--- name: aquaria-cloudflare-ops description: Operate and debug Aquaria Cloudflare Worker + Workflows deployment. Deploy safely run health checks inspect traces restart stuck instances apply fixes. references: - core-structure - build-patterns - install-script - aquaria-cli - workflows - wrangler - incident-response --- # Aquaria Cloudflare Ops (rss-feed-processor) ## When to apply - Deploy `workflows\/packages\/rss-feed-processor` Worker - Check production health daily checks workflow failures - Debug Cloudflare Workflows `item-process-workflow` `feed-poll-workflow` `qa-analysis-workflow` `daily-summary` - Mitigate incidents kill switch bulk terminate restart stuck instances ## Inputs \/ prerequisites Provide one of: 1. **Project path** (preferred): `--project workflows\/packages\/rss-feed-processor` 2. **Explicit URL**: `--url https:\/\/..workers.dev` Credentials: - **Admin key** for `\/api\/admin\/*` and `\/api\/traces\/*`: - `--admin-key` flag or env `AQUARIA_ADMIN_KEY` \/ `ADMIN_KEY` - **Cloudflare API token** for deep inspection and restart fallback: - env `CLOUDFLARE_API_TOKEN` ## Rules 1. Always start narrow then widen failure window (`--since 30m` before `24h`) 2. Never print `.env*` contents or secrets in terminal output 3. Quote operational lessons from `~\/.factory\/sessions\/*.jsonl` not inline secrets 4. Use `--step-output=false` for Wrangler to avoid dumping huge outputs 5. Enable kill switch before investigating incidents ## Workflow decision tree ``` Request: deploy rss-feed-processor A. Pre-deploy: test → build → kill-switch enable (if incident) B. Deploy: wrangler deploy → capture version ID C. Post-deploy: verify → daily-check 30m → trace list --status failure D. If failures: trace show → diagnose → fix → redeploy Request: check prod health A. Start narrow: daily-check --since 30m B. If issues: widen to 2h → trace list --status failure C. If failures found: trace show → inspect step output Request: debug stuck workflow A. List: workflow list --workflow --status errored B. Inspect: trace show OR wrangler describe C. Action: restart OR terminate Request: incident response A. Immediate (0-5m): kill-switch enable → verify → daily-check 1h B. Diagnose (5-15m): trace list --status failure → trace show C. Fix → deploy with kill-switch ON D. Recovery: post-deploy checks → kill-switch disable only after stable ``` ## Examples **Input:** Deploy rss-feed-processor **Output:** ```bash pnpm -C workflows\/packages\/rss-feed-processor test -- --run --timeout 60000 pnpm -C workflows\/packages\/rss-feed-processor build pnpm -C workflows\/packages\/rss-feed-processor run deploy pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor verify pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor daily-check --since 30m ``` **Input:** Production showing failures **Output:** ```bash pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor daily-check --since 30m pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor trace list --workflow item-process --since 30m --status failure # Pick instanceId from output pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor trace show --instance-id ``` **Input:** SQLITE_TOOBIG error in traces **Output:** ```bash # Confirm failing step pnpm -s aquaria cloudflare --project workflows\/packages\/rss-feed-processor trace show --instance-id # Or use Wrangler for full detail pnpm -C workflows\/packages\/rss-feed-processor exec wrangler workflows instances describe item-process-workflow --step-output=false --truncate-output-limit 2000 # Fix: apply trace filtering to exclude large fields # Deploy and watch traces list --status failure --since 30m ```