On Fri, May 30, 2008 at 9:49 PM, Josh Bowers <jbowers@xxxxxxxxxx> wrote: > I am trying to setup AFR across two nodes with the clients on the same boxes > as the servers. Currently, I am using "AFR single process" to do this and > everything seems to work. > > However, I think I'm missing something because the performance seems to be > poor. > > I was under the impression that "option read-subvolume" option would read > from the local filesystem if the "type storage/posix" was specified for the > subvolume. It seems to have no affect whether or not this option is used. What numbers are you getting with and without option read-subvolume? It is also possible that when the option is not given, the file might be getting read from the local subvol itself hence you would see similar numbers. When you do not specify option read-subvolume, a file gets read from one of the subvols depending on the inode number. Krishna > > > # glusterfs -V > glusterfs 1.3.9 built on May 28 2008 11:01:17 > Repository revision: glusterfs--mainline--2.5--patch-770 > > # uname -a > Linux webnode1 2.6.24-gentoo-r7 #1 SMP Tue May 6 12:38:44 EDT 2008 x86_64 > Intel(R) Xeon(R) CPU X3210 @ 2.13GHz GenuineIntel GNU/Linux > > > > > #### NODE 1 #### > volume client1 > type storage/posix > option directory /data/gluster/acalog/ > end-volume > > volume client2 > type protocol/client > option transport-type tcp/client > option remote-host 10.42.0.31 > option remote-subvolume client2 > option transport-timeout 10 > end-volume > > volume webroot > type cluster/afr > option read-subvolume client1 > subvolumes client1 client2 > end-volume > > > volume ioc > type performance/io-cache > option cache-size 64MB > option page-size 1MB > option force-revalidate-timeout 30 > subvolumes webroot > end-volume > > > volume server > type protocol/server > option transport-type tcp/server > option client-volume-filename /etc/glusterfs/glusterfs-client.vol > subvolumes webroot client1 > option auth.ip.webroot.allow * > option auth.ip.client1.allow * > end-volume > > > #### NODE 2 #### > > volume client1 > type protocol/client > option transport-type tcp/client > option remote-host 10.42.0.30 > option remote-subvolume client1 > option transport-timeout 10 > end-volume > > volume client2 > type storage/posix > option directory /data/gluster/acalog/ > end-volume > > > volume webroot > type cluster/afr > option read-subvolume client2 > subvolumes client1 client2 > end-volume > > > volume ioc > type performance/io-cache > option cache-size 64MB > option page-size 1MB > option force-revalidate-timeout 30 > option force-atime-update off > subvolumes webroot > end-volume > > > > > volume server > type protocol/server > option transport-type tcp/server > option client-volume-filename /etc/glusterfs/glusterfs-client.vol > subvolumes webroot client2 > option auth.ip.webroot.allow * > option auth.ip.client2.allow * > end-volume > > > > > > -- > Josh Bowers > System Administrator - Digital Architecture > http://www.digarc.com/ > > > _______________________________________________ > Gluster-devel mailing list > Gluster-devel@xxxxxxxxxx > http://lists.nongnu.org/mailman/listinfo/gluster-devel >
