[Gluster-users] GlusterFS CPU hit

[Tejas N. Bhise]

Freddy,

If there is high iowait, the easiest way is to add more disks. With huge RAM and fast CPUs, the typical bottleneck that I have seen on scale out systems are disks and network. 

How many disks are you striping you volume ( lun ) across per server ?

Regards,
Tejas.

----- Original Message -----
From: "Freddy Morgenstern" <fmorgenstern at icarnegie.com>
To: gluster-users at gluster.org
Sent: Monday, August 2, 2010 8:57:57 PM
Subject: [Gluster-users] GlusterFS CPU hit


Hi,

I am a GlusterFS newbee and I am experiencing the following problem: whenever I run the top command a process called glusterfsd is using over 15% of the CPU at all times and my %wa is usually over 15%.  I find the %wa particularly worrisome because it's my understanding that this means that the cpu has to wait a very long time each time it requests data to the hard drive.

Can you please tell me if the high CPU ussage is normal behavior of glusterFS, if the high %wa could be caused by glusterFS  and if so is there any way I can optimize/tune my server to eliminate the problem?


System Configuration:
- GlusterFS 2.0.8-1
- RAID 1 SATA disk arrays
- CentOS


I appreciate any help you can give me.


Thanks and Regards,
Freddy


_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://gluster.org/cgi-bin/mailman/listinfo/gluster-users