Pingdeck Changelog

Drop your live status into any site with a script tag:

<script src="https://cdn.pingdeck.com/embed.js" data-page="acme"></script>

5.4 KB gzipped, updates over SSE, follows the visitor's color scheme. There's an iframe fallback for locked-down CSPs. Quickstarts for Next.js, Astro, and plain HTML are in the docs.

Alert routing rules are live. Match on monitor tag and severity, deliver to a Slack channel, an email list, or a webhook. First match wins, with a default route so nothing is dropped. Database alerts to #infra, marketing-site blips to #general, everything else to email.

Two fixes for alert trust:

• Slow TLS handshakes no longer count as downtime. A 6-second cert negotiation is a performance problem, not an outage. It's now recorded as degraded latency and only alerts if the request actually fails.
• Custom-domain SSL renewals can't fail silently anymore. If a CAA record blocks renewal, you get a banner and an email well before expiry.

Response-time charts now show p50 with p95/p99 bands instead of a single average. Averages hide slow tails; the bands don't.

• Exponential backoff with jitter on webhook delivery retries
• Fixed a proration bug when upgrading mid-cycle with an annual coupon

Drag-and-drop CSV import with a mapping preview. Creates up to 200 monitors in one batch, dry-run mode included.

Identical alerts within 10 minutes collapse into one thread; message edits append ×N counter. Slack + email transports.

Public status pages now render in dark mode when the visitor's system asks for it. No setting to flip: it follows prefers-color-scheme.

Failed webhook deliveries now retry with exponential backoff and jitter instead of hammering a struggling endpoint on a fixed clock.

Upgrading mid-cycle with an annual coupon no longer doubles the proration. If this hit you, the difference has been credited.

Schedule recurring maintenance windows per monitor. Alerts are suppressed for the duration, uptime % excludes the window, and your status page shows a scheduled banner instead of a scary red bar.

Probes now also run from eu-central (Frankfurt). Pick regions per monitor. EU endpoints stop looking slower than they are, and latency charts split per region.

One goroutine pool per region instead of a shared queue. p99 scheduling jitter is down 38%, so response-time charts reflect your service, not our scheduler.

Uptime percentages were off by 0.01 when a maintenance window crossed midnight. Recomputed and corrected retroactively.