On 29/04/15 10:39, Sven Kieske wrote:
On 28/04/15 16:20, Dan Yasny wrote:
HA does not mean multiple running instances of the same service. It means if the service is gone, it will automatically be restored on a working server. That is a pretty narrow definition of HA, which is not shared by most parts of the community (and the world), leading to much confusion of users on this very ML.
HA in general means your service downtime gets minimized, today mostly realized through load balancing and clustering software services.
just restarting a service (in this case ovirt-engine vm) on a different host is not what todays user expect under the term "HA" imho.
in theory it should be possible as a design goal to make multiple ovirt-engine share one remote data base (remote database support is already there).
I think this would be a huge feature for ovirt, but also requires quite some design and coding to be done.
thus adding the devel list.
If ovirt-engine restarts this does not mean that you have a downtime in your service. The VMs and nodes already running will not be affected... So your infrastructure will still be operational while you restart the engine. G