Can stride be removed from ext4?

Milamber75 · 08-30-2011, 05:48 PM

Hi!

So I'm playing around with a LVM over RAID0 setup using mdadm. I created a throw-away ext4 filesystem, pvmoved it to the RAID0, then tuned it with a stride size of 8. (My rationale was that I had chosen a 32k chunk size, and 32/4 = 8. Not rocket science by any means.) However, before I start moving filesystems I care about onto the PV that is sitting on the RAID0 and permanently changing their attributes, I want to know that I can move them back. That means I want to be able to tell the filesystem that it no longer needs a stride. But try as I might, I've not been able to remove the stride extended option from the filesystem. I've tried several commands, including:

Code:

sudo tune2fs -E stride=0 /dev/notimportantforexample
tune2fs 1.41.12 (17-May-2010)
Invalid RAID stride: 0

Bad options specified.

Extended options are separated by commas, and may take an argument which
	is set off by an equals ('=') sign.

Valid extended options are:
	stride=<RAID per-disk chunk size in blocks>
	stripe_width=<RAID stride*data disks in blocks>
	hash_alg=<hash algorithm>
	test_fs
	^test_fs

Is it even possible to remove this extended option from a filesystem?

Thanks in advance.

zhjim · 08-30-2011, 06:16 PM

Raid0 lives of stride so stride=0 does not make sense. Just move it of the raid0-lvm and it should leave the stride behind. (Can't put it any better)

Milamber75 · 08-30-2011, 06:34 PM

Tried what you said. I unmounted it, pvmoved it on to a PV that's not on RAID, then remounted. It didn't work. Here's the result.

Code:

sudo tune2fs -l /dev/dontworryaboutit | grep RAID
RAID stride:              8

zhjim · 09-05-2011, 12:28 AM

Maybe removing the stride now works?

Also if you just want to make sure that the filesystem is still working on non-raid0 volume I would check that out and if it does do not care about the stride settings of tune2fs. I guess it would just be ignored when the filesystem is not in/on a raid.