Remove predictive filesystem alerts. They're not useful for us at this point.

See merge request !3

Increase InstanceDown timeout

See merge request !5

Wait 15 min before alerting on InstanceDown.

Add printf to githost.rules

See merge request !4

Reduce the excessive decimal places in storage space alert messages.

Adjust NodeFilesystemFullSoon alert

See merge request !3

* Remove 4 hour prediction because it is too sensitive.
* Replace with 12 hour prediction.

Tune node alerts

See merge request !2

* Increase time to alert for 24h NodeFilesystemFullSoon to avoid flaps.
* Re-add NodeFilesystemFull 20% warning alert.
* Tune NodeFilesystemFull to 15min to avoid flaps.
* Remove `humanize` from percentages.
* Ignore backup snapshot mountpoints in `predict_linear()` alerts.

Update githost alerting rules

See merge request !1

* Use Prometheus standard CamelCase for alert names.
* Use consistent annotations (`summary`, `description`, `runbook`)
* Use consistent formatting, trailing commas, etc.
* Add `NodeFilesystemFullSoon` alert using `predict_linear()`
  - Two separate alerts for slow and fast disk filling issues.
  - Uses 1:4 ratio for history to prediction to avoid noise.
* Add `NodeFilesystemReadOnly` to detect broken systems.
* Drop warning alerts.

Add githost login instructions

See merge request !2

add .gitlab-ci.yml

See merge request !1