service/elasticsearch-master Type LoadBalancer stuck in <pending>

Hi,

When deploying with --set service.type=LoadBalancer, the service/elasticsearch-master is stuck in <pending> state for External-IP.

# helm install elasticsearch elastic/elasticsearch   --set service.type=LoadBalancer,antiAffinity="soft"
NAME: elasticsearch
LAST DEPLOYED: Fri May 29 10:29:44 2020
NAMESPACE: default
STATUS: deployed
REVISION: 1
NOTES:
1. Watch all cluster members come up.
  $ kubectl get pods --namespace=default -l app=elasticsearch-master -w
2. Test cluster health using Helm test.
  $ helm test elasticsearch --cleanup

# kubectl get all
NAME                         READY   STATUS    RESTARTS   AGE
pod/elasticsearch-master-0   1/1     Running   0          29m
pod/elasticsearch-master-1   1/1     Running   0          29m
pod/elasticsearch-master-2   1/1     Running   0          29m

NAME                                    TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)                         AGE
service/elasticsearch-master            LoadBalancer   10.43.169.130   <pending>     9200:30470/TCP,9300:30571/TCP   29m
service/elasticsearch-master-headless   ClusterIP      None            <none>        9200/TCP,9300/TCP               29m
service/kubernetes                      ClusterIP      10.43.0.1       <none>        443/TCP                         23h

NAME                                    READY   AGE
statefulset.apps/elasticsearch-master   3/3     29m

Elasticsearch version is 7.7.0 :

# helm list
NAME            NAMESPACE       REVISION        UPDATED                                 STATUS          CHART                   APP VERSION
elasticsearch   default         1               2020-05-29 10:29:44.676001708 -0400 EDT deployed        elasticsearch-7.7.0     7.7.0      

My setup is single node with Rancher:latest as a cluster.

# kubectl get nodes
NAME       STATUS   ROLES                      AGE   VERSION
xxxxxxxx   Ready    controlplane,etcd,worker   23h   v1.17.5

My repo is :

# helm repo list
NAME    URL                    
elastic https://helm.elastic.co

# helm version
version.BuildInfo{Version:"v3.2.1", GitCommit:"fe51cd1e31e6a202cba7dead9552a6d418ded79a", GitTreeState:"clean", GoVersion:"go1.13.10"}

Thank you.

[erouvin]

Thank you @jmlrt I gave it a shot with NodePort and the behavior is similar to LoadBalancer except that the deployment controls the port number to connect to from outside using the baremetal IP. But in the end, no more <pending> printout. I was using service.type wrong. NodePort fits better to my case.

[jmlrt]

Service type LoadBalancer is supported only on some cloud providers to allow provisioning an external Load Balancer in the cloud infrastructure (example provisioning ELB when using K8S on AWS). I supposed that this not something which is possible on K8S bare metal. If you only need to access your Elasticsearch cluster from inside K8S, you should use ClusterIP type. If you need to access it from resources outside K8S, you should use NodePort type (more details here).