Introducing incident detection: one alert that says what broke
Apply.Build now connects your app's error logs, metrics, traces and crashes into one incident and alerts you in Slack or Discord. No SDK, any log format, every plan.
Johannes Peltola · · 3 min read
Most hosting platforms give you graphs and a log stream, and leave the rest to you. You have to notice something is wrong, find the right graph, and dig through the logs to work out why. Usually a user tells you first.
Starting today, Apply.Build tells you. Incident detection watches every app, connects the signals that belong together, and sends you one alert that says what broke. There is nothing to install, and it's included on every plan, the free tier too.
What a bad release looks like
Say you deploy a change that leaks memory. Within minutes your app runs out of memory, crashes, restarts and crashes again. Requests that hit it mid-crash fail, and the ones that get through are slow.
Each of those shows up somewhere different: a restart counter, a spike of server errors, a slower response-time graph, and a few hundred near-identical lines in the logs. Put together, the story is obvious. Spread across four places, it takes a while to see.
One incident, not four signals
Apply.Build reads the same signals and does the putting together for you:
- Logs. Error lines are picked out of your logs in whatever format you write them, plain text, JSON or stack traces, and grouped into patterns. Three hundred copies of one error count as one.
- Metrics. Server error rates and response times are compared with what is usual for your app.
- Traces. They show which endpoints are failing or slow.
- Lifecycle. Restarts, crash loops and out-of-memory kills.
When signals show up together, they become one incident with a severity, sent to Slack or Discord:
When your app recovers, the incident closes itself.
No SDK, no agent
Error-tracking tools usually start with an SDK in your code: one per language, configured, kept up to date, and paid for separately. Logs in a format they don't expect are just text to them.
Apply.Build already runs your app, so it already sees its logs, requests and restarts. Incident detection uses those directly. Nothing in your code or Dockerfile changes, and it works from your app's first deploy.
Not every error is a problem
Some errors are expected, like a failed login or a client hanging up early. Click Ignore next to an error and it stops raising incidents. Errors that look like obvious noise are ignored automatically, and you can undo either at any time.
It's included on every plan, alongside the firewall and vulnerability scanning, at the same fixed monthly price.
All of this is included on every app
Every Apply.Build app gets a firewall, vulnerability scanning and incident detection, at no extra cost.