[Linux-cluster] slow NFS performance on GFS2

[Olivier Desport]

Le 24/09/2013 11:42, Steven Whitehouse a écrit :
> Hi,
>
> On Tue, 2013-09-24 at 11:29 +0200, Olivier Desport wrote:
>> Hello,
>>
>> I've installed a two nodes GFS2 cluster on Debian 7. The nodes are
>> connected to the datas by iSCSI and multipathing with a 10 Gb/s link. I
>> can write a 1g file with dd at 500 Mbytes/s. I export with NFS (on a 10
>> Gb/s network) and I only can reach  220 Mbytes/s. I think that it's a
>> little bit far from 500 Mbytes/s...
>>
>> Do you how to tune my settings to increase the speed for NFS ?
>>
>> GFS2 mount :
>> /dev/vg-bigfiles/lv-bigfiles /export/bigfiles gfs2
>> _netdev,nodiratime,noatime 0 0
>>
>> NFS export :
>> /export/bigfiles
>> 172.16.0.0/16(fsid=2,rw,async,no_root_squash,no_subtree_check)
>>
>> mount on NFS clients :
>> nfs-server:/export/bigfiles /data/bigfiles nfs4
>> _netdev,rw,user,nodiratime,noatime,intr 0 0
>>
> One possibility is to try changing rsize and wsize at the client end. Is
> the 10G network used for NFS the same network as the one used for the
> iSCSI?
I've tried several values of rsize and wsize and it doesn't change 
anything.  iSCSI and NFS are not on the same VLAN. I'm testing with 
servers which are on the same chassis (NFS/GFS servers are on blades and 
client is a KVM guest with virtio network card). iSCSI and NFS use 
separates network in the chassis but each blade has to share iSCSI (50% 
to 100% of 10 Gbits/s) and LAN (50% to 100%).

iperf beetween server and client : 3.4 Gbits/s (435 Mbytes/s).



>
> I assume that the NFS is set up active/passive on the two GFS2 nodes, so
> only one node is exporting the GFS2 fs via NFS in the normal case?
>
> Steve.
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20130924/a80de961/attachment.htm>