[Gluster-users] Replication and Distribution behavior

[Yi Ling]

hi,mike~~

it's right according to you configuration.

because you made pair1 and pair2 set to "distribution", the data/file you
scp to /gluster mount point will store in pair1 or pair2, which means that
there wouldn't be 4 copies of data in all 4 nodes ath the same time.

you could try to copy more files into /gluster. you would find that there
will be some files stored in pair1 (node1 and node2), and other files stored
in pair2.

try more translators~~~

Yi Ling

----------------------------------------------------------------------

Date: Wed, 26 Aug 2009 13:49:57 -0600

From: mike foster <mfosterm at gmail.com>

Subject: [Gluster-users] Replication and Distribution behavior

To: gluster-users at gluster.org

Message-ID:

       <ff7f05da0908261249h7052a3c1l6403804f1db8d76d at mail.gmail.com>

Content-Type: text/plain; charset="iso-8859-1"


I apologize if this has already been covered, but I couldn't find anything

close enough to this scenario in searching the archives.


I'm evaluating a 4 node cluster, with nodes 1 and 2 replicating, nodes 3 and

4 replicating and pair 1 (nodes 1 and 2) and pair 2 (nodes 3 and 4) set to

"distribution".


However, when I copy data to any node on the /gluster mount point from a 5th

machine using scp, all of the data shows up in the exported share on nodes 1

and 2 only. The data does not get replicated to nodes 3 and 4, even when

directly connected to those servers.


Am I missing something or ...


Here's some configuration details:


cat /proc/mounts:


glusterfs#/etc/glusterfs/client.vol /gluster fuse

rw,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072 0 0


auto mounting the glusterfs in /etc/rc.local: glusterfs -f

/etc/glusterfs/client.vol /gluster


--- server.vol ---


# Gluster directory on raid volume /dev/md0

volume posix

 type storage/posix

 option directory /mnt/raid/gluster/export

end-volume


volume locks

 type features/locks

 subvolumes posix

end-volume


volume brick

 type performance/io-threads

 option thread-count 4

 subvolumes locks

end-volume


### Add network serving capability to above brick

volume server

 type protocol/server

 option transport-type ib-verbs

 option auth.addr.brick.allow *

 subvolumes brick

end-volume


--- end of server.vol ---


--- client.vol ---


volume cf01

 type protocol/client

 option transport-type ib-verbs

 option remote-host 10.185.17.11

 option remote-subvolume brick

end-volume


volume cf02

 type protocol/client

 option transport-type ib-verbs

 option remote-host 10.185.17.12

 option remote-subvolume brick

end-volume


volume cf03

 type protocol/client

 option transport-type ib-verbs

 option remote-host 10.185.17.13

 option remote-subvolume brick

end-volume


volume cf04

 type protocol/client

 option transport-type ib-verbs

 option remote-host 10.185.17.14

 option remote-subvolume brick

end-volume


# Replicate data across each servers in 2 pairs

volume pair01

 type cluster/replicate

 subvolumes cf01 cf02

end-volume


volume pair02

 type cluster/replicate

 subvolumes cf03 cf04

end-volume


# Distribute data across all pairs

volume bricks

 type cluster/distribute

 subvolumes pair01 pair02

end-volume


# For performance

volume writebehind

 type performance/write-behind

 option cache-size 4MB

 subvolumes bricks

end-volume


--- end of client.vol ---