Andy Knight
2009-Sep-11 14:41 UTC
[Gluster-users] (newbie) Basic Configuration Sanity Check
Hello I'm a newbie to glusterfs but have read all the docs and faq's but have been unable to find docs to answer my basic query. I'm looking to replace 2 localized NFS solutions which are kept (roughly) in sync via rsync with a simple glusterfs solution using AFR. The idea being to keep the localized NFS but have AFR keep the data in sync between locations. My setup is as follows hosts uk1 and us1 have the same server volfile and client volfiles (except for the directory name and auth.addr names) as per the 'AFR (Mirror) across Two Storage Servers' configuration example from the glusterfs docs. host uk1 is running glusterfsd (/usr/sbin/glusterfsd -f /etc/glusterfs/glusterfsd.vol) and host us1 is mounting (mount -t glusterfs /etc/glusterfs/glusterfs.vol /dir) the dir listed in the posix volume in the glusterfsd.vol file Everything appears to work as expected. Anything written to either host appears on the other host. I have exported (/etc/exports) the filesystems and NFS mounted them on other (local) hosts with writes from the NFS clients appearing on both uk1 and us1 My question is, is the above setup the 'best practice' for this solution ? or have I missed something along the way ? One thing I have noticed is that NFS writes to us1 (the glusterfs client?) are really slow - almost unusable - whereas NFS writes to uk1 (the glusterfs server?) are comparable with raw NFS Any help/guidance would be very much appreciated. glusterfs 2.0.6, fuse-2.7.4-1.el5.rf, hosts are all running CentOS 5.2, local networks for NFS is gbe, link between uk1 and us1 is tcp 100mb internet, and disks are all SATA Andy
Aaron Porter
2009-Nov-17 18:40 UTC
[Gluster-users] (newbie) Basic Configuration Sanity Check
On Fri, Sep 11, 2009 at 6:41 AM, Andy Knight <Andy.Knight at ivcapital.com> wrote:> glusterfs 2.0.6, fuse-2.7.4-1.el5.rf, hosts are all running CentOS 5.2, > local networks for NFS is gbe, link between uk1 and us1 is tcp 100mb > internet, and disks are all SATAYour nodes are separated by a transatlantic link? What's the RTT for pinging between the hosts? Gluster over WAN doesn't seem to work very well in our experience (East Coast/West Coast USA).