[zfs-discuss] Fragmentation vs traditional linux FS and ZFS as a datastore for SQL db's

Richard Yao ryao at gentoo.org
Sat Jan 18 09:41:40 EST 2014


On 01/13/2014 06:25 PM, Andrew Holway wrote:
> The
>> article above, though, makes the case that fragmentation is terrible and a
>> deal stopper.
> 
> Fragmentation is a somewhat severe performance problem when you start
> hitting 70 or 80% of the pool capacity. This is because the algorithm
> that looks for continuous blocks of free space to dump the transaction
> groups goes into "Robocop mode". There may be a fix around the corner
> for this sometime soon.
> 
> Luckily ZFS pools can be expanded on the fly which means, when things
> do start to slow down, new disks can be added to the chassis
> alleviating the problem. There are also tricks to upgrade pools from
> smaller to larger disks whilst in production.

The threshold for this behavior has been changed to 96%, but this is
calculated at the metaslab level, rather than the pool level. If your
metaslabs are somewhat unbalanced, you will have relatively filled
metaslabs that exceed this threshold and relatively less filled
metaslabs that do not.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 901 bytes
Desc: OpenPGP digital signature
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20140118/098719e8/attachment.sig>


More information about the zfs-discuss mailing list