We have a setup of three geographically distributed proxmox machines.
The hardware is cheap: celeron cpus, single 5400rpm drive, 8/16gb memory.
Each node runs a single vm with our service. The three servers run a copy of the service. We're satisfied with the speed of the service.
One server is always live and replicated to the other two hourly. They are connected through a vpn, that is sometimes as slow as 2mb/sec with a 2-5ms ping.
If one server dies (network or power; sometimes disk failures), we manually promote one of the other two servers to go live.
We'd like to make sure the our service (the single VM) has an automatic failover - clustered DB and clustered storage (we have a lot of data, that is rarely updated).
We're thinking of GlusterFS/Ceph and a clustered mysql/postgres.
After a lot of reading about Proxmox 4 Cluster we'd like to know how will it handle on slow hardware and slow connection (if the speed does not decrease from what we have now, that will be okay).
What happens if the first node of the cluster goes dark for 2010 hours? Will the rest continue working?
The hardware is cheap: celeron cpus, single 5400rpm drive, 8/16gb memory.
Each node runs a single vm with our service. The three servers run a copy of the service. We're satisfied with the speed of the service.
One server is always live and replicated to the other two hourly. They are connected through a vpn, that is sometimes as slow as 2mb/sec with a 2-5ms ping.
If one server dies (network or power; sometimes disk failures), we manually promote one of the other two servers to go live.
We'd like to make sure the our service (the single VM) has an automatic failover - clustered DB and clustered storage (we have a lot of data, that is rarely updated).
We're thinking of GlusterFS/Ceph and a clustered mysql/postgres.
After a lot of reading about Proxmox 4 Cluster we'd like to know how will it handle on slow hardware and slow connection (if the speed does not decrease from what we have now, that will be okay).
What happens if the first node of the cluster goes dark for 2010 hours? Will the rest continue working?