Il 26 mag 2016 22:16, "David Gossage" <dgossage@xxxxxxxxxxxxxxxxxx> ha scritto:
> If I have those sort of uptime requirements what I typically do is have 2 clusters of vm's and storage so I can have primary/secondary on different clusters so they can be brought down and have the other up. Otherwise regardless of updates or not I run the risk of one failure bringing everything down.
Not al services can be split in multiple hosts, in example a huge VM managed by a customer.
A scale-out strorage should not have SPOF, issues could happen to all, but having to bring down everything for a simple software update in a storage made to be high available is non-sense.
If bring down everything is really needed to upgrade, gluster can not be considered highly available
Bring down single host or server is ok, what is not ok and is nonsense is bring down the whole infrastructure as stated on official docs.
doc is saying that no one can access the volumes, no clients, no servers, so, the whole infrastructure must be taken down
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx http://www.gluster.org/mailman/listinfo/gluster-users
