> > Tom reported problems with max open files: > > oracle@(protected):/proc/sys/fs> cat file-nr > > 5676 1495 131072 > > Should be no problem for us here. > > To me open files looks low at 1024, but our old system is set this way > and never hung once.
Did your old system run the same OS and same version of Oracle with async IO? If not then it's probably not fair to compare the two.
I'd at least look at this as, based on your proc output, your system has at some point had over 5000 files open and even at that current point has 1495. I agree that it's unlikely to be your core problem but it's trivial to fix and thus to eliminate. Oracle actually suggest that, assuming a single Oracle instance on a system, you set the ulimit for that user to the same as the system limit. Lot's of Oracle setup docs suggest setting both the system file number limit and the Oracle users ulimit to 65535 or greater.
We used to get lots of ORA-03113 (See ORA-03113.ora-code.com) errors before we increased our limits. Are you having some of those? Our current production box looks like
# cat /proc/sys/fs/file-nr 7471 1849 65536
and if we left this system at the default it would definately do strange things after a few days (sometimes a few hours). A simple edit to /etc/security /limits.conf can change both the soft (default) and hard (maximum) limits for all system users, or individual users and groups.
