Hey all, not to pile on the questions, but can anyone also tell me if there's a way to truly delete a sensor from dagster? I think I broke this one sensor that absolutely will not trigger because of some dagster configurations and want dagster to recognize it as a brand new sensor that has never ran before. Any help is greatly appreciated

I think I broke this one sensor that absolutely will not trigger because of some dagster configurations

Can you share some more details about you mean here, maybe a screen shot? @daniel / @prha

Hi @Alex Despotakis… are you using the run key to enforce idempotence? I think we query runs in the run storage for tags with the sensor name and the run_key pairing.

I can definitely explain it in excruciating detail, assuming I renamed my sensor. 1. I make a new sensor that looks at my filesystem for documents with a minimum interval of 30 seconds. The sensor finds the files in a folder called "to_process", yields the run requests, and triggers the pipeline. Once my first solid executes, it takes all of the files in "to_process" and moves them to a "processed" folder. The sensor continuously runs, yielding a skip since there are no more documents to process. The last tick is at 11:47am. I turn the sensor off. 2. I notice that for some reason, the files were not processed correctly. I re-populate the "to_process" folder with the same files and turn the sensor back on. As of 11:52am, the sensor has not been triggered. The sensor-daemon continuously returns these logs:

2021-03-17 11:52:19 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4
2021-03-17 11:52:24 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4
2021-03-17 11:52:29 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4
2021-03-17 11:52:34 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4
2021-03-17 11:52:39 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4
2021-03-17 11:52:44 - SensorDaemon - INFO - Checking for new runs for the following sensors: PullDocsFromFileSystem4

3. I refresh the page, turn the sensor back off, then delete the latest runs in the dagster UI. I refresh the page again and turn the sensor back on. The sensor is still showing the latest tick of 11:47am with no indication of the next tick. Refreshing the metadata for the repository doesn't seem to do anything either. The sensor will infinitely stall and as of 11:57am, no new runs were triggered. The short term solution of how to trigger it is as @prha suggested: changing the name from

PullDocsFromFileSystem4

to

PullDocsFromFileSystem5

will work. The sensor will trigger like it was brand new; however, to have a continuously changing sensor name in a repo would not be ideal. Additionally, I am not using a run key. The sensor context object passed through the sensor isn't touched in any way. Also, thank you @prha for your responses! I will try deleting the corresponding rows in the

job_ticks

table and see if that fixes anything.

im a little confused by entry 3 - is there a “turned the sensor back on” that is missing? As it is currently written it sounds like the sensor is just in the off state.

Hey @alex, you're right. Making the edit now

and what version of dagster are you running?

I'm on version 0.10.7

what do you get when you run

dagster sensor preview PullDocsFromFileSystem4

The sensor executes just fine. It yields the expected run requests, then exits.