I have "uninventoried" the postgres plugin for most of the servers. By the way there's only a single server which suffers from this issue, and this is the one what is used by RHQ itself! The others - which are only monitored by the agent, are quite fine. The only problem is with the 9.2 version that some system tables has been changed with 9.2, so there's not current_query anymore.

This is strange, that in postgres log there's still lines like this:
2013-09-23 13:43:04 CEST ERROR:  column "current_query" does not exist at character 70
2013-09-23 13:43:04 CEST STATEMENT:  SELECT (SELECT COUNT(*) FROM pg_stat_activity where usename = $1 AND current_query != '<IDLE>') AS active,
          (SELECT COUNT(*) FROM pg_stat_activity WHERE usename = $2) AS total
2013-09-23 13:43:44 CEST FATAL:  password authentication failed for user "postgres"
2013-09-23 13:43:46 CEST FATAL:  password authentication failed for user "postgres"

I guess uninventory is not the perfect way to disable the plugin...

max_connection is the default 100 for all the servers.

Whenever it happen again, I will collect more Pg statistics!

Regards,

Attila



2013/9/23 Attila Heidrich <attila.heidrich@gmail.com>
Postgres database stopped again... I guess the problem was the enormous number of open files...

We use 9.2, and we also use the postgres plugin - which still doesn't really support 9.2 as far as I know.

Altogether Postgres (only the one storing the RHQ database) can only run for a few days, than the number of open files and the open connections raise really high, and finally I should restart it.

Usually I restart RHQ as well, I generally have no time to play with "what to restart".

The log is practically endless, I can quote, but I think I should know something to look for in it.


Regards,

Attila