// insights
Readings from the probe network.
What we learn from watching endpoints and machines all day: how a check decides something is really down, what to watch on certificates, servers and cron jobs, and what we changed in the product this month. No growth hacks, no filler.
latest first
- 04
Reminders that get quieter
One Slack message for a three-hour outage is easy to miss and hard to forgive. relay19 now sends still-down reminders on a tapering schedule: four in the first hour, then less and less. Here is why the schedule decays instead of repeating.
product · 3 min read
- 03
The backup that stopped running in March
Cron jobs fail silently by design. Heartbeat monitoring inverts the check: instead of asking whether a job broke, it notices when a job stops telling you it finished.
monitoring · 3 min read
- 02
TLS expiry is an outage you can see coming
Certificate expiry is the only outage with a published start time, and it still takes sites down every week. What to watch, how far ahead to warn, and the renewal failures that a simple expiry check will miss.
monitoring · 4 min read
- 01
Three misses before we page you
A single failed check means almost nothing. Here is why relay19 waits for consecutive failures, what that costs you in detection time, and how to tune the threshold for the monitors that actually matter.
engineering · 3 min read