The way a combination of different dependencies work together is not always clear. In order for a node to be scheduled:
- Parent: must be free in order for its children to run. A family must be free before any of its tasks can run.
- Date: must be free. This is checked when you begin a suite begin(CLI) and at midnight for the suite. There is no midnight for a hybrid clock.
- Time: must be free. This is checked every minute (configurable.)
- Trigger: must be free. This is checked every time there is a state change in the suite for all potential nodes.
- Limit: must not be full. Boolean limits just have slots in them, while integer limits must have enough space for the usage.
Here are a few examples of combinations with their behaviour.
To run task x only once, on 17th of February 2012 at 10 am:
task x time 10:00 date 17.2.2012
To run task x twice, at 10am and 8pm, on both 17th and 19th of February 2012, that is, four times in all. Notice the task is queued in between, and completes only after the last run. Under a hybrid clock, the task is run (twice) only if the date is either 17th or 19th at the time the suite begins:
task x time 10:00 ; time 20:00 date 17.2.2012 ; date 19.2.2012
To run task x after task y is complete and if the day is Monday. If the suite is using a real clock, the task waits for the following Monday (unless today is Monday) and for task y to be complete. Under a hybrid clock, if today is not Monday when the suite begins or is auto-restarted, task x is marked complete without being submitted:
task y task x trigger ./y == complete day monday
The next example shows how to run a task after an earlier task has stopped, either by completing or aborting. It may be useful to continue after a task has tried a few times but still failed. However, this technique should only be used if there is logic somewhere to correct for the missing task. Otherwise, task y will fail as well.
task x task y trigger ./x == complete or ./x == aborted # #The above trigger using named operators: # # trigger x eq complete or x eq aborted # task z trigger (x==complete or x==aborted) and (y==complete or y==aborted)
To run a task on a series of given days, use the Python API:
t = ecflow.Task("x")
for i in [ 1 , 2, 4, 8, 16 ] :
t.add_date(i,0,0) # 0 means any day,month or year
Will be displayed by the show(CLI) command as:
task x date 1.*.* date 2.*.* date 4.*.*
To run a task half after the previous task fc has done half of its work:
task fc meter hour 0 240 task half trigger fc:hour >= 120 # trigger fc:hour ge 120
There is no guarantee that a task will be sent at the exact moment requested. At the specified time, ecFlow might be busy processing other tasks. ecFlow does not check time dependencies constantly, but sweeps through them once a minute. This makes processing of events in ecFlow much more stable.