HA and different datacenter + IPs, how that works?

[postcd]

Hello,

There are several articles presenting HA functions which moves VPS from node to node if source node becomes offline (if i understand properly)... but i do not understand how networking works in this HA scenario.

i have:
- server from company A in USA datacenter with one IPv4 subnet
- server from company B in EU datacenter with one IPv4 subnet
- server from company C in Asia ...

How this can work in HA mode when i need virtual servers have fixed IP, so when there is trouble with the node which has this IP assigned, VPS is always inaccessible? Is there a workaround, can you please describe how this HA works in terms of networking? I am a noob in networking, can you please write how things are named so i google it? And last one, please can you provide hyperlink/s to a tutorial/s that explains this networking setup for HA? I am talking about public IPs for the virtual servers, not just internal NAT setup. Thank You

 

[bofh]

short answer it doesnt for several reasons.
fuirst promox doc s says ha only with low latency, so its not ment ot be used for WAN failover.

second a multi site HA isnt something trivial.
it can be done if you have the same tier 1 carrier like level 3 that could provide you some failover IPs
that would be essential using BGB.

second option would be by DNS roundrobin or similar but thats not really what you want.
if its only for HA you could go the dynamic DNS route with a very low TTL that should be sufficent in most used cases.

BGB is i assume out of scope

so now, first things first.
do you want strictly HA or also load balancing?

second what kind of application do your VMs serve and how much data is involved.

i can tell you right now that HA on level of the VMs is out of the question.
however depending on the application there several otehr methods you could use.

for example if its only a website you could use cloudflare for the failover/distribution part and replicate data on application layer between the datacenters.

but simply HA those vms across the globe - sorry but nope, not without a good chance of data corruption (if even at all)

 

[aderumier]

if you manage your own network (routers/bgp), you can change bgp annonce to failover ips to another datacenter.

if you don't manage your network, for web, you can deploy reverse proxy on each datacenter and an external loadbalancer (cloudflare for example), then manage your vm ips in a private network with vxlan (across each dc)