Failover configuration in wso2 esb - esb

Question

We have a proxy service that does the following
1. calls a backend service to get the client id
2. use the endpoint (client endpoint) configured for that client id and forward the original request to that endpoint.
3. return a response to the request
Earlier we were setting OUT_ONLY property to true while forwarding the request to the client endpoint (step 2).
But we now have to use a failover endpoint configuration for the client endpoint. So we removed the OUT_ONLY property and used
the following failover configuration. The client endpoint does not return any response for the forwarded requests.
Due to this the callback that gets registered does not get any response and it gets cleared by the Timeouthandler subsequently.
Recently we are facing some CPU spikes and OutOfMemory errors getting thrown by ESB.
We suspect that the synapse callback handler is utilizing most of the system resources and unable to recover.
Could this configuration create any OutOfMemory errors (if stress tested) as there is no response returned by the client endpoint?
Please suggest.
<endpoint xmlns="http://ws.apache.org/ns/synapse" name="client_forward_endpoint_110">
<failover>
<endpoint name="primary">
<address uri="--primary endpoint--">
<enableAddressing />
<timeout>
<duration>30</duration>
<action>discard</action>
</timeout>
<markForSuspension>
<errorCodes>101504, 101505, 101500</errorCodes>
<retriesBeforeSuspension>3</retriesBeforeSuspension>
<retryDelay>1</retryDelay>
</markForSuspension>
<suspendOnFailure>
<initialDuration>1000</initialDuration>
<progressionFactor>2</progressionFactor>
<maximumDuration>64000</maximumDuration>
</suspendOnFailure>
</address>
</endpoint>
<endpoint name="secondary">
<address uri="--secondary endpoint--">
<enableAddressing />
<timeout>
<duration>30</duration>
<action>discard</action>
</timeout>
<markForSuspension>
<errorCodes>101504, 101505, 101500</errorCodes>
<retriesBeforeSuspension>3</retriesBeforeSuspension>
<retryDelay>1</retryDelay>
</markForSuspension>
<suspendOnFailure>
<initialDuration>1000</initialDuration>
<progressionFactor>2</progressionFactor>
<maximumDuration>64000</maximumDuration>
</suspendOnFailure>
</address>
</endpoint>
</failover>

Answer 1

If your endpoint doesn't return any response, how can you tell if it is active?
If there is any kind of response from the endpoint to any kind of request, you may be able to do a more complicated mediation using first a 'call' mediator to check endpoint availability, then a send mediator with the OUT_ONLY property. You may use a switch mediator for evaluating the result of the 'call' mediator, and set a very short timeout for that mediator to make the throughput acceptable even if the first endpoint is inaccessible. It would not be an ideal solution, but it should work.