I travelled twice in the past !!

I travelled twice in the past !!

[Edouard Gaulué]

Dear all,

I'm sory I didn't get the time to read the whole list and my Internet 
researchs kept sending me to linux soft-raid howtos. So I send a mail to 
this list.

This is the second time I travelled in the past due to my linux soft 
RAID 1 system (mis)configuration.

Here is what I can say :
- I've got plenty of production servers configured in RAID 1 using the 
2.4.24 debian patched linux kernel.
- I often go on those machines and a cat /proc/mstat seems to report me 
that disks are synchronized. Should I trust this kind of information ?
- Those machines often turn more than a year long without being reboot.
- Sometime due to hardware problems, or a need for a kernel change I 
have to reboot them.
- Sometime this reboot fail.
- This morning, it failled and I needed to use a Knoppix CD to get back 
to my system, chroot, repair the problem and reboot.
- I needed to assemble my array before to mount it, what I did with : 
mdadm --assemble /dev/md0 /dev/sda1 /dev/sdb1.
- The system answered it had to resynchronize the disks.
- After this, the data I got were about one year ago...

My opinion and questions :
- I thought that in RAID 1, writes were made on the 2 disks and reads 
half/half on the first and the second. Am I wrong ?
- I've got the feeling that starting a certain day  (don't know which 
one), it stopped behave this way. Is this happening often ? Is it a bug 
? How to detect this behaviour ?
- When I re-assembled the disk : no luck, it resynchronized the older 
with the knew one, so I get in the past.

Please help me, telling me if it's a misunderstanding on the way RAID 1 
should work, if I had to make more control with special tools, if it's a 
bug, if it depends on the kernel version...

EG

-- 
========================================================================
Édouard GAULUÉ
ENSG - IGN                             Réseau et Systèmes Informatiques
6-8 avenue Blaise Pascal               Activité IGS & DORIS
F-77455 MARNE LA VALLEE CEDEX 2        Master MSIAG
Tel : +33 (0) 1 64 15 32 43
Mail : Edouard.Gaulue@ensg.ign.fr

-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: I travelled twice in the past !!

[hahn]

> - The system answered it had to resynchronize the disks.
> - After this, the data I got were about one year ago...

you mean that you lost a year's worth of updates?

> - I thought that in RAID 1, writes were made on the 2 disks and reads 
> half/half on the first and the second. Am I wrong ?

no.

> - I've got the feeling that starting a certain day  (don't know which one), 
> it stopped behave this way. Is this happening often ? Is it a bug ? How to 
> detect this behaviour ?

my only theory is that one of your disks fell out of the raid a year ago.
that would explain why it contained old data.  but in rebuilding the raid,
the more recent disk should be overwriting the old one.  unless you do
cruel things to the raid superblocks, I don't really see how this could 
happen.  afaikt, someone would have to jump the event counter on the stale
disk to or past the fresher one, and I expect MD would only do that after a
rebuild succeeds...