Bug 761792 (GLUSTER-60)

[Migrated from savannah BTS] - bug 26282 [https://savannah.nongnu.org/bugs/?26282]

Tue 21 Apr 2009 11:03:20 PM GMT, original submission:

I ran into a problem with the latest rc release of gluster where read-subvolume only appears to work for only one machine. Here's my volume file:

volume posix
type storage/posix
option directory /tmp/gluster
end-volume

volume locks
type features/locks
subvolumes posix
end-volume

volume brick
type performance/io-threads
option thread-count 8
subvolumes locks
end-volume

volume server
type protocol/server
option transport-type tcp
option auth.addr.brick.allow *
subvolumes brick
end-volume

volume machine1
type protocol/client
option transport-type tcp
option remote-host machine1
option remote-subvolume brick
end-volume

volume machine2
type protocol/client
option transport-type tcp
option remote-host machine2
option remote-subvolume brick
end-volume

volume replicate
type cluster/replicate
option read-subvolume `hostname`
subvolumes machine1 machine2
end-volume

In this cluster I've got a 1gb file of random data that I've verified via md5sum is on both machines. On machine1, I can copy the file to > /dev/null while watching dstat, and see that the file is being read completely locally. If I do the same on machine2 though, the file is read from machine1. Only if I kill glusterfsd on machine1 then it'll read the local copy.

I've also found that If I change this line:

subvolumes machine1 machine2

to

subvolumes machine2 machine1

The behavior swaps. machine2 now will read the file locally and machine1 will read it remotely. Any idea what's going on?