Error In Input - Connection Reset By Peer

I am seeing [netty-transport-#] Error in Input messages frequently in the graylog log. I believe the reason for this is due to the way the haproxy load-balancer sitting in front of the cluster nodes is performing health checks. I have altered the haproxy config to reduce the intensity of checks, but I don’t want to disable them completely.

Wondering if there is a log4j.xml config change I can make to filter out these messages?

did you use the endpoint in graylog that is present for loadbalancer health checks?

haproxy is load-balancing the tcp log inputs, and is checking against those listening ports on the cluster (3 nodes) eg. tcp/5140.

What I believe is happening is that the check itself is a:
SYN -->
SYN/ACK <–
RST -->

As the error message itself is “connection reset by peer” and haproxy is likely satisfied without completing the handshake.

I mean, Graylog has the status endpoint that can be used by loadbalancers to check the status. Special also if that Node is overloaded or not present. Might be better than just a tcp check for the service?

https://docs.graylog.org/en/3.1/pages/configuration/load_balancers.html#load-balancer-state

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.