What is your goal?
Completing incomplete execution with error between iterator and aggregator
What is the problem & what have you tried?
In my scenario, I iterate over e-mail attachments. For each attachments, I take multiple steps including AI extraction (openAI module). Sometimes, the openAI module returns an error, which is completely fine - outage etc.
However, after such error, I am unable to complete the incomplete execution. Both manual tries and automatic retries run by error handlers end up with the same error: linked feeder ‘71’ was not found for the aggregator ‘99’'.
I’ve had the same issue across organizations and across scenarios.
It seems that the aggregator does not “see” the output variables of the iterator module (outlook - list attachments). When I open the aggregator settings, it has the proper input variables listed, reselecting the correct variables and saving them does not help. The scenario works with the same settings most of the time. But if an execution is interrupted, suddenly, the agggregator stops working.
Seems like a bug to me.
Error messages or input/output bundles
Scenario initialization failed with an error.
Cannot initialize the scenario because of the reason ‘Unexpected configuration error: linked feeder ‘71’ was not found for the aggregator ‘99’’
Screenshots (scenario flow, module settings, errors)
Add an error handler on the OpenAI module using Resume or Ignore with fallback values, so the bundle carries on to the aggregator instead of leaving an incomplete execution. Also check the aggregator’s Source Module is explicitly set to the attachments iterator. Runs that already broke may need clearing rather than retrying.
I tried error handling with retry module - ignore or resume won’t do the work for me. I really need the AI output to continue in the flow. Retry behaviour is the same as rerunning incomplete execution.
The source module in the aggregator is properly set.
The thing is that this scenario needs to be ready for automatic retries in some way - AI simply has outages, it’s completely normal. Source of the primary error is expected.
Fair enough, if you need the output then ignore won’t cut it. Try a Break handler on the OpenAI module with a few retry attempts and a decent interval, so it reruns that bundle on its own. Make sure incomplete executions are switched on in the scenario settings.
Break (now renamed as retry) is there. The first time the retry module actually tries to retry the scenario, I get the configuration error. That means the execution fails in the phase of initialization, no module can even try to run. Delay and number of retries cannot unfortunately influence this.
Thanks for trying to help - I am already in touch with Make support on this one. For now, they are trying to find out why this keeps happening.
I’ll post here the resolution once I have it.
Fair enough Cenka, if it fails before any module can run then a retry handler won’t touch it. Sounds like a platform side bug, so support is the right place. Keen to see what they find, please post the fix here.
