Pod that return 503 are not called

Hello,

I’m playing a bit around with Istio by using the Istio Tutorial from Red Hat. In the section Retry (https://redhat-developer-demos.github.io/istio-tutorial/istio-tutorial/1.1.x/5circuit-breaker.html#retry) it is shown Istio will not redirect requests to pods that replied previously with a HTTP status code 503. The test case contains no description why Istio is doing this and I can’t get my head around it. Could someone give me a hint, which feature of Istio could be used for this? What is also confusing for me that I Kiali is not showing any 503 error for the pod that returns 503 or at least this is how I interpret the result. Could somebody give me hint what the reason for this could be?

Thank you very much in advance,

Olaf

Hello,

I found the solution for the first part. Istio is setting default retry policy in the istio-proxy side car container. At the time being I have not found a documentation for this in the Istio documentation. So my original observation that the pods that return 503 are not call again is wrong. They are called again however the retry mechanism makes sure that the next pod for the service will be called which is working properly. I still don’t understand why Kiali is not showing 503 errors for the pod, but this is an other story.

Olaf