So, I'm looking at deploying an IPVS cluster. I'd like to deploy it in a
way where I have some level of HA. Fortunately, my application fits fairly
well into the source hash scheduler. From my understanding, I don't need to
run the ipvs synchronization daemon across nodes.
The only scenario that I'm not sure of, is what happens if a load balancer
fails over at the same time as a backend/real server failing, and therefore
getting its weight marked at 0? Does the source hash scheduler utilize
consistent hashing, or still include the weighted 0 server in the
calculation? Or do I still need to do IPVS synchronization in order to
allow for simultaneous load balancer, and backend failure?
Additionally, I've noticed that someone has released a consistent hashing
scheduler, here: https://github.com/dsheng/ip_vs_ch
Has anyone used this scheduler, and would you potentially roll this into
the main release?
Please read the documentation before posting - it's available at:
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users