hello all -

we are on version 11.5.FC7 (from v10.fc5) running HDR on AIX 5.3

about 5 days after upgrading, the developers started getting an error
report from their cold fusion servers with several errors in the
debugging/catch information :

Description: Error Executing Database Query. Cannot open system
catalog (sysviews).

ErrorCode -311
Message Cannot open system catalog (sysviews).

ErrorCode -104
Message ISAM error: too many files open.



other times the error will read as :

ErrorCode -104
Message ISAM error: too many files open.


-242
Could not open database table table-name
(and the table name is sometimes a system table and other times a
regular database table
or will say
Could not open database table (%s)


the Cold Fusion debugging info shows the data source as usually being
the Secondary server in the HDR pair.


We have increased the nofiles (in ulimit) from 2000 to unlimited on
the secondary server
and also increased the nofiles on the app server as well but that did
not resolve the error.

On our dev tests, we are able to trap for the -242 error on the db
server (and send info to support) but the trap will not fire off on
the db server for the -104 error.

the DB message log does not ever show an error nor does the OS show
any errors in logging, debugging, etc

Also, we went from AIO to using KAIO about a month before the v11
upgrade.

Trying to get support from Adobe for Cold fusion is about as painful
(and as helpful) as sticking needles in your eye as we are exploring
what could be wrong in all areas since we really don't understand
where the problem is emanating from.

any ideas are appreciated.

thanks,
tom

On Nov 5, 3:21*pm, Tom Lehr <tomc... (AT) gmail (DOT) com> wrote:
The 104 error, assuming it is coming from the engine, does not have
anything to do with the physical number of open files. From looking
at the code it only comes from 2 places, and it's dealing with an
engine internal representation of open files, which has a limit of
like 32k. In 1 of the places it's set, it should actually trap with
the onmode -I, if that is not working, it would seem to imply it's
coming from the other location. It sounds like it could be some sort
of internal open table structure leak which is making the engine think
you have more then 32k tables open. I would possibly suggest using
onstat -g opn to monitor the number of tables each thread has
open....maybe try to see if any threads have a continuiously
increasing open table count, or any open tables counts near 32k.

Jacques Renaut
IBM Informix Advanced Support
APD Team

On 05/11/2010 21:13, jrenaut wrote:
Could also check:

ls -l oninit - i.e. ensure correct permissions
and
ulimit -a - i.e. environment limits when starting the engine
there may be something in
errpt -a