summaryrefslogtreecommitdiffstats
path: root/doc/02-Behavior.md
diff options
context:
space:
mode:
Diffstat (limited to 'doc/02-Behavior.md')
-rw-r--r--doc/02-Behavior.md101
1 files changed, 101 insertions, 0 deletions
diff --git a/doc/02-Behavior.md b/doc/02-Behavior.md
new file mode 100644
index 0000000..ae2444b
--- /dev/null
+++ b/doc/02-Behavior.md
@@ -0,0 +1,101 @@
+Behavior
+========
+
+Top Level View uses additional status logic for it's views.
+
+This does not affect the overall status behavior of Icinga 2 or Icinga Web 2,
+but it is important to understand the differences.
+
+## Worst status on top
+
+The main responsibility of TLV is to show you the worst status on top.
+
+Worst status is defined in the following order:
+
+* critical_unhandled
+* warning_unhandled
+* unknown_unhandled
+* critical_handled
+* warning_handled
+* unknown_handled
+* ok
+* downtime_handled
+* missing
+
+In addition counter badges will present you indepth details over the
+status below a tile or tree element.
+
+Similar to Icinga Web 2 you can easily see unhandled problems by the strength of color.
+
+![Unhandled problems](screenshots/colors-unhandled.png)
+![Handled problems](screenshots/colors-handled.png)
+
+## SOFT states
+
+While the normal monitoring views will always show you all current states,
+the Top Level Views will only show hard states.
+
+Which means, as long as the object doesn't have reached a hard state, the
+node should be OK and green.
+
+## Handled and Unhandled
+
+Icinga Web 2 introduced an handled state to every host and service.
+
+By default handled would be true if:
+
+* Problem has been acknowledged
+* Object is in downtime
+
+In Top Level View, a few things are different:
+
+* Downtimes are handled special (see next topic)
+* Notification settings can influence a status (see next topic)
+* Flapping means the state is handled
+
+## Downtime and Notification Periods
+
+Since downtime and notification settings are essential for alerting,
+Top Level Views tries to integrate these into its status logic.
+
+The following behaviors will trigger the downtime logic:
+
+* Host or Service is in an active downtime
+* Notifications are disabled for the host or service
+* Host or Service is out of its notification period
+
+If those conditions are met:
+
+* TLV counts the service as `downtime_active`
+* TLV ignored non-OK states and marks them as `downtime_handled`
+
+## Additional Features
+
+Some features can be enabled by view, and control some additional behavior.
+
+### Host always handled
+
+Option `host_never_unhandled`
+
+Every host problem is set to unhandled by default.
+
+This helps when alerting is mostly based on service states, and the host
+is only a container.
+
+This was a behavior of the old Top Level View, and is enabled on convert.
+
+### Enabling notification period
+
+Option `notification_periods`
+
+Checks the configured notification_period and handles "out of period" as downtime active state.
+
+This was a behavior of the old Top Level View, and is enabled on convert.
+
+Also see [limitations](90-Limits.md) for this setting.
+
+### Ignoring certain notification periods
+
+Option `ignored_notification_periods`
+
+This notification periods will be ignored for "out of period" checking, see above.