Skip to main content
CoreWeave applies request size limits at the platform gateway that all Kubernetes API traffic passes through: approximately 50 KB for combined URL and header size, and 100 MB for request body size. These limits aren’t adjustable per customer or per cluster. See Request size limits for the full limit table. Most failures come from the URL and header limit, not the body limit. If a request exceeds it, the behavior you see depends on your client: a connection reset or an error such as 431 Request Header Fields Too Large, 414 URI Too Long, or 400 Bad Request. This usually happens with CLI tools, job schedulers, or workload launchers that build kubectl exec or attach calls by embedding an entire script or command as a command-line argument. That argument becomes part of the request’s URL and headers, so a large inline script can push the combined request size past the gateway’s limit. To avoid hitting this limit when you automate Kubernetes workloads, follow these practices:
  • Keep any script or command you pass inline to kubectl exec (or an equivalent API-driven automation call) short.
  • For larger scripts, save the script to a file inside the container or a mounted volume. Then execute the file, for example bash /path/to/script.sh, instead of passing the full script inline.
  • If you use a workload orchestration or job-launching tool that inlines large payloads into Kubernetes API calls, check the tool’s documentation for a way to execute the payload from a file instead.
If the issue persists after these checks, contact support. Quotas & Rate Limits
Last modified on July 27, 2026