Bill Moran Network Engineer 5849 Forbes Avenue Pittsburgh, PA 15217

Using John Sidney-Woollett's nagios check as a starting point, i've developed something that I feel is better. The changes in this improved script:

• Includes the ability to return WARNING as well as ERROR.
• The script determines the status based on how many events each node is behind instead of lag time.

I believe these changes make the check more flexible while also catching a larger number of failure conditions, but this may not be what everyone is looking to accomplish. Thus I have renamed the check check_slony_lag to differentiate it from John's work.

We generally use 50 events as the ERROR level and 20 as the WARNING level (although some of our clusters use different values). Obviously, practical values will vary depending on a large number of factors, but these should provide a starting point. My opinion would be to start out with a very low WARNING and adjust up as experiece demonstrates what normal event lag is for your system, then double that value for the ERROR level.

The script can be downloaded here.

All content copyright Collaborative Fusion, Inc. and Bill Moran. All rights reserved.