Status page

The public status page lives at status.persql.com. It’s open to the internet (no login) and refreshes every 30 seconds.

What it shows

• An overall operational / degraded / outage badge.
• 24-hour and last-hour uptime percentages.
• p50 and p95 probe latency over the last hour.
• A 24-hour minute-by-minute sparkline. Each cell hovers a tooltip with the timestamp, HTTP status, and round-trip latency.

How it’s measured

A synthetic probe runs every minute against the api worker’s /health endpoint via the public hostname (api.persql.com). That exercises:

• DNS + Cloudflare edge routing.
• The Worker isolate cold-start path.
• Fetch → handler → JSON response → fetch resolution.

A 2xx response within 5 seconds counts as healthy. Anything else (non-2xx, transport error, timeout) counts as a failed probe and contributes to a “down” minute on the page.

Status levels

Level	Trigger
operational	Last 15 minutes have zero failed probes
degraded	Any failed probe in the last 15 minutes
outage	Last probe failed AND ≥3 of the last 15 failed, or ≥8 of 15 failed

These are heuristics — a single failed probe shows as degraded for ~15 minutes even if everything has fully recovered. We err toward showing problems sooner rather than hiding them.

Programmatic access

The same data is available as JSON at /status.json:

curl -sS https://status.persql.com/status.json | jq .data.level
# "operational"

Schema:

interface StatusResponse {
  success: true;
  data: {
    level: "operational" | "degraded" | "outage";
    uptime24h: number;            // percentage 0–100, one decimal
    uptime1h: number;
    latencyMsP50: number | null;  // null when no healthy samples in the last hour
    latencyMsP95: number | null;
    lastSampleAt: number | null;  // unix-ms of the most recent probe
    lastFailureAt: number | null; // unix-ms of the most recent failure (any age)
    samples: Array<{
      ts: number;     // unix-ms of the minute bucket
      ms: number;     // probe latency in ms; -1 means transport failure
      status: number; // HTTP status; 0 means transport failure
    }>;
  };
}

This is a stable surface — feel free to scrape it from your own monitoring or wire it into PagerDuty / Slack.

Caveats

• The probe isn’t running queries against your data. It exercises the api worker; if a single tenant’s DO is unhealthy the page can still be green. For per-database health, run your own probe against /v1/db/:ns/:slug/tables.
• Buffer is 24 hours. Older history isn’t surfaced — the page is for current status, not long-term reliability reporting.