Hi list,
I think I encounter the following problem:
If a backend hits it maxconn setting for a server, further requests including http-checks are queued. After some time the health checks time out and haproxy changes the servers state to DOWN.
The next health check succeeds an the server comes UP again. All in all this results in a flapping server. As a work around separated the health check into a separate backend without connection limit:
backend srv-health
option httpchkp
server srv srv.example.org check
backend srv
server srv srv.example.org track srv-health/srv maxconn 1000
1) Is there a proper way to keep health checks away from connection limits such that health checks are never queued?
2) Can I configure some kind of logging for hitting the connection limit?
thx, ~s
I think I encounter the following problem:
If a backend hits it maxconn setting for a server, further requests including http-checks are queued. After some time the health checks time out and haproxy changes the servers state to DOWN.
The next health check succeeds an the server comes UP again. All in all this results in a flapping server. As a work around separated the health check into a separate backend without connection limit:
backend srv-health
option httpchkp
server srv srv.example.org check
backend srv
server srv srv.example.org track srv-health/srv maxconn 1000
1) Is there a proper way to keep health checks away from connection limits such that health checks are never queued?
2) Can I configure some kind of logging for hitting the connection limit?
thx, ~s