Zero: error logs about broken pipe

Users Dgraph
1

Hey,

I have deployed a simple Dgraph cluster via helm chart v24.0.5 on a three node self hosted Kubernetes cluster.

On a totally new installation every zero pod logs every 10 seconds an error message like this:

E1223 12:11:23.512475       1 x.go:352] Could not send error msg=write tcp 10.111.54.190:6080->89.XXX.XXX.115:49574: write: broken pipe code=ErrorNoData due to http error write tcp 10.111.54.190:6080->89.XXX.XXX.115:49574: write: broken pipe

Has anybody an idea what does this mean? (probably any liveness probe)
I am wondering about the message content, because the “->” implies to me that the connection goes from the private IP of the zero pod (outgoing on port 6080) to the public IP (incoming on a random port) of the host the zero pod is running. Well this should be the other way around, right?

Could this be caused because I don’t use any ingress (disabled in the values for the helm chart)?

Thanks for any hint to get rid of it.

2

Just to provide you with some information we figured out during the last weeks.

The mentioned error is triggered by the readiness probe on the path: /state
When we changed it to /health like for liveness probe all zero pods work like expected without this error message.

We have also found a pull request from year 2020 where the issue is already mentioned. However the path: /state is still set in the value.yaml file for helm charts provided on GitHub (which caused the mentioned issue in my original post).

Are there any reasons to do the readiness probe for zeros on the path: /state ?