The Graylog document says only one of the Graylog cluster nodes is master. I am assuming this is active/standby HA. So, do anyone know that if you can simply use Linux Heartbeat HA instead of a proxy/load-balancer solution for a multi-node Graylog deployment?
Could you please provide more details?
From my experience heartbeat is used just for nodes monitoring (so standby node can know that active node still alive), not for load balancing
Thanks zoulja. What I am using is from linux-ha.org (http://www.linux-ha.org/wiki/Download)
It creates a virtual IP address for the two Graylog nodes and the secondary take the virtual IP address if the primary is down.
I mean, if only the master Graylog node is processing the incoming messages, then I can use this instead of a full-fledged load balancer.
There is no load balancing in your scheme, just failover/redundancy.
Load balancing is needed when you have for example 2 GL nodes and specific Input is running on each of them.
If you have balancer then it can route requests to both nodes.
Without balancer all requests will go to the same node(where you have your floating IP), so one node can be overloaded and second node will do nothing.
So this assumption:
if only the master Graylog node is processing the incoming messages
is not really correct.
Message is processed by the node where Input is running. Master node is handling things related to Indexes, metadata and so on
Thanks. You clear my mind with the Graylog master/slave explanation.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.