0

I'm trying to research the best way to find a HA solution plus (little) load balancing for our TCP services. All solutions I found (e.g. HAProxy, LVS) use a master machine which holds the TCP service IP and when the master crashes a backup system (by keepalived, VRRP) becomes available by using that service IP. That's nice.

But I didn't found a solution for making the backup system the active one by a smooth transfer (without loosing a packet): Sometimes I need to update/change/reboot the current master machine so I want new connections going to the backup system while old/current connections are still handled by the old master until they are done so I am able to reboot the old master. Is that possible or can't this be done?

And while dreaming: The optimal solution would be a transfer of the current connections (plus the service IP) to the backup machine. Is there a solution for this dream?

MadHatter
  • 79,770
  • 20
  • 184
  • 232
Petra Verheim
  • 21
  • 2

1 Answers1

1

I'm not entirely sure what you're after, but if it's "smooth transfer without loosing a packet":

I've never used keepalived or VRRP implementations, because most (all?) of them lack STONITH, AFAIK.

But: The industry standard for setting up high availability clusters in the open source world, the corosync / pacemaker stack, provides this.

gxx
  • 5,591
  • 2
  • 22
  • 42