Azure HA failover load balancer service issue

rmontalt Member Posts: 3

Greerings, I have an azure deployment where all 3 pods are up, redundancy, config-sync etc..., pod0, pod1 and pod2. We are using an azure load balancer service for clients to connect to smf ssl port. This works fine until pod0 goes down and pod1 is in charge. It looks like all the requests keep going to pod0 since it's pod label stays marked as active=true. I'm guessing that when the failover happens pod1 never gets it's label updated as active=true and stays as active=false. Thoughts, suggestions? Thanks.



  • Aaron
    Aaron Member, Administrator, Moderator, Employee Posts: 541 admin
    edited April 29 #2

    Is your load-balancer looking at the healthcheck endpoint on the brokers?

    Default port 5550 for non-TLS, 5553 for TLS:

    HEALTHCHK  TCP N N - Mgmt                   5550 U U
    HEALTHCHK  TCP Y N - Mgmt                   5553 U D No Cert

  • rmontalt
    rmontalt Member Posts: 3

    I'm using an azure load balancer service and from my understanding whenever pod0 changes status to backup and pod1 becomes active the readiness probe is supposed to update the pod label on pod1 to active=true so all client requests via load balancer should go to pod1 as it should get marked active=true and pod0 should get marked active=false. We see in a /tmp file that readiness probe is seeing that pod1 status as active but the pod label never gets updated which should happen via a curl command to the k8s api but I think that's the part that's not happening 🤔