file table full

[Marco_113]

Hi all,
I've a service guard cluster for a mail system;
on the first node runs POP3 and SMTP packages and on the second node runs IMAP package.

The SMTP package has stopped by itself because the nfs has gone down;

from the syslog.log:

Sep 21 16:03:07 vmunix: file: table is full
Sep 21 16:03:07 above message repeats 10 times
Sep 21 16:03:07 CM-SMTP_pkg[4478]: cmhaltserv SMTP_pkg.monitor
Sep 21 16:03:09 vmunix: file: table is full
Sep 21 16:03:09 cmcld: Service PKG*26369 terminated due to an exit(1).
Sep 21 16:03:09 cmcld: Halted package SMTP_pkg on node xxxxx.
Sep 21 16:03:09 cmcld: Package SMTP_pkg halt script exited with NO_RESTART.

I've seen in the pop and mail log, but the number of processes (fork) generated in that period are similar to those of other days.

What can be the cause of a similar problem??
Thanks

[Geoff Wild]

You need to increase nfile and maybe nproc....

See http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=593058 for more.


Rgds...Geoff

[bhavin asokan]

hi,
this is a common problem.in this forum just search by the keyword table is full.you will get enough examples.

you have to increase the values of nfile,nproc,maxusers entries in kernel.through SAM you can do it.

regds,

[Pete Randall]

Berfore you go blindly increasing kernel parameters, it might be a good idea to take a look at which ones are actually in need of adjustment - obviously nfile is. If you have Glance installed, take a look at the System Tables report - that will tell you where you stand on most of the common parameters. If you don't have Glance installed, there is a trial version on the application CDs that you can install.


Pete

[Steven E. Protter]

Leave maxusers alone. nproc and nfile are by default set on formulas calculated from maxusers.

best break those formulas. Thats what Bill Hassell says he does.

You may use sam and the new kernel will require a boot.

:-)

SEO

[Marco_113]

Thanks,
I've already changed the values on nfiles nproc and ninodes with sam.
It's the second time (in this year) that this problem occours and is the second time that i change the values of kernel.
So i should like to know why this problem occours; normally the number of files opened in the file table is lowest than the limit; but it's the second time that suddenly the table is filled.
Looking the logs of mail and smtp ther's nothing anomalous or different from the other days.

[Prashant Zanwar_4 · Accepted solution]

Read carefully about all the parameters nfile, nproc, maxusers. Try and see sar, glance outputs for a period of time. Compare values on different systems and applications running on different system. And tune accordingly. Refer to the document below, it's quite a wonderful information on K-Parameters. Last but not least, please dont forget to assign points.

Hope it helps
Thanks and regards
Prashant