Skip to content

Blog

Blog

Writing from the realuptime team on monitoring, incident response, and what we're building. New posts every few days.

What a TCP port check proves, and what it does notA TCP port monitor answers one question: does this host accept a connection on this port, from here, right now? That is the right question for databases, mail, SSH and brokers, and the wrong one for a web app. Where each check type fits, and what each one cannot see.September 4, 2026 · 7 min readYour cron job stopped and nobody noticed. Heartbeat monitoring, explained.Uptime checks cannot see a backup script or a queue worker, because there is nothing to probe. Heartbeat monitoring turns the check around: the job reports in, and silence is the alarm. How it works, how to size the grace period, and the mistakes that make it useless.September 2, 2026 · 8 min readHow often should an uptime monitor check your site? Less than the interval decidesOne minute, five minutes, thirty seconds: the check interval only sets a ceiling on how fast you find out. Confirmation, timeouts, and region quorum decide the real number. The math, with worked examples.September 1, 2026 · 10 min readChatGPT is down. The status page checkmarks are green. Both are true.Is ChatGPT down right now? Measured from four regions: the OpenAI API is fine, the ChatGPT product is not, and the component board stayed green through a major incident.August 31, 2026 · 6 min readWhat it takes to track 900+ services honestlyA week ago the outage catalog covered 195 services. It now covers more than 900. The story is not the count, it is the rejection list behind it: what we would not add, and why.August 31, 2026 · 6 min readRealUptime is publicWe launched on Product Hunt today, and our listings on AlternativeTo and SaaSHub are live. Here is what we built, why, and a launch discount.August 27, 2026 · 3 min readUptime monitor false positives: the page you should not have gottenFalse positive alerts train teams to ignore real ones. How we decide a failure is real before waking anyone up, and what we refuse to page on.August 27, 2026 · 4 min readWhen the monitor is the problemAn uptime monitor that fails quietly is worse than no monitor at all. We keep a public register of the ways ours can fail and what happens for each.August 27, 2026 · 4 min readOne internet, four speedsWe ping the same 190 services from Virginia, California, Frankfurt, and Tokyo every few minutes. The answers disagree more than you'd think.August 25, 2026 · 3 min readStatus pages that can't lieMost status pages are marketing documents that occasionally mention incidents. We built ours so that the tempting lies are structurally unavailable.August 25, 2026 · 3 min read