bin/144214: zfsboot fails on gang block after upgrade to zfs v14

[Bartosz Stec]

W dniu 2010-05-13 08:44, Volodymyr Kostyrko pisze:
> 2010/5/13 Andriy Gapon<avg at icyb.net.ua>:
>    
>> This is what I am confused about.
>> What do you mean by the same 'boot code'?
>> I originally asked if you have the *latest* bootcode installed, but you seem to
>> be giving contradicting answers...
>>
>> You detailed your magic workaround of re-writing data, but you haven't mentioned
>> how you updated your boot code.
>> Could you please report the actual commands that you use to update the boot
>> blocks?  Just so we are sure that the problem is indeed present in the latest code.
>>
>>      
I am watching this thread from some time and maybe it's worth to mention 
that my raidz pool was hit by this bug also on CURRENT on april 30. I 
did make buildword; make buildkernel; make installkernel; reboot and I 
didn't realized that at this moment zpool was 98% full (about 300-400MB 
free space left). Loader catched gang block while trying to loading 
kernel and after a moment it give up, informing that it can't find from 
what device is booting from.

So at this moment my system was in non-bootable state, with installed 
kernel but without new world installed. Zpool however was fine, I didn't 
have any problems accesing it from fixit environment. So what I did at 
first time was installing world and updating bootblocks from freshly 
builded /boot using fixit dvd. Of course it left my zpool with even less 
free space. After reboot loader text show itself for about half a second 
and after that I could only watch blank screen.

I also tried some other things from fixit environment: installing 
bootcode from 8.0-release (it failed because zpool v13 vs v14 mismatch), 
revert back from snapshot from time before new world and kernel was 
build (gang block again), and finally I fixed issue by deleting all 
snapshots and deleted as max files as possible (so I had about 5GB free 
space left) and then:

cp -pR /boot /boot2 ; mv /boot /boot.bak ; mv /boot2 /boot ; reboot

As my system was bootable again, I have csupped new sources, rebuilded 
world and kernel, updated bootcodes and everything was fine. It seems 
that bug can be catched only with almost full zpool.

-- 
Bartosz Stec