I have the following setup with GlusterFS.
Server: 4
- CPU: Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz
- RAM: 32G
- HDD: 1T, 7200 RPM (x 10)
- Network card: 1G x 4 (bonding)
OS: Centos 6.4
- File system: XFS
> Disk /dev/sda: 1997.1 GB, 1997149306880 bytes
> 255 heads, 63 sectors/track, 242806 cylinders
> Units = cylinders of 16065 * 512 = 8225280 bytes
> Sector size (logical/physical): 512 bytes / 512 bytes
> I/O size (minimum/optimal): 512 bytes / 512 bytes
> Disk identifier: 0x000efb6f
- GlusterFS: 3.4.0-8.el6
- Sysctl.conf:
> vm.swappiness = 0
> vm.vfs_cache_pressure = 1000
> net.core.rmem_max = 4096000
> net.core.wmem_max = 4096000
> net.ipv4.neigh.default.gc_thresh2 = 2048
> net.ipv4.neigh.default.gc_thresh3 = 4096
> vm.dirty_background_ratio = 1
> vm.dirty_ratio = 16
I use only default config for GlusterFS (follow
http://gluster.org/community/documentation/index.php/Getting_started_overview).
After testing between NFS client and FUSE client, I choose NFS because
the performance is much better.
NFS mount options:
svr385-1212.localdomain:/gv0 on /glusterfs type nfs (rw,nfsvers=3,addr=192.168.12.12)
I also use a Nagios plugin to monitor the availability the NFS mount
point by run `stat -f /glusterfs` every minute.
The problem: Sometimes, the plugin report that 'NFS server is not
responding', because `stat -f` do not return after 2 seconds.
Nothing useful in client machine log files.
The /var/log/glusterfs/nfs.log (http://pastebin.com/M2sicYBx) contains
some errors that I do not know how to interpreter.
So I want to ask:
1. Why is this happening to me? How can I debug the problem?
2. What are recommend settings for my cluster? Can I do something to
max out the performance?
--
PHAM Van Diep
