Licode falls after not using it in a certain time

Question

I'm working with the webRTC Licode and I've noticed that, sometimes (it does not always happen) after about two days approximately without using Licode it falls and stops working. I get the following error.

Image here

Does anyone know how could I solve it?

Answer 1

Licode doesn't fall after not using it, it has a time synchronizing issue. Since it's a cloud based application with a microservice architecture, all services use the CloudHandler as a registry/discovery service. Services send keep-alive messages to the cloud handler through the rabbitmq and the cloud handler checks them through a check-keep-alive messages. The time between both events isn't correctly defined, neither the timeout between them.

When Licode starts, Erizo Controllers events start to get too close (due to computing time) that the check-keep-alive event occurs before the keep-alive, so the cloud handler just removes the erizo controller from its database which makes the whole application fall.

Keep in mind that this is not totally true because the cloud handler makes n checks (n is defined in the licode_config.js) before removing the controller, but the time exceeds n times the interval time

A way to improve this is -Increase the check keep alive interval -Increase the timeout between the check-keep-alive and the keep-alive events

You will find these parameters in the cloudHandler.js and licode_config.js

Hope it helps