p24-infra Docs

n8n-ai-webhook-down

3 min read

BrandPilot — n8n AI Webhook Down

Trigger: Campaign generation in BrandPilot fails with “N8N webhook failed” or similar error. Users report that AI-generated content is not being produced.

Workflow: SvcDlrMxBMN9aTUJ on bms-4 (54.36.123.110) n8n host: https://n8n.bms-4.infra.zintegrowana.online AI path: n8n → SSH → claude CLI on AI-Dev-IO1 (vps-i1, 217.154.82.162) → response

1. Confirm

Check the BrandPilot app — trigger a test campaign generation. If the request times out or returns an error containing “webhook” or “n8n”, proceed.

2. Diagnose

2a. Is n8n on bms-4 alive?

curl https://n8n.bms-4.infra.zintegrowana.online/healthz
# Expect: { "status": "ok" }

Not reachable → n8n container is down. Go to Fix A (requires p24-infra SSH session).

2b. Is workflow SvcDlrMxBMN9aTUJ active?

Check via n8n API (requires BMS4_N8N_API_KEY from p24-infra/.env.local or SOPS):

curl -H "X-N8N-API-KEY: $BMS4_N8N_API_KEY" \
  https://n8n.bms-4.infra.zintegrowana.online/api/v1/workflows/SvcDlrMxBMN9aTUJ

Check "active": true. If false → Go to Fix B.

Also check recent executions:

curl -H "X-N8N-API-KEY: $BMS4_N8N_API_KEY" \
  "https://n8n.bms-4.infra.zintegrowana.online/api/v1/executions?workflowId=SvcDlrMxBMN9aTUJ&limit=5"

Look for status: error and inspect data.resultData.error for root cause.

2c. Is the SSH connection to AI-Dev-IO1 working?

The workflow SSHes into AI-Dev-IO1 (vps-i1, 217.154.82.162) and runs claude CLI.

Test from a p24-infra SSH session (requires access to bms-4 or vps-i1 directly):

# From any machine with the n8n SSH credential key:
ssh root@217.154.82.162 "claude --version"
# Should return something like: Claude Code 1.x.x

If SSH fails → check n8n credential vps-i1-root-ssh (id: nleNO94baxznp9P0). Key may need rotation.

If claude not found → the CLI was removed or path changed on AI-Dev-IO1. Requires p24-infra session.

2d. Is the claude subscription active on AI-Dev-IO1?

ssh root@217.154.82.162 "echo 'Say hello' | claude -p 'You are a test' 2>&1"
# Should return a short response. Error 401/auth failure = subscription issue.

If authentication fails → Claude subscription on AI-Dev-IO1 needs renewal (human action).

3. Fix

Fix A: Restart n8n on bms-4

Requires p24-infra SSH session:

ssh root@54.36.123.110
cd /opt/n8n  # or wherever docker-compose.yml lives
docker compose restart n8n n8n-worker-1 n8n-worker-2 n8n-worker-3
# Wait 30s, then:
curl https://n8n.bms-4.infra.zintegrowana.online/healthz

Fix B: Activate workflow

curl -X POST \
  -H "X-N8N-API-KEY: $BMS4_N8N_API_KEY" \
  https://n8n.bms-4.infra.zintegrowana.online/api/v1/workflows/SvcDlrMxBMN9aTUJ/activate

Or in n8n UI: toggle the workflow to “Active”.

Fix C: Re-establish SSH credential

If the SSH connection from n8n to AI-Dev-IO1 is broken (key mismatch or rotated):

  1. Open n8n UI → Settings → Credentials → find vps-i1-root-ssh (id: nleNO94baxznp9P0)
  2. Update the private key with the current C:\Users\konar\.ssh\id_ed25519 key
  3. Test the credential, then re-run a test execution of SvcDlrMxBMN9aTUJ

4. Escalate to human when

  • Claude subscription expired on AI-Dev-IO1 — SSH command returns auth failure. Human must re-authenticate claude CLI on vps-i1.
  • n8n worker queue backlogged >100 executions — may indicate memory issue on bms-4; requires capacity decision.
  • bms-4 host unreachable → escalate to p24-infra session (SSH, Proxmox console if needed).
  • vps-i1 unreachable → check IONOS console; AI-Dev-IO1 may need restart.

5. Verify fix

Re-trigger a test campaign from the BrandPilot UI and confirm content is generated.

Check n8n execution history for SvcDlrMxBMN9aTUJ — latest execution should show status: success.

Graph View

Backlinks