Hello
I am going to deploy in field an Hosted Engine oVirt environment with just one Storage
(just one Target ID) and 6 Blades that access to the Storage by iSCSI.
Even I did not read the Warning about the requirement of a double target ID, one for the
LUN of the Engine and another one for the LUNs that host the Disks of the other VMs, and
change the Bill of Material now, it is a bit complicated !
After the reading of this thread I set up, in Lab, a test environment in which 2 Servers
run the oVirt Node 4.3.7 and they host the Engine in version 4.3.7, too
A third Server runs a CentOS version 7.x and it has been configured to export Disk space
by iSCSI (targetcli tool). On the STorage Emulator I have configured 3 LUNs behind the
same Target ID.
On the oVirt Environment I have 1 DC and 1 Cluster including the 2 Servers and 3 Storage
Domains corresponding to teh 3 LUNS.
The Cluster configured is the Default one, i.e. the one on which it runs the Hosted
Engine. On the same Cluster I have run 2 VMs with CentOS guest.
Now the Engine keeps its data on the first LUN, one of the guest OS keeps its data on the
second LUN and the second guest OS keeps its data on the third LUN.
Well, if I put in Maintenance one of the guest OS LUN, the other LUNs (the one of the
second guest OS and the one of the Engine) continues to be available.
I wonder if there is any other reason that requires the presence of a second Storage in
the network design, or if we con consider it just as a preferred design but not as
Mandatory.
Many thanks, Luca