File table nearly full and Disk bottle-neck

[Ivan-Ho]

Hi everybody,

My database system running Oracle 8.1.7 with configuration as follows:

- OS: HPUX 11i
- Database: Oracle 8.1.7
- Memory: 2GB
- Swap: 4096 MB

However, database often be down with two message from GlancePlus appears like this : "File table over flow" and "Disk bottle-neck probability"

I am not aware of relation of both meesages? Any one tell me about that?

Can I increase nfile parameter value from kernel to fix both problems?

Thanks in advance,

[Sridhar Bhaskarla]

Hi,

No. There are not related unless your '/var/opt/perf/alarmdef' is misconfigured as 'Disk bottle-neck probability' comes from the configuration of that file.

To get rid of 'File table overflow', you will need to increase 'nfile' kernel parameter. Use SAM and increase it by say twice or atleast 50%. You can simply put the number instead of using the formula. Also consider increase nproc and uproc by a little bit. You can actually check the usage by looking at "t" window of glance or with the command 'sar -v 5 5'.

Disk bottle-neck probability, if it is real, then you will need to check the disks and see if their response time is ok. 'u' window of glance will show you all the disks that are at 100% with disk queues on them. Either you may have to move some of the tables from those disks to quieter disks.

Database crashes are due to 'nfile' parameter. Once you fix it, database will not go down again. "Disk bottleneck" will slow down the system. Post your 'sar -d 5 5' , 'sar -b 5 5', 'sar 5 5' , 'vmstat 5 5' and 'swapinfo -t' and someone will be able to help you.

-Sri

[Sยภเl Kย๓คг]

As Sri said
increase the 'nfile' parameter's value using SAM-->KernelConfiguration-->configurableparameters
nfile..double the value this and see.
This problem can be solved by tunning the kernel parameters.Refer to the Dbase reqirements by ur software u can takecare of this
regards
SK

[Ivan-Ho]

Thank Sri,

Attached file is from sar command.

Any help more?

[Sridhar Bhaskarla]

Hi,

You didn't use the arguments '5 5' with your sar commands. So, it pulled the data from the old history files.

Provide us the current data as specified earlier when you are getting 'Disk bottleneck probability' messages.

'nfile' increase should fix your intermittent database failures.

-Sri

[Ivan-Ho]

Hi Sri,

Forgive my mistake!

Here is sar command with argument "2 20" used.

I captured these at time that "Disk bottle-neck probability" message had appeared.

Any help more?

Thanks

[Sridhar Bhaskarla]

Hi,

From your 'sar -d' and other results, I am not really seeing any disk bottlenecks. Looks like you have only one disk c4t0d1 and it's only around 50% used and there isn't any queue built up. Generally sysadmins configure glance's alarmdef with a combination of "Disk Utilization" and "Disk queue". If you can attach your /var/opt/perf/alarmdef, one will take a look and advise about possible alterations to get real alerts.

Now the main question is "Do you get any complains from the users?" when the disk bottlenecks show up in glance.

-Sri

[Muthukumar_5]

"File table over flow" and "Disk bottle-neck probability" because open file descriptor limit reaching there.

You can know about this error informations on open() system call man page.

Refer this link for more,
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=706782

[Hoang Chi Cong_1]

Hi Hùng.

I think you should recommend with your manager to upgrade the memory.

I wonder that what kind of the business with your system?

Need for tuning is not easy, it depend on your system, aplication runs in your system.

Ji