File system and Smart arrays

[Anders Norrbring]

I don't know if my problem is the file system or not, so I ask..

I have a server running on a Smart Array 3200 and 12 x 18.2GB drives.

There are 2 volumes, both are currently formatted as Reiserfs, this because it's the default in SuSE Linux.

I've noticed that the system "waits" for extended amount of time when file activity increases. That makes the system look like it's hanged, but then continue after 10-30 seconds. Can this be related to reiserfs, or should I look elsewhere?

If it's reiser, what other file system should I use for good performance?

Anders.

[Florian Heigl (new acc) · Accepted solution]

Hi,

the controller could experience a scsi reset, which would result in a short hang.
Hopefully You'd get some info in the syslog or kernel.log !

Also, there could be a real performance problem, i.e. if You have a RAID5 array spanning both disks which then is split into the two volumes.
On a copy from vol1 to vol2 You would run into *bad* performance, whenever ReiserFS flushes it's Metadata, filesystem log and real data.

If this system is not loaded with billions of files, but rather has a regular distribution like a file server, I'd dare saying that XFS will outperform *anything*

[Anders Norrbring]

Thanks a lot, I'll reformat with XFS, just to check.
And no, the 2 volumes are separated, one is mirror+spare, the other is RAID-5+spare, so the doesn't interact.

Nothing seem odd in any logs, that's the strange part..

I'll post my findings later..

Anders.

[Florian Heigl (new acc)]

To add in on that - please look through the adapter man page.
My Smartarray 431 is running in a freebsd box, there I also get close to no messages from the driver, I do not know if they linux driver is more verbose.
There appears to be a cli utility
http://h18000.www1.hp.com/support/files/storage/us/download/22168.html - maybe it can poll the controller's event log.