Context

When deploying a LangGraph agent, you may encounter a timeout error stating "Queue Deployment is not ready after 600 seconds" even when using template agents. This issue can occur due to configuration problems or version compatibility issues with the LangSmith Platform Helm chart.

Answer

This timeout issue is commonly caused by two main factors that should be checked in order:

1. Check Your Helm Chart Version

The most common cause of this issue is using Helm chart version 0.11.14, which was yanked due to a bug in queue reconciliation logic. If you're running this version, upgrade immediately:

1. Check your current Helm version

2. Upgrade from version 0.11.14 to 0.11.15 or later

3. The upgrade will include listener RBAC fixes that resolve the deployment timeout

2. Verify N_JOBS_PER_WORKER Configuration

If upgrading doesn't resolve the issue, check your worker configuration. Look for this log entry:

N_JOBS_PER_WORKER is 0. Skipping queue.

If you see this message, the worker configuration is incorrect:

1. Set N_JOBS_PER_WORKER = "5" in your .env file

2. Place this setting after your model configuration (e.g., after LLMAAS_MODEL_NAME = "Meta-Llama-33-70B-Instruct")

3. Rebuild and redeploy your application

3. Additional Troubleshooting

If the issue persists after trying the above solutions:

1. Remove any custom auth configuration from your langgraph.json file temporarily

2. Delete all existing deployments and tracing projects before creating a new deployment

3. Collect pod logs using the troubleshooting script available in the LangChain documentation

The Helm chart version upgrade (from 0.11.14 to 0.11.15+) resolves this issue in most cases, as version 0.11.14 contained a known bug that prevented proper queue deployment.