The mailing list for listmasters using Sympa

["Steve Shipway" (via sympa-users Mailing List) <address@concealed>]

> I've had a couple of issues lately of message distribution failing with the 
> error
> "MySQL server has gone away". The first was with a daily digest, and the
> most recent this afternoon with an ordinary message. It also looks as if the
> attempt to tell the original sender of the failure may itself have failed.

This ( 'mysql has gone away' ) happens when the connection is prematurely 
disconnected.  There are a number of configuration and timeouts that might 
cause this.

First, check your max_allowed_packet size in the my.cnf.  Set it nice and big 
, maybe max_allowed_packet = 50000000.  Large messages can cause problems 
like this.

Now check for intervening firewalls with low session timeouts.  If there are 
any firewalls (including host-based) between sympa and mysql, make sure they 
don’t have small session timeouts causing inactive sessions to be prematurely 
disconnected.

Set wait_timeout in my.cnf to something suitably large, such as 180 or even 
more.  Also, connect_timeout can be set to 10 or so.

Set log_slow_queries and check that you aren’t getting queries with an 
unreasonable executation time -- this would indicate problems with missing 
indices, or insufficient memory given to the mysql buffer sizes.  If this is 
the case, increase {read|join|sort}_buffer_size , and make sure you have 
query_cache_size set high enough.

This should help out the issue.

Steve

Steve Shipway
address@concealed

Attachment: smime.p7s
Description: S/MIME cryptographic signature