Bug #42160 MEM should deal with counters stepping backwards

• Submitted: 16 Jan 2009 15:26
• Modified: 5 Feb 2014 16:22
• Reporter: Domas Mituzas
• Status: Closed
• Category: MySQL Enterprise Monitor: Server
• Severity: S2 (Serious)
• Version: 2.0
• OS: Any
• Assigned to: Assigned Account
• CPU Architecture: Any

[16 Jan 2009 15:26] Domas Mituzas

Description:
Due to nature how 'SHOW GLOBAL STATUS' gets collected (walking through active threads), some counters may actually step back (or whatever other reason is for non-sequential values). 

MEM treats such data as counter reset and uses full value, though as other values in equation do not reset, this leads to wrong evaluations and bad alerts.

How to repeat:
Example wrong data:

Created_tmp_disk_tables 23186   2008-12-19 06:35:01    
Created_tmp_disk_tables 23186   2008-12-19 06:40:01     
Created_tmp_disk_tables 23186   2008-12-19 06:45:01     
Created_tmp_disk_tables 23187   2008-12-19 06:50:01     
Created_tmp_disk_tables 23186   2008-12-19 06:55:01    
Created_tmp_disk_tables 23186   2008-12-19 07:00:01   

Suggested fix:
In case one value in equation has been reset, and others not, such event should be discarded, as due to bad math wrong advisor would be shown.

[1 Feb 2009 20:56] MySQL Verification Team

one reason for counters going backwards can be bug #31222

[5 Feb 2014 16:22] Mark Leith

This was resolved with 3.0.