[Gluster-users] Gluster lvm cache vs tiering performance issue

[Mohammed Ghanawi]

I am on the latest gluster version 6.x. I have an nvme lvm cache,
configured on a sata based volume. I have two bricks replicated, and
configured exactly the same way.

I have made sure that the block size on volumes and xfs is 128k. The issue
is that the write performance to the mounted gluster volume is horribly
slow 39 mib/s.


However if I test the lvm volume performance or the individual bricks
performance I get 600-700 mib/s, which confirms the bricks are
configured correctly. It is only when I attempt to mount the gluster file
system remotely or even locally I get slow speeds.

I had a similar config 2 nvme disks for the hot tier and 2 sata disks for
the cold tier, and the mounted gluster file system v3.12 was giving near
600 mib/s.

It's not the network, as mounting locally has the same issue. I tried write
through on the lvm volume, but no luck.

What could be the problem?

Thanks
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20190826/24f5dcdd/attachment.html>