SUMMARY: rmvol hangs when volume too large?

From: Carlos Chua <chuacarlos_at_hotmail.com>
Date: Tue, 23 Jan 2001 22:50:34 +0800

Thanks to the following for their valuable inputs:

Dr. Thomas P. Blinn
Mandell Degerness
Steve Hancock
Gary Phipps
Pat O'Brien
alan_at_nabeth.cxo.dec.com

The answer is no, there should not be a problem using rmvol for my volume
size of 200GB.

Based on the response I got, I'm now thinking if I've run defragment before
doing rmvol, I might have a better chance of not facing this kind of issue.
However, for my case, we finally have to shutdown the server. Worst thing is
after the reboot, the domain can no longer be mount (complains about link
mismatch, diff effort was tried but nothing works) and so we have to
recreate the domain and filesets and then restore from tape : (

Below are extracts of the replies that I like to share:

---
There should not be any size limitation.  200GB is not that large.  I do
not know whether there are any know problems with rmvol that would explain
the high CPU use; you might be hitting a bug.
Since V4.0E is no longer a supported release, I suspect you can not even get
the patch kit to check the bug list, and that won't help you if you are in
the middle of the process anyway.
---
I did this successfully on a domain which had a total of over 200GB.  The
individual RAID sets which I was adding and removing were 40 GB and 20 GB
respectively.  I never saw the behaviour you are describing.
---
I'm guessing with that large domain, you probably have a very large
frag file. This file, which keeps the very small file fragments, can
exist on any or all of the volumes in the domain. So, with a very
large number of small files it can get quite big. If it is large, it
could take a long time to migrate the extents that reside on from
volume A to the other volumes. This file has a tag number of one and
can be examined using the command "showfile -x /<mount-pt>/.tags/1".
---
I have recently migrated 2Tb of data using addvol/rmvol on a GS140 (AS8400)
clustered pair.
My advfs domains are 100Gb each. Some of the files within these advfs
domains are over 60Gb each.
Things I noticed with this :
It takes a long time (depending upon how your disk attached).
It does consume a large amount of cpu. It seems to run in stages (I think
it's at a file level).
We us cpuinfo to monitor our systems, and at certain stages all 10 cpus were
in I/O !
Of course, I don't need to say 'Backup your data beforehand' do I ?
One word of caution.  We run ASE cluster - one of our services is an NFS
filesystem.  Migrating this caused HUGE problems. -resulting in a reboot !
I'm not sure if it was NFS or ASE that caused the problem..... but be aware.
---
I do not have the direct answer for you, but can add insight.  There is a
special metadata area on the disk which does get starved during large advfs
utilities data operations. I too have this issue, and my senior had a
procedure for increasing the "I think he called this the bmj area", but he
has left with his procedure.  Although not a direct answer, you are looking
for a procedure to increase your metadata area, so large file operation will
succeed.  hope this helps
---
If the volume being removed has lots of files, it may
take considerable time to clean up the metadata being
changed.  In V4 and earlier, some of the algorithms
used become very CPU intensive when there are lots of
files.  I don't recall if this an absolute "lots of
files" or "lots of files in a single directory" sort
of problem.
I don't think this a limit of due to large volume, but
simply a consequence of a large volume.
---
----- Original Message -----
From: "Carlos Chua" <chuacarlos_at_hotmail.com>
To: <tru64-unix-managers_at_ornl.gov>
Sent: Saturday, January 20, 2001 7:26 AM
Subject: rmvol hangs when volume too large?
> Hi,
>
> I'm currently using the addvol/rmvol command to migrate a disk that
contains
> 200G of data. My environment is V4.0E. I'm using the showfdmn command to
> check the status of the file migration. It seems to me that whenever the
> rmvol process is about to finish, the rmvol process would start taking up
> 99.9% CPU time. Now the process has been stuck for more than an hour, I'm
> afraid the server might crash anytime soon as has happened last week.
>
> My question is, as I was able to do the same process on a smaller size
> domain, but only have problem with a bigger size domain, does it mean
there
> is a size limitation when using the rmvol command?
>
> Any help greatly appreciated.
>
> Regards,
>
> chuacarlos_at_hotmail.com
>
> _________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
>
>

Received on Tue Jan 23 2001 - 14:53:01 NZDT