I am running 11.11 on a L1000 machine which has 2 disks, 1 for operating system and supporting software, the other for application development etc.
The OS disk hardly changes, but the AppDev disk has 100's of updates/creates/deletes per week. I have not bothered with trying to tune this disk's performance as yet, and have to admit to being new to this area of HP-UX Sys Adm.
So, I was hoping that the kind people who frequent this forum would provide me with an explicit recipe to defragmenting (if necessary) this disk, and give any other tuning tips as you deem fit. I hope I have supplied enough detail, if not let me know and I shal try to answer any questions as necessary.
The research I have done so far seems to indicate that I need to use Veritas software (fsadm_vxfs, etc), and this is what I find under swlist "Base VERITAS Volume Manager 3.2 for HP-UX"
Is this the software being referred to?
Thanks in advance for your detailed responses... assume I know very little on this subject.
With just one disk for application, you have a very little to do. What application is this?? Depending on the application and the data it accesses, it is going to be difficult to know which data is accessed most and which least.
The defragmentation can be done with fsadm -d ad -e options. read the man pages of vxfs_fsadm. If you do not have OnlineJFS(swlist -l product OnlineJFS), you will have to unmount the FS and do the defragmentation work. But to work it, you should have some free space on the FS.
A better defragmentation would be doing a full cold backup and restore.
Stop all applications accessing the FS. Back it up. newfs on the FS and restore. Start the application.
Also you may want to look at your buffer cache setting.
This has nothing to do with VxVM but only pertains to fsadm -F vxfs. THe man pages for fsadm_vxfs should be all to need. Just invoke fsadm -F vxfs -d -e /fsmountpoint and you are good to go. This can be done while the filesystem is mounted.
In general, don't expect big gains from running this. Vxfs handles fragmentation well and in the UNIX world many, many file changes per second is the norm. Any filesystem that could handle that would simply never make it to prime time. There are a few corner cases where defragging noticeably helps but most of the times any gains are difficult to measure and extremely difficult to perceive.
{root} /> swlist -l product OnlineJFS # Initializing... # Contacting target "hp2"... ERROR: Software "OnlineJFS" was not found on host "hp2:/". {root} />
Is there any other way to do the defrag without using fsadm/Veritas.
Perhaps a little more background might help. We use the L1000 machine for porting work only. The software we develop is generated from a Model-Based development tool, and we generate the code for Win/Oracle, Win/MS-Sql, Unix/Oracle and MVS/DB2, all of which are automatically compiled and linked as part of our development process. So we seldon log on to the L1000, except for testing our software, most of which happens on the Win/Oracle platform... Recently, we have noticed that the compiles and links of our Unix/Oracle code is taking longer than it used to, and I figured that it is probably down to the HD needing some TLC, hence my posting...
We have a Dev Partner relationship with HP since our software is certified to run on HP h/w, and as such, could possibly get Veritas, but at present it does not seem that we have it.
Anyone who is interested (curious) in our software can look at our website (www.iet.co.uk)
Also, since this machine is only used for porting work, it has no Tape Drive and does not get backed up, so backing up and then doing a "newfs" isn't really an option I want to follow right now... thanks anyway.
Doesn't look like you have OnlineJFS license on the box to do the defrag of the filesystem. You need to have onlineJFS on the system. OnlineJFS is a $$ product and is sold seperately.
cat /usr/local/bin/fsadm.defrag #!/bin/sh # defrag all mounted file systems LOG=/tmp/fsadm.defrag.log if [ -f $LOG ] then mv $LOG $LOG.old fi cat /dev/null > $LOG
for i in `mount -l | grep -v stand |awk '{print $1}'` do echo "defraging " $i >> $LOG fsadm -F vxfs -d -D -e -E $i >> $LOG done
Rgds...Geoff
Proverbs 3:5,6 Trust in the Lord with all your heart and lean not on your own understanding; in all your ways acknowledge him, and he will make all your paths straight.
It's probably too early to tell. A look at the man page tells us "If the totals in the columns labeled "Dirs to Reduce" are substantial, a directory reorganization can improve performance of pathname resolution." The examples show "Dirs to Reduce" in the 2500 to 3000 range, far above your 46. This may run for a long time but I don't think you're going to see any substantial improvement.
You might want to start looking at other performance metrics of this machine to see what is slowing it down.
Ok, now that I have OnlineJFS and fsadm is running where else can I start looking for some perfomance improvements? Perhaps I should start a new thread for this one?
Thanks everyone for the help, I left the fsadm task running last night and it completed successfully, now I need some help analyzing the results... See plain text attachment:
From your output, it looks like it was a rousing success! Since the intent is to gather all the free space into the biggest blocks possible, and your report shows:
Extent Fragmentation Report Total Average Average Total Files File Blks # Extents Free Blks 64621 117 1 1481069
That's telling me that you've got 1 free extent of 1481069 blocks, which is the ideal outcome.
