How do I tell Prefect to run exactly one copy of a Flow star

Question

How do I tell Prefect to run exactly one copy of a Flow starting immediately until the Flow ends variable amount of time then immediately start another single copy and so forth by copy I just mean I wouldn t want the Flow runs to overlap

John Muehlhausen · Answer

perhaps I m asking if a Flow can begin on a schedule of a Flow ending then after starting the Flow manually it would never end

Kevin Kho · Answer

Hey < John Muehlhausen> use it in the flow level It will look something like this ```def mystatehandler obj old state new state if new state is finished client create flow run return new state with Flow state handlers = mystatehandler as flow ``` This will also be a first class feature in Orion Prefect 2 0

John Muehlhausen · Answer

create flow run appears to want a flow id how does my flow know it s own flow id

Michael Adkins · Answer

Your flow needs to be registered to be spawnable by the API You can pass a flow name and project name combination to uniquely identify it instead

John Muehlhausen · Answer

```prefect client client Client create flow run flow id=None context=None parameters=None run config=None labels=None scheduled start time=None idempotency key=None run name=None version group id=None ```

Michael Adkins · Answer

Oh my apologies I was referring to the `create flow run` task

John Muehlhausen · Answer

I don t see project flow name

Michael Adkins · Answer

Michael Adkins · Answer

You can use that task and just call ` run `

Michael Adkins · Answer

Or you can query for the flow and use the client method

Michael Adkins · Answer

The source code of the `create flow run` task may be helpful here

John Muehlhausen · Answer

`from prefect tasks prefect import create flow run`

John Muehlhausen · Answer

`create flow run flow name= `

Michael Adkins · Answer

``` if flow name flow = FlowView from flow name flow name project name=project name client = Client flow run id = client create flow run flow id=flow flow id ```

John Muehlhausen · Answer

This told me I have to run in a Flow context

Michael Adkins · Answer

You d have to call `create flow run run ` to trigger the task code outside of a `with Flow ` block

Michael Adkins · Answer

I d recommend instantiating a `FlowView` as I shared above to get the flow id instead it s simpler

Anna Geller · Answer

< John Muehlhausen> here is one example that you could use any time your flow run ends successfully a new flow run gets triggered if you want it to be triggered even on failure i e as long as it s finished use this instead as Kevin pointed out ```if new state is finished ```

John Muehlhausen · Answer

< Anna Geller> thank you Was thinking about dropping in at the event tonight but parent teacher conference had priority slightly smiling face

John Muehlhausen · Answer

Any chance this would make flows too hard to kill I deleted some flows Prefect Cloud UI that insert into a database The most recently installed one does indeed insert on the condition of interest but 5 other versions of the same record at nearly the same time show up in the db I ve logged to Prefect all insert statements being executed and there is only one of them so I m wondering if I have some rogue flows that are not visible on the Prefect Cloud dashboard Ever hear of anything like that

Anna Geller · Answer

gt I m wondering if I have some rogue flows that are not visible on the Prefect Cloud dashboard This shouldn t happen gt Any chance this would make flows too hard to kill In general cancellation is a best effort attempt to kill execution initiated on your infrastructure The best way to tackle it would be to use MERGE INTO or upsert instead of insert Would it make sense to apply that in your use case

John Muehlhausen · Answer

gt cancellation is a best effort attempt to kill execution initiated on your infrastructure

John Muehlhausen · Answer

Yes for some reason Cloud lost control of an agent that was still executing Hmm

Kevin Kho · Answer

So if the flow has compute that is happening on a different machine kubernetes pod or dask then it is very hard to cancel those processes as well For Docker Kubernetes you may still see work being run if the agent process dies