What it is for
It answers "is meulog well?". An error monitor that is broken and quiet is worse than none: everyone thinks they are covered.
Health
Nine items, each with a state and a sentence saying what to do when it is bad:
- Read: the database responds.
- Queue: jobs are moving, and how many failed.
- Scheduled jobs: meulog own scheduled tasks ran.
- Inputs: something arrived recently through each input that is on.
- URLs: the watched ones are being checked on schedule.
- Databases: each account opens, and how much it takes.
- Evidence: the attachments are where they should be.
- Backup: when the last copy was.
- Time zone: the system and meulog agree.
The queue
How many jobs are waiting, how many failed, and which. A failed job carries its error message: this is where a configuration problem shows up before it becomes missing data.
What is slow
The most expensive queries in the last window, with the time for each. meulog reference is 200 ms: above that the screen stops feeling instant.
When someone says "I got no alert". Before hunting for the rule, check whether the queue moved and whether the input brought anything. Half of every "the alert never came" is "the error never came".