http://starrocks.io logo
Join Slack
Powered by
Hi <@U0A71G31CDV> I want to deploy the starrocks ...
# questions-and-troubleshooting
s
Hi @Rocky I want to deploy the starrocks cluster in eks using Kubernetes operator and I want to update the password for the root user as part of initial deployment Also share me the latest stable version of the starrocks and docker tags
r
To deploy StarRocks on EKS using the StarRocks Kubernetes Operator and set the initial 
root
password, the recommended approach is using the kube-starrocks Helm chart. 1. Update Root Password during Initial Deployment The StarRocks Operator can automatically initialize the 
root
password using a separate initialization job. Step 1: Create a Kubernetes Secret First, create a secret in the same namespace where you will deploy StarRocks. The secret must contain a key named 
password
.
Copy code
bash
kubectl create secret generic starrocks-root-pass --from-literal=password='your_secure_password'
Step 2: Configure Helm Values In your 
values.yaml
(or via 
--set
flags), enable the initialization password feature and reference the secret:
Copy code
yaml
starrocks:
  initPassword:
    enabled: true
    passwordSecret: "starrocks-root-pass"
Note: During deployment, a pod named 
kube-starrocks-initpwd
will be created. It may initially show 
Error
or 
CrashLoopBackOff
while waiting for the FE nodes to become ready. This is expected; the job will eventually succeed once the cluster is up. ────────── 2. Latest Stable Versions and Docker Tags As of June 2026, the following versions are recommended for production: Release Type | Version | Docker Image Tags (Example) Stable (LTS) | 3.3.22 | 
starrocks/fe-ubuntu:3.3.22
, 
starrocks/be-ubuntu:3.3.22
Latest Stable | 3.5.15 | 
starrocks/fe-ubuntu:3.5.15
, 
starrocks/be-ubuntu:3.5.15
Latest Major | 4.0.8 | 
starrocks/fe-ubuntu:4.0.8
, 
starrocks/be-ubuntu:4.0.8
Official Docker Repositories: • FE: 
starrocks/fe-ubuntu
• BE: 
starrocks/be-ubuntu
• CN: 
starrocks/cn-ubuntu
• Operator: 
starrocks/operator
(Latest version is typically 
v1.9.x
or 
v1.11.x
depending on your Helm chart version). Quick Deployment Command To deploy using the latest stable Helm chart:
Copy code
bash
helm repo add starrocks <https://starrocks.github.io/starrocks-kubernetes-operator>
helm repo update
helm install starrocks-cluster starrocks/kube-starrocks -f my-values.yaml
Caution: If you are using a shared-data architecture (Cloud-Native), ensure your FE/BE images match and the 
run_mode
is correctly configured in your spec. References • quick_start/helm.md • loading/Spark-connector-starrocks.md • loading/Kafka-connector-starrocks.md • deployment/post_deployment_setup.md
s
@Rocky I have used the init password still my pod is in crashloopbackoff state
r
If your 
kube-starrocks-initpwd
 pod is in a 
CrashLoopBackOff
state, this is often expected behavior during the initial minutes of deployment. 1. Why it crashes (Expected Behavior) The initialization pod starts immediately after you run 
helm install
. However, it cannot set the password until the FrontEnd (FE) nodes are fully initialized and the MySQL port (9030) is ready to accept connections. * The behavior: The pod tries to connect, fails because the FE is still starting, crashes, and Kubernetes restarts it with an exponential backoff. * The solution: You should ignore this for the first 5–10 minutes. Once the FE pods show 
1/1 READY
, the next retry of the 
initpwd
pod will succeed and its status will change to 
Completed
. 2. Troubleshooting (If it persists) If the FE pods are already healthy (
READY 1/1
) and the 
initpwd
pod is still crashing, check the following: * Check the Logs: This is the most direct way to see why it's failing.
Copy code
bash
    kubectl logs <kube-starrocks-initpwd-pod-name>
* Verify the Secret Key: The StarRocks Operator specifically looks for a key named 
password
 inside your secret. If you named the key 
root-password
or something else, it will fail.
Copy code
bash
    # Verify the key exists
    kubectl get secret starrocks-root-pass -o jsonpath='{.data}'
* Check the FE Service: The init job connects via the internal FE service. Ensure the service exists and the selector matches your FE pods.
Copy code
bash
    kubectl get svc
* Resource Issues: On EKS, if your FE pods are also crashing, it is likely due to Insufficient Memory (FEs need at least 8GB to be stable) or PersistentVolume (PV) binding issues. Check 
kubectl describe pod <fe-pod-name>
. ────────── 3. Latest Stable Versions and Docker Tags (June 2026) For a production deployment on EKS, use these verified stable versions: Release Type | StarRocks Version | Docker Image Tag Stable (LTS) | 3.3.22 | 
starrocks/fe-ubuntu:3.3.22
Latest Stable | 3.5.15 | 
starrocks/fe-ubuntu:3.5.15
Current Major | 4.0.10 | 
starrocks/fe-ubuntu:4.0.10
Recommended Helm Chart & Operator: * Helm Chart: 
kube-starrocks
version 1.11.4 * Operator Image: 
starrocks/operator:v1.9.1
(or latest 
v1.11.x
) Example Docker Tags for your `values.yaml`:
Copy code
yaml
starrocksFeSpec:
  image: "starrocks/fe-ubuntu:3.5.15"
starrocksBeSpec:
  image: "starrocks/be-ubuntu:3.5.15"
Note for EKS: Ensure your BE nodes have enough 
max_map_count
configured on the underlying EC2 hosts (at least 
262144
), or the BE pods will crash immediately. References • quick_start/helm.md • deployment/helm.md • deployment/post_deployment_setup.md • deployment/sr_operator.md • administration/user_privs/authentication/native_authentication.md
s
@Rocky its been 30 mins my pod still in crashloopbackoff
r
If your 
kube-starrocks-initpwd
 pod has been in 
CrashLoopBackOff
for 30 minutes, it indicates a persistent issue rather than the normal startup delay. 1. Troubleshooting the CrashLoopBackOff Please run the following command to see the exact error message:
Copy code
bash
kubectl logs <kube-starrocks-initpwd-pod-name>
Common Root Causes * FE Pods are not "Ready": Check if your FE pods show 
1/1 READY
. If the FEs are also crashing or stuck, the init job cannot connect. On EKS, this is often due to Insufficient CPU/Memory (FE requires at least 4-8GB) or Storage (PVCs not binding). * Password Already Set: If this is not a truly fresh deployment (e.g., you deleted and re-installed the Helm chart but kept the Persistent Volumes), the root password might already be set. The 
initpwd
job tries to log in with no password to set the new one; if it's already set, it will fail with 
Access denied
. * Incorrect Secret Key: The job specifically looks for a key named 
password
. * Fix: Ensure your secret was created as: 
kubectl create secret generic starrocks-root-pass --from-literal=password='your_pass'
* Network/DNS Issues: In some EKS VPC configurations, the pod may fail to resolve the FE service name (
kube-starrocks-fe-service
). The logs will show 
Unknown MySQL server host
. ────────── 2. Latest Stable Versions (June 2026) For a stable production environment on EKS, these are the current recommended versions: Tier | Version | Docker Image Tags Stable (LTS) | 3.3.22 | 
starrocks/fe-ubuntu:3.3.22
, 
starrocks/be-ubuntu:3.3.22
Latest Stable | 3.5.15 | 
starrocks/fe-ubuntu:3.5.15
, 
starrocks/be-ubuntu:3.5.15
Latest Major | 4.0.10 | 
starrocks/fe-ubuntu:4.0.10
, 
starrocks/be-ubuntu:4.0.10
Kubernetes Operator: * Chart Version: 
kube-starrocks-1.11.4
(or latest available) * Operator Image: 
starrocks/operator:v1.9.1
────────── 3. Recommended 
values.yaml
for EKS Ensure your resource requests are high enough for EKS nodes to schedule them properly:
Copy code
yaml
starrocks:
  initPassword:
    enabled: true
    passwordSecret: "starrocks-root-pass"

  starrocksFESpec:
    image: "starrocks/fe-ubuntu:3.5.15"
    resources:
      requests:
        cpu: 2
        memory: 8Gi # Minimum recommended for FE
    storageSpec:
      name: fe-meta
      storageSize: 20Gi

  starrocksBeSpec:
    image: "starrocks/be-ubuntu:3.5.15"
    resources:
      requests:
        cpu: 4
        memory: 16Gi
    storageSpec:
      name: be-data
      storageSize: 100Gi
Note for BE on EKS: If the BE pods are also crashing, ensure you have set 
sysctl -w vm.max_map_count=262144
on your EKS worker nodes (usually via a Launch Template or UserData script). References • quick_start/helm.md • deployment/helm.md • deployment/post_deployment_setup.md
2 Views
Open in Slack
PreviousNext
Powered by