Re: Monitoring error messages

On Thu, 2006-08-17 at 08:33, robin(dot)c(dot)smith(at)bt(dot)com wrote:
> I am testing the monitoring of the postgresql 8.1.4 - I have found
> that I would like to suppress some things from the log
>
> For example if I type some random rubbish into the database instead of
> select…. I get an error message logged into the server logfile like:-
>
> 2006-08-17 14:19:00.965 BST 10109 # ERROR: syntax error at or near
> "kshdfkjh" at character 1
> 2006-08-17 14:19:00.965 BST 10109 # STATEMENT: kshdfkjh;
>
> More importantly I have deleted a file from the database tables and
> when I try to query data in the file I get the error:-
>
> 2006-08-17 14:14:30.922 BST 10085 # ERROR: could not open relation
> 1663/16384/16385: No such file or directory
> 2006-08-17 14:14:30.922 BST 10085 # STATEMENT: select * from
> rs_vacuum;
>
> These are both logged as ERROR: which is easy to check for with
> automatic monitoring. However I wouldn't want to be called out to fix
> a non-existent error where someone has typed in some nonsense. Is it
> easy to suppress the syntax errors so that the real database errors
> are obvious?

I've had the same issue in the past. A simple fix is to pipe the output
of your grep into a series of "grep -v" statements.

For instance, this line is how we used to check for errors in our
production logs:

b=`grep -c PANIC $e`;

Just add the grep -v in there:

b=`grep -c PANIC $e|grep -vi syntax`;

and now syntax errors aren't reported.