agents-store Troubleshooting
Diagnostic steps and fixes for common Trigger.dev v4 problems.
Quick Diagnostics
Run these checks first:
- CLI auth:
npx trigger.dev@latest whoami - Dev server:
npx trigger.dev@latest dev— starts without errors? - If MCP:
list_orgs()— returns organizations? - Recent failures:
list_runs(status="FAILED", period="1d")
Task Errors
| Status | Cause | Fix |
|---|---|---|
| FAILED | Task threw an error | Check run details → read error trace |
| CRASHED | OOM or unexpected crash | Increase machine preset or fix memory leak |
| SYSTEM_FAILURE | Infrastructure issue | Check supervisor container health |
| TIMED_OUT | Exceeded maxDuration | Increase maxDuration or optimize task |
| EXPIRED | TTL expired before execution | Raise ttl on the trigger / task / global default, or drain the queue. See config-and-build and task-development for TTL precedence. |
| PENDING_VERSION | No matching worker deployed | Deploy to the correct environment |
Connection Errors
| Error | Cause | Fix |
|---|---|---|
| ECONNREFUSED | Instance not running | Start: docker compose up -d |
| ETIMEDOUT | Network/firewall | Check DNS, firewall, proxy config |
| ENOTFOUND | Wrong URL | Verify TRIGGER_API_URL |
| SSL_ERROR | Invalid certificate | Check reverse proxy SSL config |
| Redirected to cloud | Missing API URL | Use login -a <url> or set TRIGGER_API_URL |
Authentication Errors
| Error | Cause | Fix |
|---|---|---|
| 401 Unauthorized | Invalid secret key | Check format: tr_dev_xxx (dev), tr_prod_xxx (prod) |
| 403 Forbidden | Wrong environment key | Use dev key for dev, prod key for prod |
| Token expired | Key rotated | Generate new key in dashboard |
| CI auth fails | Missing access token | Set TRIGGER_ACCESS_TOKEN (tr_pat_xxx) |
Deployment Errors
| Error | Cause | Fix |
|---|---|---|
| Build failed | TypeScript errors | Fix compile errors in your task code |
denied: requested access to the resource is denied |
Not logged into self-hosted registry | docker login -u $DOCKER_REGISTRY_USERNAME $DOCKER_REGISTRY_URL (see Registry Push Failures below) |
unauthorized: authentication required |
Registry credentials wrong or expired | docker logout $URL + fresh login with current password |
no basic auth credentials |
Docker credentials store empty for this host | Run docker login on the machine; in CI, add login step before deploy |
Push to localhost:5000 fails on CI |
DEPLOY_REGISTRY_HOST on server still set to default |
Set public hostname in webapp .env + restart webapp |
| "No tasks found" | Wrong dirs config |
Check dirs in trigger.config.ts |
| Version mismatch | SDK/CLI out of sync | npm install @trigger.dev/sdk@latest |
Registry Push Failures (Self-Hosted)
When trigger.dev deploy fails at the push step, the developer/CI machine can't authenticate with the instance's built-in registry:
# Interactive (dev laptop)
docker login -u registry-user registry.your-domain.com
# Non-interactive (CI)
echo "$DOCKER_REGISTRY_PASSWORD" | docker login \
"$DOCKER_REGISTRY_URL" \
-u "$DOCKER_REGISTRY_USERNAME" \
--password-stdin
# Verify creds are valid
curl -s -o /dev/null -w "HTTP %{http_code}\n" \
-u "$DOCKER_REGISTRY_USERNAME:$DOCKER_REGISTRY_PASSWORD" \
"https://$DOCKER_REGISTRY_URL/v2/"
# → HTTP 200 means login will succeed
# → HTTP 401 means wrong username/password
Common root causes:
- CI runner has no persisted Docker credentials → add
docker loginstep before deploy - Registry password rotated on server but local docker keychain still has old password →
docker logout+ re-login - Using
localhost:5000publicly → changeDEPLOY_REGISTRY_HOSTto public hostname in webapp.env, restart webapp, re-login clients
Dev Server Issues
| Issue | Cause | Fix |
|---|---|---|
| Cannot find trigger.config.ts | Wrong directory | Run from project root or pass --config |
| Tasks not registering | File not in dirs |
Check trigger.config.ts dirs |
| Hot reload not working | File watcher issue | Restart dev server |
Self-Hosted v4 Diagnostics
Check Instance Health
# Webapp
cd trigger.dev/hosting/docker/webapp
docker compose ps
docker compose logs -f webapp
# Supervisor
cd trigger.dev/hosting/docker/worker
docker compose ps
docker compose logs -f supervisor
Common Self-Hosted Issues
| Issue | Cause | Fix |
|---|---|---|
| All runs CRASHED | Supervisor OOM | Increase Docker memory limits |
| Runs stuck QUEUED | Supervisor disconnected | Check worker token; restart supervisor |
| Dashboard 502 | Webapp crashed | docker compose restart webapp |
| Worker token missing | Separate machines | Check webapp logs for token on first start |
| Registry auth fails | Wrong credentials or stale docker keychain | See "Registry Push Failures" above — verify via curl -u user:pass https://$URL/v2/ returns 200 |
| Object storage error | MinIO bucket missing | Create packets bucket via MinIO UI (:9001) |
| Disk full | Old runs not cleaned | Configure retention, clean old data |
Log Analysis
# Find errors in webapp
docker compose logs webapp 2>&1 | grep -i error | tail -20
# Check for OOM kills
docker compose logs supervisor 2>&1 | grep -i "oom\|killed" | tail -10
# Check registry
curl -s http://localhost:5000/v2/
# Check MinIO
curl -s http://localhost:9000/minio/health/live
Query & Dashboards (TRQL)
| Symptom | Cause | Fix |
|---|---|---|
| "AST too complex" | Query too large / nested joins | Split the query, simplify aggregates |
| "Row limit exceeded" | > 10 000 rows | Add LIMIT + a time filter via period |
| "Concurrent queries exceeded" | Dashboard or batch query spam | Reduce widget refresh rate; stagger batch queries |
get_query_schema error "table required" |
v4.4.4 now requires a table arg | Pass table: "runs", "metrics", or "llm_metrics" |
Column value not found on metrics |
v4.4.4 renamed/clarified column | Use metric_value |
See the observability skill for the full TRQL syntax reference.
MCP Tool Annotations (v4.4.4+)
Every MCP tool now carries readOnlyHint / destructiveHint annotations. If your MCP client respects them, agents can be prevented from calling write tools without explicit approval. For a hard-enforced read-only deployment, install with npx trigger.dev@latest install-mcp --readonly — deploy, trigger_task, and cancel_run are hidden server-side.
Deeper Reference
- @references/common-errors.md — extended error catalog