RED ALERT File table is nearly full

[Ahmed_58]

Hi to all,
I'm getting this message when ever I use "glance" is this mean kernel paramter for nfile need to b increase? how can i found how much is used or % used now?

regards,
Ahmed

[Peter Godron · Accepted solution]

Ahmed,
for both answers please see:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=183679

[Steven E. Protter]

Shalom,

It means if you don't increase the kernel parameter an application will fail because its not permitted to open a file.

You can use sam to increase it. How much you increase it actually depends on how many files you want the OS to have open at the same time.

That you should be able to calculate to some degree based on the apps you run. Then increase it another 20%-50% for the sake that users like to run things.

SEP

[Frank de Vries]

It is like Steven said.

You can monitor the progression
also with

[root@orasrv2:]/etc<>>> sar -v 2 5

the fourth collumn file is the value of nfile.

Usually nfile and nproc are related to maxusers. So best practise is up
the other kernel parameter as well, so they
stay proportional.

It is never a good idea to increase parameters to much in one go.
Increase them by 25% or 50% and see how that
works after a boot.

[Ahmed_58]

Thanks gents for a quick reply,
Is there a way we can get an Alert in syslog, the message only comes on glance, and not in syslog! I wonder whats going to happen if it gets to 100%, will it crash the system?
Currently the nfile shows 94% & nbuf 100%

Ahmed

[Peter Godron]

Ahmed,
the syslog.log file will get an automatic alert once the table is full. This means your system will not allow any more files to be opened, which will, at the least, severly restrict the ability of your system. Applications will/may fail. Best way is to control the situation now and fix/reboot.
If you had a known bad disk, what would you do ?!

[Ahmed_58]

Thanks Peter,

But for us, it is very critical system, We should proactively be monitoring any critical UNIX parameters as part of normal Operations procedures in order to avoid crash, and arranged planned outages if required.

Ahmed

[Peter Godron]

Ahmed,
I fully agree with you. You are already using glance. If you want another warning system, please read (for inspiration):
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1035358

I would:
1. Investigate why the file table is filling up. Were new/changed applications installed ?
2. Change the kernel parameter based on some sar monitoring to help estimate values required.

[Raj D.]

Ahmed,

You need to increase the kenel parameters ,

1] nflocks
2] nfiles

You need a reboot.



You can check the current utilisation with:

# sar -v 5 5
or
# glance -t


Cheers,
Raj.

[Ahmed_58]

Thabks to all.