[Gluster-users] Gluster questions - glusterfs 3.7.5

[Peter Michael Calum]

Hi,

I have tested from a client server mounted to the Gluster setup with both Glusterfs and NFS. I used fio for the tests.
NFS over tcp was about 30% faster than GlusterFs mount. My 4 bricks has 4x900G in raid5 each and 1G interfaces.
Fio test mounted with NFS shows about 1050 iops on 50% read/write. I don’t know if this is OK, but I think it is because of
the small number of disks in the raid5. I get bigger servers in 2-3 weeks with 14 disks for the bricks.

thanks,
Peter


Fra: gluster-users-bounces at gluster.org [mailto:gluster-users-bounces at gluster.org] På vegne af Thing
Sendt: 8. november 2015 21:07
Cc: gluster-users at gluster.org
Emne: Re: [Gluster-users] Gluster questions - glusterfs 3.7.5

I am in the exact same boat.  Well the front end is RHEL7.1 and the backend Ubuntu 14.04 LTS.
What I have done however is mount using glusterfs.
No idea which is faster, currently I am testing local disk v nfs v gluster to get a feel on the performance penalty of gluster.
NFS using UDP might well be a good idea though if in reality it is indeed more fault tolerant?

Yes I need an operational guide as well, if nothing is about I am just going to have to write myself one.


On 8 November 2015 at 22:41, Peter Michael Calum <pemca at tdc.dk<mailto:pemca at tdc.dk>> wrote:
Correction : Version are glusterfs 3.7.5

Fra: gluster-users-bounces at gluster.org<mailto:gluster-users-bounces at gluster.org> [mailto:gluster-users-bounces at gluster.org<mailto:gluster-users-bounces at gluster.org>] På vegne af Peter Michael Calum
Sendt: 7. november 2015 14:11
Til: gluster-users at gluster.org<mailto:gluster-users at gluster.org>
Emne: [Gluster-users] Gluster questions

Hi group,

I’m new to Gluster, and still have a lot to learn..

I’ve build a 4 node setup for testing with Gluster 3.5.7  on CentOS 7.1.1503 (Core) an a replica 2 distributed volume
optimized for VM’s. – I have a RHEV 3.5 (Redhat Enterprise Virtuliazation) setup connected to my volume with NFS
as a storage domain, and I have 3 test VM’s running.

I have 3 questions :

1 : If I reboot one of the cluster nodes it won’t start the gluster up correctly. Gluster is started but RPC does dot run.
I have tried to enable it with systemctl but it don’t work. I have to start RPC manually and restart glusterd before the
local NFS service in glusterd starts. – Does anyone have seen this problem ?

2 : I have tried to connect from RHEV 3.5 and create a storage domain with the GLUSTERFS protocol, but this is not possible,
I get an error (Not possible to mount storage domain) In RHEV manager. I made a tcpdump and I can see that some version check
is performed. Am I right it it could be the RHEV3.5 is build for Red Hat Gluster version and therefore has an earlier client version
for gluster than 3.5.7 ? - Or are there backwards compability between clients and server ? - I think this could give problems in
the future when upgrading RHEV or Gluster.. Maybe Ovirt is better to use with gluster. – What do you think ?

3 : Does anyone knows a good maintenance guide for gluster to handle daily maintenance tasks, like taking a node out of service,
handling common errors, rebalancing and that kind of things..

Thanks for reading,

kind regards
Peter Calum



_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org<mailto:Gluster-users at gluster.org>
http://www.gluster.org/mailman/listinfo/gluster-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20151108/74827a07/attachment.html>