[Gluster-devel] Gluster failover question

[Scott Whitney]

We're moving to an HA solution in my SaaS environment, and I'm looking at using Gluster 2.09 (32 bit hardware) and CentOS 5.5. 

This is up and running as such: 
gluster1 -- data server 
gluster2 -- data server 
app1 -- app server (mounts the glusterfs in /home) 
app2 -- app server (ditto) 

Here's my question. 

Test 1: 
create /home/foo and add 10 files on app1 
ls /home/foo on app2 -- I see them 
ls /data/export/foo on gluster1/gluster2 -- I see them 

Test2: 
rm -rf /home/foo on app2 
ls on the other 3. The directory is gone 

Test3: 
create /home/foo and add 5 files on app1 
shutdown gluster2 
add 5 more files 
startup gluster2 
ls /data/export/foo on app1/app2/gluster1 -- I see 10 files 
ls /data/export/foo on gluster2 -- I see only the 5 files created when the server was up. 

How is the failover/replication supposed to work in the situation that one of the backend RAID1 servers goes down? 

config files attached 

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: gluster1.txt
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20100520/33698342/attachment-0009.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: gluster2.txt
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20100520/33698342/attachment-0010.txt>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: client.txt
URL: <http://supercolony.gluster.org/pipermail/gluster-devel/attachments/20100520/33698342/attachment-0011.txt>