This page looks best with JavaScript enabled

Azure Kubernetes Service - Issues have met and the solution

 ·  β˜• 6 min read  ·  πŸ‘½ Nhat Le

[Mongodb] Restore data in Mongodb:

mongodb restore:


  • Exec to pod: kubectl exec -it <POD_NAME> bash
  • Then u are inside the pod and run: mongo -u -p
  • Next, run: rs.initiate()

[Azure Resource] 1 Insufficient memory, 1 node(s) exceed max volume count.

When deploying resource infrastructure The reason is the limit of the disks can be attached to the node. For example, the size of node_1 is standard_b2s, which can only attach a maximum of 4 disks to the node. So when u deploy a new infra resource to node_1 which has already had 4 disks attach to. Then the error will show 1 node(s) exceeds max volume count

Check the capacity of the disk can attach to the node:

disk attach:


  • Go Virtual machine scale set on Azure
  • Resize the VMs (The size have more Data disks) or increase the number of node

[Release] “<release_name>” has no deployed releases


  • Just need to uninstall that chart with the command “helm uninstall <chart_name>” and release again (Even though when using helm list, the chart not show)
  • Maybe there is a mismatch between Kubernetes and Helm. when deploying the chart but you immediately interrupted it

[Helm chart] Install another version MongoDB chart

The scenario is you are using version 4 but install version 5 or any version different from the first


  • It can not work as expected. Have to uninstall MongoDB chart, delete PVC. (back up the database before do this). Then create a new PVC and MongoDB with the new version

[Elasticsearch] Unable to start pod after restoring data (Elasticsearch pod not green)

There are more errors show on logs, but the root core is java.lang.OutOfMemoryError: Java heap space


  • Increate JVM heap to 512, or more (as the image below) in the configuration file, if you use helm3, the config in elasticsearch-values.yaml
Java heap space
Java heap space:

[Nginx-Ingress] Can not upload file (limit size upload)


  • Add 6m to the values.yaml file (this will create an annotation to service’s ingress)

for example, add this line to candidate-app-service:

increase body size nginx
increase body size nginx:

[Helm] [Local machine] The system cannot find the file specified stable-index.yaml

Somehow the file stabel-index.yaml has been deleted, you can check at path, for example on window:
C:\Users\xxx\AppData\Local\Temp\helm\repository (the Helm3 path you have install in your machine)

Solution (run on your machine):

  • Run: helm repo add stable
  • This script add repo “stable” to the path

[Azure DevOps Pipeline] helm task deploy timeout

Somehow there is a problem with Helm3 task deploy option –wait, it not actually waiting to deploy the chart successfully.

Besides that, If we implement a health Probe for pods, a pod such as A-client take time to let the “health check” to green (more than ~10 minutes). this also makes the “helm3 task” wait ~10 minutes, this led to a timeout error. We can increase the timeout in a “task”, but seem like it not work as expected


  • To workaround, uncheck option –wait on helm deploy task. Then we have to go on Kubernetes dashboard to manual check there is any problem with the pod or use scrip (Kubernetes get pod -w)

NOTE: Will update scenarior and solution later

[Certificate] The certificate is expired (Logs from a service, ex: Identity service)

Identity service certificate
Identity service certificate:


  • The certificate file in Identity pod no longer valid so you need update that file
  • The way to put the file to a pod is using “secret” or “configmap”. Basically the idea is create a secret file, then “mount” that secret to the path inside the pod. Take a look at this in values.yaml of accounts chart:
  - name: cert
    mountPath: /tmp/cert
    readOnly: true
  - name: cert
      secretName: identityserver-certificate


  • Remove the old certificate: kubectl get secret and find the secret name “identityserver-certificate” or any your secret name and delete
  • cd to where you place the certificate on your machine,
  • (change certificate.pfx to the file name you have) next run kubectl create secret generic identityserver-certificate --from-file=./certificate.pfx.
  • Next, update the password in the Identity service (if your cert have password)

[Certificate] Unable to retrieve document from: xxxx/.well-known/openid configuration

The cert is still valid but maybe it lacks Chain Cert(CA).

Error from Identity service
Error from Identity service:


  • Check that the certificate is still valid with or

  • The result shows that lacking “Chain Cert” (or rank B). The result get from sslshopper (image below):

    Check CA domain
    Check CA domain:
  • Use to generate the Chain Cert, give it our domain and it gives us the Chain Cert.
    Look at the file you has already downloaded (OPEN WITH NOTEPAD), copy that string (only CA string) and add
    to your tls.crt (in k8s secret) (ex: tls.cert in bravotalents-tls-certificate)

[Logging configuration] Cannot scrape logs from some services

  • Make sure that all service has already saved logs on the node.

  • You can check by creating a pod, then mount the log’s path on the node to that pod.

  • Then you can exec to the pod has created, and check is there already have all services logs
    Example a pod (take a look at hostPath and mountPath):

    pod volume
    pod volume:
  • If everything is ok, so the problem maybe comes from the Filebeat service or the format of logs that Filebeat doesn’t understand

    • Go to setting of service such as appsetting.json (.NET) and remove this line
      "UseElasticsearchLogFormatter": false or set to true
    • Check that the setting doesn’t change the output of logs or turn off the format.

[Application] Service cannot serve traffic

The scenario is when exec to the pod, run command curl localhost:80 it doesn’t response anything. event run command netstat -apnt still return the service listen on port 80

NOTE: Will update scenarior and solution later

Share on

Nhat Le
Nhat Le
DevOps Engineer (Junior)