HAProxy downtime until the IP is moved


#1

Hi,
I know that this is rather a keepalived question, but I am just wondering how you guys are dealing with this.

I have the two instances checking for haproxy process every second and after 3 failed attempts failing over the backup instance, which means 4 seconds of downtime … How are you guys dealing with this? What would be the configuration for the ALOHA appliance for example?

Thanks.


#2

You accept it. Seriously, the failover in itself is probably doing more harm (given that existing sessions will break) than the 4 seconds of actual outage of the VIP.

When does it really failover? When you have a hardware fault, and when haproxy crashes. If both issues aren’t extremely rare in your configuration, you are probably doing something wrong.


#3

You are right, with my current infrastructure I am having outages once a year ,maybe… But I was thinking to fail over quickly to apply some maintenance on the LB hosts, for example.
But I was actually asking this to know if this behavior was something expected from LB.

Thanks.


#4

For maintenance I’d assume keepalived has some other way to do it, without waiting the full 4 seconds. That said, I’m not a keepalived expert.