So far we have written dependencies like
foo => bar. This is, in fact,
foo:succeed => bar. It means that the task
bar will run
foo has finished successfully. If
foo were to fail then
would not run. We will talk more about these task states
in the Runtime Section.
We refer to the
:succeed descriptor as a qualifier.
There are qualifiers for different task states e.g:
When the task has started running.
When the task finishes if it fails (produces non-zero return code).
When the task has completed (either succeeded or failed).
It is also possible to create your own qualifiers to handle events within your code (custom outputs).
For more information see the Cylc User Guide.
In Cylc, cycle points are just task labels. Tasks are triggered when their dependencies are met, regardless of cycle point. But clock triggers can be used to force tasks to wait for a particular real time, relative to their cycle point, before running. This is necessary for certain operational and monitoring systems, e.g. for tasks that process real-time data.
For example in the following workflow the cycle
2000-01-01T12Z will wait
until 11:00 on the 1st of January 2000 before running:
[scheduling] initial cycle point = 2000-01-01T00Z [[xtriggers]] PT1H_trigger = wall_clock(offset=-PT1H):PT30S [[graph]] # "daily" will run, at the earliest, one hour before midday. T12 = @PT1H_trigger => daily
See the Clock Triggered Tasks tutorial for more information.
By default Cylc uses the Gregorian calendar for datetime cycling, but it also supports:
360-day calendar (12 months of 30 days each in a year).
365-day calendar (never a leap year).
366-day calendar (always a leap year).
[scheduling] cycling mode = 360day