BDF is showing Wrong Disk Usage!!!

[Khalid Dogar]

I am almost positive that output from the BDF is not correct. I had a volume 100% full, after removing significant amount of data it is still showing 100% full.

The " du -ks " reveals that there is almost half of the volum empty.

We are running Oracle on HP-UX 11i.

Any Ideas? why I am not able to reclaim the space back.

Thanks
Khalid

[Gordon Gable]

You have some open files left which have removed likely..Once these processes exit..your usage will come down.

YOu can check this using lsof and also using fuser -cu /FS...

Hope this helps
Thanks
Prashant

[RAC_1]

As Gordon Gable (aka Prashant) explained, it is due to the fact that, there are some processes that have open files.

You check them with lsof and fuser command. Until the processes are stopped, the space will not be recovered.

Anil

[Gordon Gable]

Hey sorry for it...I have to log off seems. He is my supervisor.

Thanks
Prashant

[Prashant Zanwar_4]

OK it goes here

Thanks
Prashant

[Hein van den Heuvel]

I like to be a contrarian and say NONSENSE to those deleted files suggestions.

You'll find this is due to SPARSE file usage, most likely by Oracle for temp files.

Your system will operate fine, but there is lurking problem. The DBA has overcommited storage, and if Oracle will ever actually use it, then it will get a more ugly IO error to go with the device full.

Please do an ADVANCED SEARCH ("more options" at top of this page) for 'bdf' in category hupx for 'subject line only'. You'll find dozens of similar reports and the first replies are always about deleted files still being open.
But so far I have not seen a single confirmation that this was actually the cause of the problem. So far I have frequently seen oracle mentioned at the same time as this problem being reported.

Note, I am not disputing that deleted files, which are still open, may well distort the used block pictures. I just don't think that they are the true cause behind the bulk of the reported mismatches. Those mismatches often do not sound transient, and they are often in the GB space, not MB space. Most system managers / DBA's know when Gigabytes of files are being deleted.

So, for starters I recommend you compare a DU output with ls -l for your oracle datafile directories, notably the one with temp.

And please, do report back.
Did the problem clear after an application stop/restart? Did you find those open+deleted files? Did the problem clear after a system restart, or did it perhpas persist?

btw.. Oracle is not the only application using sparse files, but just a more frequent/reported user.

Hope this helps,
Hein.

[Dave Wherry]

I hate to contradict a contrarian but, I've seen this before.

If you have a process that is sending output to a log file and that log file is deleted, the disk space will not free up until the process is killed. Some times it even took a reboot to clear it up.

[Prashant Zanwar_4]

I had similar experience too..
You can find out open files using lsof & then try seeing whats wrong exactly.
I even dont want to contradict on anything..I appreciate explanation by hein.

Hope it helps
Thanks..

[Todd McDaniel_1]

I might add my 2cents.

We have found the best course is to trace down the offending process and shut it down gracefully instead of trying to delete files first. Since this creates an untenable situation where you might have to reboot.

We have, on occasion, been able to find the process, shut it down and the 100% full syptom went away. Then restart the oracle process.

[Kent Ostby]

One way to trace down processes hogging your CPU if you dont have all the tools or just want a quick cut at it is:

ps -ef | cut -c42-80 | sort -nr | head

This will show you the top 10 processes that have used CPU and if you had something writng very large files, it will usually show up here as well.

Best regards,

Kent M. Ostby

[Sanjay Chavda]

---Clip---
... even if files are removed and disk usage in bdf is showing 100%, their inodes are not freed yet. once done, bdf should show you the correct disk usage.
lsof can be used to check this.

[Khalid Dogar]

I thank all of you to help me out here!

There can be many possibilites but in out case it was the deleted file for open processes. We have allowed DBA to delete files in application tmp directories and this is what happened. It wouldn't happen all the time though only once in a blue moon. Over the weakend when system came to relax a bit, we found the the lost disk space back in the " bdf " output.
I was going to recycle the machine but a quick bdf revealed the disk space is back.

Thanks
Khalid