Dear all,
I would like to add a new brick to the running Gluster volume.
I have 3 bricks in 3 different servers. Everything is fine on the
running system. Connection to the application on the filesystem is ok.
All 3 server have a load between 0.5 and 1.5 on top. Iotop is idle and
iftop output is between 1 and 5 MB traffic.
Now I plan to add a new brick to the volume based on a LVM to create
snapshots.
It's easy to add this brick to the running gluster volume.
Shortly after successfully add of the brick the load in top explodes-
Self heal is running on the new brick and on all servers i have a load
of 40 and more.
The sync rate is very slow (150 MB in 5 minutes).
I have 3 VMs with 8 GB RAM, on the Gluster volume is 30 GB used and
every VM has 2 CPU core.
What's wrong with the Gluster cluster?
I use Debian Jessie with the available glusterfs-server package from the
default repositories (Glusterfs-server 3.5.2).
Is it possible to turn on/off the self heal for a brick?
Could it help to upgrade to a newer version of glusterfs-server (3.8.8
from Jessie-Backports or 3.10.8 from gluster.org)? Is it possible to
directly upgrade the glusterfs-server from 3.5.2 to 3.8.8 or 3.10.8?
Thanks
Ben
