geom_raid5 livelock?

[R. B. Riddick]

Thank you for the tests... :-)

--- CyberLeo Kitsana <cyberleo at cyberleo.net> wrote:
> I haven't forgotten. I was just trying to figure out how to build and 
> use the new module without having to reboot the box. Guess there's no 
> safe way.
>
Yup... I always remove all disks, when I am testing (then I can unload the
module)... But then u had to do a rebuild, which is not so good...

I do not know, why GEOM taste-s the disks again, after the device has been
stopped... I havent found out the trick yet...

> http://home.cyberleo.net/cyberleo/workspace/Zip/graid5-all3.log
> That's the latest log, with the debug lines inserted.
> 
OK - So we have a false parity at 264438038528.
Now we could
1. (hard way)
Look at the data (16KB) at that offset of all disks and see, where exactly the
parity mismatches ("dd if=/dev/adXs2 bs=16k iseek=16140017
of=/tmp/adXs2.test..."; then we could XOR them all together and see, where we
have non-zero bits)
or
2. (easy way)
Do a re-sync (graid5 configure -R raid5) and look at kern.geom.raid5.veri_w
(the number of bad parity blocks).

The re-sync seems to be necessary, because it is very likely, that ur graid5
device has bad parity blocks... It is difficult to say, how that happened
(could be an unintentional "graid5 configure -R" or a bug in graid5 or...)...
If you can reproduce it (maybe with a smaller device?), I would be glad, if you
could let me know, how u did it...

-Arne


 
____________________________________________________________________________________
Yahoo! Music Unlimited
Access over 1 million songs.
http://music.yahoo.com/unlimited