I'm thinking of doing an SSD based hyperconverged setup (for 4.3), but am wondering
about certain design issues.
seems like the optimal number is 3 nodes for the glusterfs.
but.. I want 5 host nodes, not 3
and I want the main storage for VMs to be separate iSCSI NAS boxes.
Is it possible to have 3 nodes be the hyperconverged stuff.. but then add in 2
"regular" nodes, that dont store anything and arent part of the cluster?
is it required to be part of the gluster cluster, to also be part of the ovirt cluster, if
thats where the hosted-engine lives?
or can I just have the hosted engine be switchable between the 3 nodes, and the other 2 be
VM-only hosts?
Any recommendations here?
I dont what 5 way replication going on. Nor do I want to have to pay for large SSDs on all
my host nodes.
(Im planning to run them with the ovirt 3.4 node image)
--
Philip Brown| Sr. Linux System Administrator | Medata, Inc.
5 Peters Canyon Rd Suite 250
Irvine CA 92606
Office 714.918.1310| Fax 714.918.1325
pbrown(a)medata.com|
www.medata.com