[Gluster-users] Write performance in a replicated/distributed setup with KVM?

[Brian Candler]

On Fri, Mar 02, 2012 at 11:04:02AM +0200, Harald Hannelius wrote:
> This has probably been discussed before, but since I'm new on the
> list I hope You have patience with me.

There are some tuning hints in this post:
http://gluster.org/pipermail/gluster-users/2012-February/009584.html
although mostly focussed on read performance.

I don't know why your write performance is so slow. But to narrow it down, I
suggest the following:

- Create a single-node volume (non-replicated, non-distributed)
  a. Time writing to it via a glusterfs mount on the same node
  b. Time writing to it via a glusterfs mount from a different node

If both are fast: then retest using a two-node replicated volume.

If a is fast but b is slow: investigate your gigabit network

If both are slow: try writing to your RAID1 volume locally

> # ls -lah /gluster/debtest.img
> -rw------- 1 root root 20G Mar  1 12:35 /gluster/debtest.img
> # du -ah /gluster/debtest.img
> 4.5G	/gluster/debtest.img

A sparse file, right? So another useful test would be a 4.5G regular file.

Regards,

Brian.