While investigating a bit the source of the problems of the AMs that were disconnected after some time, we obtained the following from an OF AM:
Aggregate Manager *** was unreachable at [[ 08:47:03 PST ]] the cause
was: <ProtocolError for xxx:yyy@zzz:www: 500 Internal Server Error>
After checking the opt-in manager log:
[Mon Nov 04 09:49:56 2013] [error] [client zzz] OperationalError: (1040,
'Too many connections')
which is related to [1]
The maximum number of MySQL connections is 151 at a time. For some reason, Django ORM does not properly close these on opt-in manager.
To check how many connections are open in MySQL:
mysql> show processlist;
At this point there are at least two quick fixes:
(a) restart the mysql server daemon so as to flush the zombie connections
(b) increase "max_connections" variable to some higher value, for example 200 (see [1])
While investigating a bit the source of the problems of the AMs that were disconnected after some time, we obtained the following from an OF AM:
After checking the opt-in manager log:
which is related to [1]
The maximum number of MySQL connections is
151at a time. For some reason, Django ORM does not properly close these on opt-in manager.To check how many connections are open in MySQL:
At this point there are at least two quick fixes:
(a) restart the mysql server daemon so as to flush the zombie connections (b) increase "max_connections" variable to some higher value, for example 200 (see [1])
[1] https://dev.mysql.com/doc/refman/5.5/en/too-many-connections.html
but we still need...
To search for a solution