HA HTTP client behavior

Hi All,

I am trying to understand how HA would work with the HTTP client API. The documentation states that:

you would just put a load balancer in front of the three nodes

But how the load balancer would know one of the three node is down or zombie? In that case, one out of 3 client requests will fail consistently until the node is back up. This would increase latency because of the client retries involved.

Is there a better way?



Most load balancers can detect downed nodes and stop routing to them

Is there load balancing (e.g. indeed.com boxcar style) anticipated in the TCP client?

Eg for reads? This would be pretty easy to setup now

Exactly, mostly for read load balancing.