Skip to contentSkip to conversation

raid1 oops in 2.6.9 (debian package 2.6.9-1-686-smp)

DISCUSSION Robert Murray 4 participants 5 messages Nov 28, 2004 – Nov 30, 2004

lore
Share & tools
lore
#0 lore raw

Hi

The complete console log can be found at http://haylott.plus.com/~robbie/md-oops.txt

hde is a failed drive. In this log, hdg (the other drive in the raid1 array) is not present. This oops also occurs when hdg is present. I don't know why it tries to use hde when it has been failed for some time now. This doesn't occur with 2.6.8 (also a debian kernel). I don't have a log of the oops when hdg was present, but I can provide one if necessary.

Please let me know if there is any other information I can provide to help to debug this. For now I have removed hde and everything is working fine.

Best regards

Rob

#1 lore raw
Quoted 11 lines · Robert Murray

Hi

The complete console log can be found at http://haylott.plus.com/~robbie/md-oops.txt

hde is a failed drive. In this log, hdg (the other drive in the raid1 array) is not present. This oops also occurs when hdg is present. I don't know why it tries to use hde when it has been failed for some time now. This doesn't occur with 2.6.8 (also a debian kernel). I don't have a log of the oops when hdg was present, but I can provide one if necessary.

Quoted 3 lines

Please let me know if there is any other information I can provide to help to debug this. For now I have removed hde and everything is working fine.

On a second note: Could someone please provide an explanation of why the raid10 driver exists? People have created RAID-10 sets for years using the RAID-0 driver on top of several RAID-1 arrays - this works beautifully, it's simple, and it's easy to explain to people.

Why oh why, do we need raid10 ?

(I don't mean to bitch and moan over it - I just assume that there is a good reason for it which was somehow never conveyed, or that I overlooked in my search for this explanation)

And; if raid10 does not provide new functionality that was not possible with raid1 + raid0, why oh why does this get accepted in a stable kernel
series? (ok, 2.6 is not stable, but I assume the intention is to make
it stable eventually, and accepting new functionality does not help this process - all in all I do not understand the raid10 submission at all, but I hope to be enlightened by someone (Neil?))

Also, I'd love to add a mention of raid10 in the HOWTO, but I need to know why raid10 even exists before I can reasonably do that.

Signature · 1 line
 / jakob "baffled Software-RAID HOWTO co-author"
#2 lore raw

From: Jakob Oestergaard <[email protected]>
Date: Mon, Nov 29, 2004 at 11:07:08AM +0100

Quoted 1 line

Why oh why, do we need raid10 ?

Raid-10 allows things currently not possible with raid-0/raid-1, like spreading 2 pieces of data over 3 pieces of harddisk.

Their was an introductory message on the linux-raid mailinglist, but it's more than one month old so I don't have a local copy.

Quoted 3 lines

And; if raid10 does not provide new functionality that was not possible with raid1 + raid0, why oh why does this get accepted in a stable kernel series?

New drivers that are not enabled by default have always been allowed in stable kernels, since they don't have an impact on stability for the average user.

My $0.02,
Jurriaan

Signature · 3 lines
If something was not wrong things would not be right.
        Sergeant Ortega - Zorro
Debian (Unstable) GNU/Linux 2.6.10-rc2-mm3 2x6078 bogomips load 1.44
#3 lore raw
Quoted 6 lines · Jurriaan

From: Jakob Oestergaard <[email protected]>
Date: Mon, Nov 29, 2004 at 11:07:08AM +0100

Quoted 1 line

Why oh why, do we need raid10 ?

Raid-10 allows things currently not possible with raid-0/raid-1, like spreading 2 pieces of data over 3 pieces of harddisk.

Sounds weird to me, but hey, that's probably just me :)

Quoted 3 lines

Their was an introductory message on the linux-raid mailinglist, but it's more than one month old so I don't have a local copy.

I must have missed it when I looked for it then - I'll look again.

Thanks!

Quoted 8 lines
Quoted 4 lines

And; if raid10 does not provide new functionality that was not possible with raid1 + raid0, why oh why does this get accepted in a stable kernel series?

New drivers that are not enabled by default have always been allowed in stable kernels, since they don't have an impact on stability for the average user.

True

Signature · 1 line
 / jakob
#4 lore raw
In reply to Robert Murray
Quoted 4 lines · On Sunday November 28, [email protected] wrote:

Hi

The complete console log can be found at
http://haylott.plus.com/~robbie/md-oops.txt

This looks like a known bug that is fixed in current 2.6.10 pre-releases.

Quoted 5 lines

hde is a failed drive. In this log, hdg (the other drive in the raid1 array) is not present. This oops also occurs when hdg is present. I don't know why it tries to use hde when it has been failed for some time now.

It tries to use hde because it sees no reason not to.
When a drive fails, md never writes to it again, so the record of it being part of a raid1 array is still there.
If it is assembled with another drive that "knows" that hde has failed, then it won't accept hde into the array. But the hdg missing, hde is the best bet it has, and it tries it anyway.

NeilBrown