> Agent-readable docs index: /llms.txt. Download /docs.zip to grep all markdown files locally. --- $schema: https://holocron.so/frontmatter.json title: Health Checks description: Monitor URLs on a schedule. Get alerted when endpoints go down, recover notifications when they come back, and auto-disable checks that fail for too long. icon: lucide:heart-pulse --- # Health Checks Strada health checks fetch a URL on a schedule and alert you when it fails. Think of it as a simpler [Checkly](https://www.checklyhq.com/). No browser checks, no scripting, no multi-step flows. Just: fetch a URL, check the status code, alert on consecutive failures. ```diagram strada checks create Cloudflare Workflow --url https://api.example.com/health (every 5 min) --name "API health" │ │ ▼ ▼ ┌──────────────┐ ┌──────────────┐ │ fetch URL │ │ D1 database │◀─── config ─────────────│ measure ms │ │ (alert_rule) │ │ check 2xx │ └──────────────┘ └──────┬───────┘ │ ┌──────────────┼──────────────┐ ▼ ▼ ▼ ClickHouse consecutive email/webhook results table failure check alert sent ``` Health checks run as a **Cloudflare Workflow** with each tenant org as a separate durable step. If one org's database is slow, other orgs are not affected. Steps retry independently on failure. *** ## Creating a health check ```bash strada checks create --url https://api.example.com/health --name "API health" ``` The check starts running within 5 minutes. By default it expects a **2xx status code**, checks every **5 minutes**, and alerts after **2 consecutive failures**. Alerts go to the same destinations as error alerts. If you have no destinations yet, create an alert rule with a destination first: ```bash strada alerts create --name "Notifications" --channel email --to ops@example.com ``` ### Options | Flag | Default | Description | | ---------------------- | ------------- | --------------------------------------------------------------- | | `--url` | required | URL to fetch | | `--name` | required | Human-readable name for the check | | `--method` | `GET` | HTTP method (`GET`, `HEAD`, `POST`, `PUT`, `DELETE`, `OPTIONS`) | | `--schedule` | `*/5 * * * *` | Cron expression in UTC (minimum granularity 5 min) | | `--timeout` | `10000` | Request timeout in milliseconds | | `--failures` | `2` | Consecutive failures before alerting | | `--status-min` | `200` | Minimum acceptable status code | | `--status-max` | `299` | Maximum acceptable status code | | `--cooldown` | `60` | Minutes to wait before re-alerting the same check | | `--auto-disable-hours` | `24` | Auto-disable after N hours of continuous failure (0 to disable) | | `--project` | all | Scope the check to a specific project | ### Example: strict health check ```bash strada checks create \ --url https://api.example.com/health \ --name "API health" \ --timeout 5000 \ --failures 3 \ --schedule "*/10 * * * *" \ --cooldown 30 ``` This checks every 10 minutes (cron), allows up to 5 seconds for a response, alerts after 3 consecutive failures, and waits 30 minutes before re-alerting. *** ## Listing and managing checks ```bash # List all health checks strada checks list # Delete a check (historical results remain in ClickHouse) strada checks delete 01KPVGTT9CJW4ZNEF414VHGRFD # Disable a check without deleting it strada checks disable 01KPVGTT9CJW4ZNEF414VHGRFD # Re-enable a disabled check strada checks enable 01KPVGTT9CJW4ZNEF414VHGRFD ``` *** ## How alerting works ### Consecutive failure detection A single failed check does not trigger an alert. The check must fail **N times in a row** (configured by `--failures`, default 2). This avoids false alarms from network blips or brief deploys. ```diagram check 1: ✓ pass check 2: ✗ fail ← failure count: 1 check 3: ✗ fail ← failure count: 2 → alert sent check 4: ✗ fail ← within cooldown, no re-alert check 5: ✓ pass ← recovery alert sent ``` ### Recovery alerts When a check transitions from **failing to passing** after an alert was sent, Strada sends a **recovery notification** so you know the issue resolved. No action needed; recovery alerts are automatic. ### Cooldown After an alert fires, Strada waits `--cooldown` minutes (default 60) before sending another alert for the same check. This prevents notification spam during extended outages. ### Auto-disable If a check has been failing continuously for longer than `--auto-disable-hours` (default 24 hours), Strada **automatically disables it** and sends a final notification. This prevents filling your database with thousands of identical failure rows from a decommissioned service or misconfigured URL. Re-enable with: ```bash strada checks enable ``` *** ## What gets stored on failure When a check fails, Strada stores the **response body** (truncated to 16KB) and **all response headers** (except `set-cookie`). This means an agent or human debugging the outage can see exactly what the server returned. Successful checks store only the status code and latency. No body or headers. ### Querying check results Health check results live in the `otel_health_checks` ClickHouse table. Query them with SQL: ```bash # Recent failures for a specific URL strada query " SELECT Timestamp, StatusCode, LatencyMs, ErrorMessage FROM otel_health_checks WHERE Success = 0 ORDER BY Timestamp DESC LIMIT 20 " # Availability percentage over the last 30 days strada query " SELECT countIf(Success = 1) * 100.0 / count() AS availability_pct FROM otel_health_checks WHERE Url = 'https://api.example.com/health' AND Timestamp >= now() - INTERVAL 30 DAY LIMIT 1 " # Latency percentiles strada query " SELECT quantile(0.5)(LatencyMs) AS p50, quantile(0.95)(LatencyMs) AS p95, quantile(0.99)(LatencyMs) AS p99 FROM otel_health_checks WHERE Url = 'https://api.example.com/health' AND Timestamp >= now() - INTERVAL 24 HOUR LIMIT 1 " # Read the response body from the last failure strada query " SELECT Timestamp, StatusCode, ResponseBody FROM otel_health_checks WHERE Success = 0 AND Url = 'https://api.example.com/health' ORDER BY Timestamp DESC LIMIT 1 " ``` *** ## Common workflows ### Monitor a production API ```bash # Create the check strada checks create \ --url https://api.example.com/health \ --name "Production API" # Make sure you have alert destinations strada alerts add --channel email --to oncall@example.com strada alerts add --channel webhook --to https://hooks.slack.com/services/... ``` ### Monitor multiple services ```bash strada checks create --url https://api.example.com/health --name "API" strada checks create --url https://example.com --name "Marketing site" strada checks create --url https://app.example.com/health --name "Dashboard" strada checks create --url https://ws.example.com/health --name "WebSocket server" --timeout 3000 ``` ### Check a non-standard health endpoint Some services return 204, 301, or other non-2xx status codes from health endpoints: ```bash # Accept 200-399 as healthy (includes redirects) strada checks create \ --url https://legacy.example.com/ping \ --name "Legacy service" \ --status-min 200 \ --status-max 399 ``` ### Disable auto-disable for critical checks For checks you never want automatically disabled (even if they fail for days), set `--auto-disable-hours 0`: ```bash strada checks create \ --url https://payments.example.com/health \ --name "Payment gateway" \ --auto-disable-hours 0 ``` *** ## Alert destinations Health checks share the same destination system as error alerts. One destination can receive both error alerts and health check alerts. See [Alerts and Destinations](/docs/alerts) for the full guide on setting up email, webhook, and Slack destinations.