uses a peek query to get an idea for how many messages are available to be processed, which is more complex when multiple endpoint instances are competing to process the same message. The previous peek query would read uncommitted data and continuously fire queries to fetch new messages but not receive any new data because the data was not yet committed.

Fixes

• READPAST,READCOMMITTEDLOCK instead of NOLOCK to not read uncommitted data #1277 for branch master
• Excessive queries to fetch new messages when no messages are available to be processed #1278 for release branch release-7.0 to be released in version 7.0.5
• Excessive queries to fetch new messages when no messages are available to be processed #1292 for branch release-6.3 to be released in version 6.3.7

See also

• Related to Competing-consumer deployments might result in wasted Select queries #863

[DavidBoike]

Released in 7.0.5 and 6.3.7