file table is full error in syslog.log

[stephen peng]

dear all,
i've got following messages in syslog.log:
Nov 10 11:16:20 gx_sd01 vmunix: file: table is full
Nov 10 11:16:20 gx_sd01 vmunix: file: table is full
Nov 10 11:20:20 gx_sd01 above message repeats 10 times
and at the same time, oracle's alert.log produced the following messages:
ORA-01115: IO error reading block from file 5 (block # 47858)
ORA-01110: data file 5: '/dev/ctais_datavg/rlvctais10g07'
ORA-27041: unable to open file
HP-UX Error: 23: File table overflow
Additional information: 3
ORA-01115: IO error reading block from file 201 (block # 30346)
ORA-01110: data file 201: '/dev/ctais_datavg/rlvctais10g01'
ORA-27041: unable to open file
HP-UX Error: 23: File table overflow
what could cause this problem? should i modify the tunable parameters?

thanks a lot

[siva0123]

Hi ,

Yes you have to tune the following tunable parameters.

1.nproc
2.maxfiles

Thanks,
Siva

[Patrick Wallek]

Yes, you need to modify the nfile kernel parameter.

You can check file table usage by using 'sar -v'.

# sar -v 5 5

and look at the 'file-sz' column.

# man sar for details.

[siva0123 · Accepted solution]

Stephen ,


Also you have to tune your kernel paramater nfile.

This message indicates that you have exceeded the maximum number of open files allowed on the system. To correct this problem, you must increase the value of the kernel parameter "nfile".

to check abt nfile,
http://docs.hp.com/en/5991-0710/ch04s01.html


Thanks,
Siva

[James R. Ferguson]

Hi:

The "file table is full" message says it all. You need to increase the 'nfile' kernal parameter value.

'nfile' must be greater or equal to 2-times 'maxfiles_lim'. Be generous with 'nfile's value, each increment consumes very little kernel memory. The maxiumu value is that of a 32-bit signed integer. A value of 65,536 is the default for systems with at least 1GB of memory.

Regards!

...JRF...

[siva0123]

http://docs.hp.com/en/B2355-60105/nfile.5.html

Thanks,
Siva

[DCE]

One way to check usage is via glance

press "t" to get current table usages

And that should let you know how close some of the parameters are to their limit