Re: Loadbalancer shows 0/0 for health checks

[Digil (Google Cloud Platform Support)]

I would suggest editing your post to avoid exposing your project ID. I checked your project and the instances look healthy now.

Once you setup the health checks it might take a couple of minutes for GCP to show your instance as healthy.

These help center articles [1][2] have details on how to setup the health checks for different types of LB. I always suggest paying special attention to the firewall rules that need to be set up.

There is also very useful information blog [3] regarding debugging health checks in LB

[1] https://cloud.google.com/compute/docs/load-balancing/health-checks#health_check_source_ips_and_firewall_rules

[2] https://cloud.google.com/compute/docs/load-balancing/health-checks#tcp_and_ssl_health_checks  

[3] https://cloudplatform.googleblog.com/2015/07/Debugging-Health-Checks-in-Load-Balancing-on-Google-Compute-Engine.html

On Wednesday, October 11, 2017 at 9:07:04 AM UTC-4, aakash khatri wrote:

I have created a loadbalancer for my projectid - theworlzen-9ba7f. 

I have assigned the loadbalancer to an instance group which has one instance but it still shows 0/0 for health checks.

Please advise. 

[Manu J]

I was involved in setting up this load balancer. A couple of additional details

1. It didn't work for around 20-30 minutes and then it automatically started working.

2. Before it started working, the health checks never reached the instance. The instance was available and accessible all through.

3. The same issue was experienced by someone else before - https://groups.google.com/forum/#!searchin/gce-discussion/0$2F0%7Csort:relevance/gce-discussion/tBMze5dfegs/IB1L5SOrBAAJ

4. If it was a health check issue, why does the loadbalancer show 0/0 and not 0/1 ? The instance group had  1 instance in it all through and this was reflected in the console UI in other sections. I'm guessing it's not a health check issue because health checks never reached the instance and then after 20-30 minutes health check requests started hitting the server without any changes from our part

[Digil (Google Cloud Platform Support)]

That's the intended behavior while setting up LB for the first time. Also make sure that you haven't setup "autoscalling" or "autohealing". Did you check the performance of LB after sending the traffic to the instances?.

[Manu J]

This was for unmanaged instances. So no autoscaling or autohealing. 

Intended behaviour?  Then why is the loadbalancer reported as ready if it's actually not? It's quite confusing because you have no feedback as to something went wrong

[Digil (Google Cloud Platform Support)]

I have created and tested the LB. But not able to recreate the issue. Health check had taken only a couple of minutes and the LB worked fine. As I mentioned in my first response, the LB after the initial setup will take some time to become fully functional. 

Did you test the performance of the LB?. Did you face any errors while testing?

[Manu J]

Once it started working, there was no problem.  A couple of minutes was understandable but this took 30 minutes or so.

Also why does it show 0/0 instead of 0/number of machines in pool ?

[Navi Aujla (Google Cloud Support)]

Can you confirm if there are firewall rules in place to allow the health check (HTTP(s) or TCP)? For more information on the IP ranges associated with the HTTP(s) and TCP health checks probes, refer to this public documentation link [1].

[1] https://cloud.google.com/compute/docs/load-balancing/health-checks#health_check_source_ips_and_firewall_rules

[Manu J]

Yes. The health checks automatically started working after 30 minutes or so without any changes from our part. During this time, the instances were up and running and we we were able to access it directly 

[Digil (Google Cloud Platform Support)]

I created the LB in different zones but unable to reproduce the issue. After creating the LB, the health checks shows the correct number of instances. Have you tried creating LB in a different zone?