No instances should be ok?!

The --no_instances=OK switch  has been available for some time now. It prevents for example a LUN check that can’t find any LUNs with a not always matching UNKNOWN not to seek any unneeded attention. Recently, we received an increasing number of requests that express the desire for such a switch for collectors (i.e the getter scripts) as well. An understandable argument is, that the getter should always be configured, since a variety of objects, such as LUNs or auto-grow volume are created very dynamically or disappear from a system for short periods, events that should not be viewed as errors and cause warnings to be sent. From my point of view, it would make sense to change our current policy of interpreting the existence of an object as the normal state - this would be correct for volumes and aggregates. In this case, we would change the default configuration of the getters and checks so that the **non-existence **of objects does not trigger an alarm, therefore is OK. Of course, this setting can be configured according to your needs. The behavior of existing configurations, that don’t have an explicit option set, would change, until explicitly specified if and how the non-existence of instances should send a warning. In principle, we don’t change the default settings in new releases. In this case, this principle contradicts another principle: the default settings for checks and getters should follow current best practices. Comments are welcome - please use the comment boxes bellow or write to us at il@netapp-monitoring.info .


-timeout Bug
Testing/Fixing Release 3.4.1_01

Comments