A monitor checks one thing on a schedule and tells you when it fails, recovers or gets slow.
Types
- HTTP(S): requests a URL and checks the status code. Keyword also looks for (or must not find) text in the page; JSON API checks a value, like
$.status == "ok". - Multi-step API: several requests in a row, passing values (such as a login token) from one to the next.
- Browser flow: opens a real browser to log in and click through a page. Available on paid plans.
- Heartbeat: your cron job or backup calls a URL when it finishes; you're alerted when the calls stop. Report a failure with
?status=down. - TCP port, Ping and DNS: is a port open, does a host answer, does a record resolve (to the right value)?
- Database: connects to Postgres, MySQL or Redis. The connection string is stored encrypted.
- Game servers: Minecraft, Steam/Source and FiveM, with player counts.
- Discord bot: is a bot online in your server? Needs your workspace's Discord bot, see Discord.
- Mail blocklist: is a sending IP listed on spam blocklists?
How often and when it counts as down
The interval is how often a check runs; the fastest allowed depends on your plan (every 3 minutes on Free, every minute on Pro, every 30 seconds on Business). Retries decide how many failed checks in a row it takes before a monitor is down, so a single hiccup doesn't wake anyone. After a failure, the retry runs within seconds instead of waiting a full interval.
Set a response time threshold to be told when a monitor stays slow for three checks in a row.
Checks from inside your network
Uplance's servers refuse private and internal addresses on purpose. To monitor something that isn't reachable from the internet, run a remote agent inside your network (Agents): a single file that runs checks there and reports back.
Fewer, better alerts
- Depends on: while a parent (say, your load balancer) is down, alerts of the monitors behind it are folded into the parent's.
- Flapping: a monitor that goes up and down repeatedly is marked as flapping instead of alerting every time.
- Deploy windows: send a deploy marker from CI to hold alerts for a few minutes during a rollout (see API and automation).
Pausing and deleting
Pausing stops checks and alerts but keeps the history. Deleting removes the monitor and its history for good.