Hi
We have successfully deployed glusterfs. We have deployed two servers with one
volume as a replica. We use Citrix XenServer for our virtual environment. In
XenServer we have created a storage repository via NFS using the glusterfs
volume. This has been successful and XenServer can now use that storage
repository for storing virtual machine images.
We are finding that the storage performance for the virtual machine images being
stored across the network is too slow. I am comparing this to using a straight
Linux NFS server in the same scenario. glusterfs is very appealing as a
replacement for NFS due to its support for NFS clients and its redundancy.
Virtual machine images range from 10GB to 100GB, most are 10GB.
The two glusterfs servers are on servers with RAID 6 hardware and we get great
storage performance out of them.
Current volume configuration:
gluster volume info
Volume Name: storage-pool-01
Type: Replicate
Status: Started
Number of Bricks: 2
Transport-type: tcp
Bricks:
Brick1: scluster01.corp.assureprograms.com.au:/gluster-export/storage-pool-01
Brick2: scluster02.corp.assureprograms.com.au:/gluster-export/storage-pool-01
Options Reconfigured:
nfs.volume-access: read-write
auth.allow: 192.168.10.*
nfs.ports-insecure: on
nfs.addr-namelookup: on
server.allow-insecure: on
nfs.export-volumes: on
As you can see above we have applied some volume configuration changes like
"nfs.volume-access: read-write". Does anyone have any suggested
configurations changes that can be made to glusterfs to improve performance of
the I/O in this scenario?
Kind regards
Stewart Campbell
