topic Re: Help in Operating System - OpenVMS

Help

vmsserbo — Tue, 08 Aug 2006 16:32:48 GMT

Can anyone recommend how to fix a couple disk space issues and fragmentation problems.

Here is an example of our problem on one of our nodes.

Re: Help

Robert Brooks_1 — Tue, 08 Aug 2006 16:49:12 GMT

Get a disk defragger. We sell DFO (Disk File Optimizer); there are other 3rd party products
of varying quality.

-- Rob

Re: Help

John Gillings — Tue, 08 Aug 2006 17:05:30 GMT

Miles,

Again the title "Help" is not very descriptive. It certainly won't help anyone trying to find this topic in the future. Please change it to something like "How to fix file fragmentation"

The files you describe are fragmented, but not too severely. You can't do anything about it while they're open. Assuming you have some time window when they're closed, I'd suggest you CONVERT each file to clean it up both inside and outside.

$ CONVERT [RMS]EDOCHIST.DAT [RMS]EDOCHIST.DAT

Given the level of fragmentation, this is NOT urgent. After the CONVERT if the file is still fragmented, try:

$ COPY/CONTIGUOUS [RMS]EDOCHIST.DAT [RMS]EDOCHIST.DAT

To avoid the file fragmenting in the future, work out its likely growth pattern and give it an appropriate EXTEND quantity. If I'm reading your report correctly, the file is around 450K blocks, so 50K blocks wouldn't be unreasonable:

$ SET FILE/EXTEND=50000 [RMS]EDOCHIST.DAT

You need to know the data to decide on the best value. How much do you expect the file to grow? How much disk space are you prepared to allocate to it now?

Of course, this will only help if you have sufficient contiguous free space.

If you have it, run DEFRAG against the disk to consolidate the free space.

It's not clear from the display what your "free space issues" are. If you're running low, the most cost effective solution is often "buy more"!

Re: Help

Robert Gezelter — Tue, 08 Aug 2006 18:12:51 GMT

Miles,

For that number of files, my first attempt would be to do CONVERT with an appropriate FDL (as John said).

I would also consider what the expansion requirements of these files are, and allocate the new files accordingly.

If the disk has a big enough block of free space, this should work. If the entire disk is fragmented, I would consider using a defragmenter.

- Bob Gezelter, http://www.rlgsc.com
P.S. I second John's comments about titling the post in a clearer fashion.

Re: Help

Andy Bustamante — Tue, 08 Aug 2006 20:15:03 GMT

In addition to the other replies you have, I'll chime in with the budget answer.

BACKUP can be used as a defragmentation utility and it's included with the operating system. If you have a spare disk, make sure all files are closed and create a disk-to-disk /IMAGE copy. Either update the logicals used for data access or change the disk identifier.

If you don't have a spare device, you can create a tape backup and image restore. Be very sure your backup device is reliable before trying this. The VERIFY switch is recommended.

As mentioned above, set a reasonable file extent to help avoid fragmentation issues in the future.

Andy

Re: Help

Wim Van den Wyngaert — Wed, 09 Aug 2006 01:10:22 GMT

On the freeware cd (or on the net) you find the utility DFU that can defragment 1 file at the time. With the command defrag/besttry you should defrag the file if possible. If your disk is too defragmented it may fail.

BTW : may be the files don't get defragmented because they were open during the defragment process (if any).

Wim

Re: Help

John Abbott_2 — Wed, 09 Aug 2006 01:27:20 GMT

Hi Miles,

If you haven't already, I'd get a copy (for free) of DFU from here

http://h71000.www7.hp.com/freeware/freeware70/dfu/

...It's very quick on a whole bunch of useful tasks (has online help) and can defag individual files just fine.

If you want something that'll defrag more than an individual file HP's Disk File Optimizer (DFO) has a varied level of defrag levels. You can download an use the reporting version for free here:

http://h71000.www7.hp.com/openvms/storage/dfopage.html

It's been a while but at the time we felt the cost was reasonable and justified for the volumes we cannot easily backup/image and restore, so we use it regularly to consolidate free space.

One final point, if any of these files are created regularly (new files ;1 say from a month end batch run) and you know how big they get, you could investigate changing the application to create preallocating the required space and/or apply an extension too. If they're created via .FDL files then you can also specify values within the FDL too.

Kind Regards
John

Re: Help

Ian Miller. — Wed, 09 Aug 2006 03:53:48 GMT

A more important question is what real impact is this fragmentation having?

Is the performance of your application currently acceptable and are these files frequently accessed?

Look at window turns (MONITOR FCP) and split I/O rate (MONITOR IO). Unless this a significant percentage then the fragmentation is not a problem. Don't fix problems that are having no significant impact.

Re: Help

Jan van den Ende — Wed, 09 Aug 2006 12:54:23 GMT

Miles,

even if you DO have a relatively high WindowTurnRate, this CAN be overcome more easily.
You can enable "Cathedral Windows", ie, you always load ALL of the file header in memory.
You achieve this by setting SYSGEN param ACP_WINDOW to 255 (max, special meaning: ALL)
Unfortunately it requires reboot, and it can better be left to AUTOGEN to calculate some related params.
This completely eliminates WindowTurns.
There are those that say it is as good as the best defragger.
I do have some doubts on that, but it sure eliminates all defragger overhead. And it does NOT consolidate free space!

(You DO need a bit more physcal memory (in the order of the total size of the excessive fileHEADERS, so several KB [er very fragmented file probably less than 1 MB if you have less than 100 VERY frahmented files)

fwiw

Proost.

Have one on me.

jpe

Re: Help

Ian Miller. — Wed, 09 Aug 2006 13:38:44 GMT

Cathedral Windows does cure excessive window turns but not split I/O.

It depends on what your problem (if any) is.

Re: Help

John Gillings — Wed, 09 Aug 2006 17:35:24 GMT

Miles,
Huh? Something wrong with my answer?