On Wed, 07 Dec 2011 21:39:34 +0100 Sigurd Mytting sigurd@mytting.no wrote:
On 2011-12-07 10:29, Sigurd Mytting wrote:
Hi,
Hi :-)
I've run into trouble after upgrading to 3.9.3 via apt-source, system is debian 6.0.3, this was not a problem with 3.9.2.
First the message during upgrade:
Setting up nav (2+3.9.3-1) ...
I did a little digging and found the statement timeout was due to a 30 second time limit on transactions in postgresql.conf, changed that to the default value of 0 (wait forever).
Ouch, that could have been a problem with many long-running queries.
After then trying to run syncdb.py for seven hours I terminated it. Some more digging showed the statement "DELETE FROM rrd_file WHERE netboxid IS NULL;" did not terminate, but "select *" gave over 7700 lines. Deleting them one by one using the rrd_fileid and "netboxid IS NULL" cleaned up the database in a few minutes.
"DELETE FROM rrd_file WHERE netboxid IS NULL" should never take 7 hours. I'm guessing something was holding on to a database lock that the statement needed to wait for.
Does a "ps axuw" listing show many postgres processes with the status "idle in transaction"?
After adjusting the timeout value I can delete devices, but still get errors when adding devices - however bulk import seems to work.
What kind of errors?
Don't know if it's related, but the postgres-log contains a lot of "LOG: unexpected EOF on client connection".
It's not related.