Cloud
Notice it before the client calls
Good monitoring reports the fault before anyone picks up the phone. Bad monitoring reports everything and is therefore ignored.
Area
Cloud
Typical for
Systems in continuous operation
Principle
Few alerts that count
What it is about
The most common form of monitoring is a client phoning in. The second most common is a mailbox full of alerts nobody opens any more.
We set up a small number of alerts that genuinely require a response, and separate them from everything that is merely informational.
What gets watched
- Availability and response times of the application
- Errors in production, grouped rather than one by one
- Storage, load and certificate expiry
- Whether the backup actually ran
And when something happens
Every alert has a recipient and an expectation attached. Without both it is not monitoring, only a log.
Related