Closed prononext closed 1 year ago
Hey @prononext,
I think the binary data only exists on disk for a limited amount of time, Maybe 24 hours with the default settings so if you are rerunning a workflow that failed that could be why. The Gateway error could be something else as well it is hard to say without seeing the full error.
That is very strange. Binary data of saved executions should stay around indefinitely. Only for manual executions, where saving execution data is disabled will it be auto-deleted after a certain time.
But obviously should also never crash!
This was fixed in https://github.com/n8n-io/n8n/pull/4964.
But unfortunately, all executions (with nodes that don't use streaming) between 0.206.1
and 0.208.1
are missing the metadata file, and are prone to this crash.
I'll create a PR to handle this without crashing the service.
Fix got released with n8n@0.213.0
I had a workflow failing because of execution data was missing. It was a workflow that downloaded images, resized them and uploaded to s3 again.
I do not know why the data was not avilable while execution, but I noticed that when I went to the execution view in UI and choose the node with the problem. I tried to view the missing binary data and at that moment it got a "Gateway Error" and the whole UI Crashed with the n8n main container in my queue setup. I tried this 4 times with the same result.
There was a log in the n8n main service.
System Details: