PacketFence - BTS - PacketFence
View Issue Details
0001096PacketFenceerror-handlingpublic2010-10-21 11:262015-02-13 15:42
obilodeau 
 
normalminorhave not tried
closedopen 
 
investigate 
0001096: a lot of failed database query in large environments
A LOT of entries like the following can be seen in some of our largest environments:

WARN: database query failed with: MySQL server has gone away. (errno: 2006), will try again (pf::db::db_query_execute)

Up to several thousands of messages like this per week.

I thought they were related with running out of MySQL handles but we checked today and it seems they are not.

It will need further investigation.
No tags attached.
Issue History
2010-10-21 11:26obilodeauNew Issue
2010-10-29 14:44obilodeauNote Added: 0001739
2011-01-18 10:03obilodeauTarget Version => 2.0.1
2011-01-26 15:30obilodeauTarget Version2.0.1 => 2.0.2
2011-03-03 15:19obilodeauTarget Version2.0.2 => +1
2012-04-05 10:23haroldNote Added: 0002636
2012-10-19 12:11fgaudreaultTarget Versiongeneral => investigate
2015-02-13 15:42lmunroNote Added: 0003794
2015-02-13 15:42lmunroStatusnew => closed

Notes
(0001739)
obilodeau   
2010-10-29 14:44   
It might be related to the per-switch connections that are made by the switch module when starting to handle every trap.

If the issue is reproducible with 1.9.x and not with trunk then we might have the culprit.
(0002636)
harold   
2012-04-05 10:23   
Please provide more information:

- logs switched to debug or trace level on pf::db
- logs containing timestamps and mysql connections stats (like dstat -t --mysql5-conn --output mysql_stats.csv)
- mysql server variables values (especially interactive_timeout, wait_timeout)
- mysql query logs (if you log queries and related thread/connection ids)
(0003794)
lmunro   
2015-02-13 15:42   
These bugs have been sitting untouched since 2012.
Closing them and possibly reopening in github tracker where relevant.