[Gluster-users] Gluster caching behaviour

Robert Hajime Lanning lanning at lanning.cc
Tue Dec 3 05:13:36 UTC 2013


On 12/02/13 06:21, Maik Kulbe wrote:
> Hi,
>
> I have a quick question on how Gluster leverages caching on the brick
> servers. My bricks currently have 4GB of RAM and of that 4GB about
> 600-800MB per server are unused and even more gets freed from time to time.
> Is it possible to configure Gluster so it uses all the RAM on those
> machines?

On the server, it would not be a GlusterFS configuration.
It would be kernel tuning.  If you are running a linux kernel, then 
setting some sysctl values can help.

I have these tuned for my system:
vm.swappiness
vm.vfs_cache_pressure
vm.dirty_background_ratio
vm.dirty_ratio
vm.dirty_writeback_centisecs
vm.dirty_expire_centisecs

This pretty much goes for any storage server that is filesystem based 
(ext2/3/4,xfs,btrfs,zfs...)

Very important to tune these values when you have large amounts of RAM.

-- 
Mr. Flibble
King of the Potato People



More information about the Gluster-users mailing list