HTTP load balancer and 0/0 of VMs in backend service

[Pavel Madr]

Hi,

my HTTP load balancer doesn't find any instance from group.

The backend service shows that my instance group is connected, but shows 0/0

"my-group  us-central1 0 / 0 Off Max CPU: 80% 100%"

Command "$ gcloud compute backend-services get-health" returns just:

---

backend: https://www.googleapis.com/compute/v1/projects/my-project/regions/us-central1/instanceGroups/my-group

status:

  kind: compute#backendServiceGroupHealth

It's strange because Network load balancer works fine with that instance group, VM, instance group and health check too. Firewall is enabled for health checks too.

Yesterday backend service reported 1/1 or 0/1 (in case that health check failed). Today with new deployment it shows 0/0.

Please, can someone help me?

With Best Regards,

Pavel

[Pavel Madr]

UPDATE: I did nothing and it started to work after 30 minutes.

It's really strange. Is the HTTP load balancer reliable?

[Faizan (Google Cloud Support)]

Hello Pavel,

Do you have autoscaling[1] enabled for your instance group? If yes, it is an expected behavior, due to the fact that when instance in your instance groups becomes unhealthy it will be removed from the group and recreated. During the time of recreation or deletion they are not visible to the backend services in the load balancer and it will report 0/0 instances. This only will happened when all the instances in the group are deleted or there is only one instance in the the group.

I hope it helps.

Faizan

[1] https://cloud.google.com/compute/docs/instance-groups/

[Pavel Madr]

Hello Faizan,

I have not autoscaling enabled. I was tracking health check requests on VM instance and only instance group health check was connecting.

VM instance was working well for whole time without restart. I suspected firewall rules, so I enabled access for all IPs. It didn't help too.

Today everything works fine. Maybe there was some network issue.

Best Regards,

Pavel

[Faizan (Google Cloud Support)]

Hello Pavel,

I believe you are using autohealing with the instance group. Autohealing will have the same behavior that I have explained earlier because the non healthy VMs gets deleted and recreated.

[Hrishikesh]

Hi Faizan,

I realize this is an old thread, but I wanted to get this point clarified here since it's related to something I am facing with my GCP deployment.

You mention that "when instance in your instance groups becomes unhealthy it will be removed from the group and recreated" is a property of autoscaling. However, as far as I can figure out from the docs, this is a property of autohealing, not autoscaling. Can you confirm that this was a typo on your part, or if I am missing something here?

Regards
Hrish