spl/zfs-0.6.0-rc4

Matthew Robbetts wingfeathera at gmail.com
Fri May 6 17:02:30 EDT 2011


Hi all,

Just giving this a try now. I import my single-disk pool (made by
zfs-fuse, pool version 26) using zfs and it seems to work. All
filesystems mount successfully and are listed with correct stats (usage,
etc).

However, after maybe a minute or two, the machine locks up. Actually it
seems to grind down to a halt, with small commands like ls taking longer
and longer to return, until the machine stops responding entirely.

This is after importing the pool but not actually doing anything with
it, just sat there idling. The first time it happened, I had started a
scrub immediately after importing the pool and I thought it was that. I
got good speeds for a while before the machine hung, so I thought it was
scrub's fault. But I don't think it is, now, since it also happens when
idling.

There doesn't seem to be any unexpectged CPU or memory usage being
reported by top before the lockup. There's no disk activity on the pool
disk either. Things just seem to... go wrong. The machine always
responds to ping, even at the end.

This has just happened three times in a row, so something's up. The
machine is fine until I load

Any ideas? Anything I should watch for or try to run? There's nothing in
dmesg or /var/log/messages.


I should mention that I'm on Gentoo, using the ebuild provided by devsk
in the science overlay. As there's no _rc4 version in the overlay yet, I
built the -9999 version of both the SPL and ZFS packages. As far as I
can see, there were only two commits since the rc4 tag was made - not
sure how much impact they could have.


Thanks,
Matt

P.S. Tried it one last time for luck - this time the machine froze
immediately upon `/etc/init/zfs start`.


On 05/06/11 00:20, Brian Behlendorf wrote:
> The spl/zfs-0.6.0-rc4 release candidate is available!
> 
> http://zfsonlinux.org/
> 
> Noteworthy fixes include:
>   #84  - Fixed 'zfs set volsize=N pool/dataset' crash
>   #190 - Allow NFS exported zfs filesystem
>   #165 - Allow uid/gids up to 2^32-1
>   #174 - Improved 'zfs scrub' stability by reducing stack usage
>   #191 - Fixed zvol deadlock
>   #136 - Fixed 2GB file size and zvol limit on 32-bit systems
>   #169 - Enforce ARC meta-data limits
>   #180 - Fixed stale znode/SA crash
>   #203 - Truncate xattr's when modified
>   #194 - zfs(8) and zpool(8) man page corrections
>   #xxx - Fix mounting zfs snapshots
>   #xxx - Added all known zfs tunables as module options
>   #xxx - Added default Gentoo/Lunar/Redhat init scripts
>   #xxx - Various build system fixes
> 
> A complete list of changes as always can be found in the git commit
> logs.  I'd also like to thank the following individuals for their
> contributions to this release candidate:
> 
>   Alejandro R. SedeƱo
>   Aniruddha Shankar
>   Darik Horn
>   Devsk
>   Fajar A. Nugraha
>   Gunnar Beutner
>   Jean-Michel Bruenn
>   Richard Laager
> 
> Finally, I have started using the newly introduced milestones on the
> Github issue tracker.  Since there are still a dozen or so known bugs
> outstanding there will be at least one more release candidate.  You can
> follow the progress of the 0.6.0-rc5 release candidate here:
> 
> https://github.com/behlendorf/zfs/issues?milestone=9&sort=created&direction=desc&_pjax=true&state=open
> 
> Additionally, longer term milestones have been created as a tentative
> road map.  This should give you a good idea when certain features are
> planned to be implemented.  The due dates are fictional, but the more
> help we get from the community the faster things will get done.
> 



More information about the zfs-discuss mailing list