runfabric

Troubleshooting

Common errors and fixes per provider and workflow.

Credentials missing: General → “missing or empty …”
Trigger not supported: General → “provider X does not support trigger Y”
Provider-specific issues: jump to provider section

General

“missing or empty: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION”

Cause: Provider credentials are required for real deploy/invoke.
Fix: Set env vars (or use a profile). See CREDENTIALS.md and PROVIDER_SETUP.md.

“provider X does not support trigger Y”

Cause: The chosen provider does not support the trigger in the config (e.g. cron on fly-machines).
Fix: Change trigger in runfabric.yml or switch provider. See EXAMPLES_MATRIX.md for support matrix.

“functions must be an array to patch”

Cause: runfabric generate function patches the reference-format functions array in runfabric.yml.
Fix: Keep functions as an array of entries with name, entry, and triggers, or edit the config manually.

AWS (aws-lambda)

Missing env: Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION. Optional: AWS_SESSION_TOKEN for temporary creds.
IAM / access denied: Ensure the role/user has lambda:*, apigateway:*, sqs:*, etc. as needed. See PROVIDER_SETUP.md.
Region: Use a valid region (e.g. us-east-1). Wrong region can cause “resource not found” or timeout.

GCP (gcp-functions)

Missing env: Set GCP_PROJECT_ID and ensure Application Default Credentials (gcloud auth application-default login) or a service account key.
Wrong project: Confirm GCP_PROJECT_ID matches the project in the console. Check billing and APIs (Cloud Functions API) are enabled.

Azure (azure-functions)

Missing env: Set AZURE_SUBSCRIPTION_ID, AZURE_TENANT_ID, AZURE_CLIENT_ID, AZURE_CLIENT_SECRET (or use Azure CLI login).
Resource group / region: Ensure the app/function app exists in the expected resource group and region.

Kubernetes

kubeconfig: Set KUBECONFIG or use default ~/.kube/config. Ensure the context has namespace and RBAC for creating deployments/services.
Image pull: If using a private registry, ensure imagePullSecrets or cluster config is set.

Cloudflare (cloudflare-workers)

Missing env: Set CLOUDFLARE_API_TOKEN and CLOUDFLARE_ACCOUNT_ID.
Real deploy: Set RUNFABRIC_CLOUDFLARE_REAL_DEPLOY=1 for real API deploy. See PROVIDER_SETUP.md.
invoke logs has no live tail output: Install wrangler CLI so RunFabric can use wrangler tail. Without it, RunFabric falls back to Cloudflare API tail.
Force API tail only: Set RUNFABRIC_CLOUDFLARE_DISABLE_WRANGLER_TAIL=1.

Vercel / Netlify

Tokens: Set the provider-specific token (e.g. VERCEL_TOKEN, NETLIFY_AUTH_TOKEN). See PROVIDER_SETUP.md.
Project / site: Ensure the project or site ID is correct for the account.

Compose

“unsupported runtime” or deploy fails: Each service in runfabric.compose.yml must point to a valid runfabric.yml with a supported provider and runtime (e.g. nodejs20.x). Fix the service config or runtime in the referenced runfabric.yml.
Concurrency: Compose deploys services in dependency order; multiple services to the same provider run sequentially unless you use a higher --concurrency (see COMMAND_REFERENCE.md).

See also

CREDENTIALS.md — credential resolution and env vars
PROVIDER_SETUP.md — per-provider setup and optional env
COMMAND_REFERENCE.md — flags and commands

This site is open source. Improve this page.