goodmorning,i am evaluating (and trying to wrap my head around) the pros and cons of having a generic application cluster (say db, web server, whatever) with one node in site1 and the second in site2, interally connected to allow replication of local data/san, and on the wan side each site is announcing a /24 with bgp voodoo to accepting external request.
in case the l2 is going down all the two nodes will be in split brain? is this an application/middleware problem more than a netwoking possible solution?
I mean it is better to think about it as two monolithic site in each one with its own midlleware and try to sync then in some other way? for example, aws i think suggest to make the application site-fault resilient deploy two different single node and then load balance them with dns?
I am saying somethink terrible wrong? how the big boys do this stuff?
thank you!
No comments:
Post a Comment