A server went down, but HAProxy kept it in UP state


#1

A server went down, but HAProxy 1.6.3 kept it in UP state even if check was set and LastChk was L6OK.
The only abnormal thing was check duration still = 0 for hours.
While I reloaded haproxy service it started to fluctuate from 0 to 30 ms.
Anybody saw such behaviour before?


#2

How do you know the server was down but kept UP?

you might want to upgrade to latest code in 1.6 branch (1.6.13) and retry
your test.

Baptiste


#3

I’ve got detailed monitoring. Upgrading seems good idea, but the problem appeared after some month of regular traffic, so it could be difficult to replicate.
Thanks.

Witold Grzebinski


#4

To answer your question from the first post: No, we did not see such a behavior.

Anything more specific would require that you actually provide some details about this. Like your full configuration, and exactly what kind of down you had on the backend server.