[Btrfs-devel] Re: btrfs stability

Sebastián Peyrott pseus7 at gmail.com
Sun Mar 9 04:55:08 PDT 2008


Rekrutacja <rekrutacja119 <at> gmail.com> writes:
> 
> it doesnt help, setting 'ikeep' neither
> 

Have you tuned the settings of your XFS partition to match your RAID settings
(sunit/su, swidth/sw, etc.)? If not, you might want to look into that.
Other than that, it may be wise to bring this up in XFS' mailing list. I'm
pretty sure its devs will like to hear about any performance regressions you
might have experienced lately. IMHO, the recommended options for a high
performance, *backed-up*, XFS partition are: 
 * "-d unwritten=0 -l size=128m,version=2,lazy-count=1" (plus sunit, swidth or
other RAID settings) for mkfs
 * "nobarrier,noatime,filestreams,logbufs=8,logbsize=256k" for mount.

Please note that this may put your data at risk in the events of a crash
(particularly the 'nobarrier' option).

Tuning the number of access groups ('agcount' or 'agsize' for mkfs.xfs) might
make a difference, but I wouldn't know what the appropriate settings for your
particular configuration are.

In any case, politely asking for pointers in XFS' mailing lists might give you
some answers.

Good luck.




More information about the Btrfs-devel mailing list