scheduling flows

Though I have never done it, due to the large number of workunits it would create, it seems you should be able to use Frequency in Seconds option under Scheduling (Rich Client).

I use an external process to create a separate file and then use a File Channel to kick off.  I am thinking that you could also use a "sleep" timer to loop and re-check --- therefore to only kick off the AGS call when there is something to change -- AND only have 1 work unit running.

 

Thanks Todd. I like the single WU sleep idea. Will explore.

I've used the 'sleep' node before and we always seem to have issues with it. We use scheduled processes all the time and they work great. As long as you have an efficient workflow that's the route that I would take. The flow will complete and will no longer be consuming system resources. The one workunit running on a sleep node seems to be very inefficient. However, we have a pretty beefy landmark server and do regular maintenance to clean up old workunit records. So, our applications will differ.

brox07, in my experience the wait node has issues when you are trying to wait while needing to save the state of the workunit. When the workunit resumes after the wait period, you have to reload variable values, etc. So it works as expected only in very limited circumstances. An alternative--assuming you have access to the SQL command node, which is not available for multi-tenant cloud--is to use SQL wait commands via SQL command node. Not a great solution, but faster and easier than designing around the wait node's limitations.

I have a flow I've run for Years, scheduled every 20 minutes. This check for auto jobs in recovery or wait state, gathers data and emails user to manage it, and emails me a list of them all. Never had an issue. Not sure if I would want a job 'open' all the time as it's taking up a spot in the queue. I have others that run every 15 minutes as well. The key is knowing the normal processing time  of the flow and being aware of the settings (threshold / Continue to run / retries). Had a flow once that opened itself over one that was already processing but having an issue.

BTW - just now debugging another users flow that inconsistently fails. Uses a sleep node and then a system node. When it does fail, it's right after the sleep.