Posting status to a measurement with exception message when exception is thrown out of node

[dzmitry-moisa]

The plugin generates a lot of junk when any command generates exception out of node construction. It is creating a measurement with an exception message instead of writing to job or stage measurement.

Steps to reproduce:

1. Configure connection to InfluxDB
2. Crete a pipeline job
3. Generate an exception not wrapping it in node() {...} or stage() {...}
4. The plugin creates a measurement with the name of exception message

[jeffpearce]

I've been trying to figure out a better way to handle these for awhile. Originally I added the ability to log exceptions outside of the pipeline because we had a couple of scripts that did this, and I wanted to be able to get counts got particular errors. However, in our case, I wish I'd taking the approach that they were badly written scripts, as the exceptions could have happened inside a stage, they just didn't.

Curious what would work for you. It seems clear that it should write to the job measurement, but if an exception happens outside of a stage, I can see a case for not trying to log that.

Option 1: ignore exceptions that happen outside of a stage, and just write the job result
Option 2: use a generic name when writing to the stage measurement, something like "non-stage error"
Option 3: something else?

[dzmitry-moisa]

I think that Option 2 would be the better way. It could allow to get the satats of such cases and fix them asap.

[jeffpearce]

That's the way I was leaning as well. I'll make that change. Thanks for reporting!

[FCamborda]

Heyo, has this been already addressed? :)

[UlrichBlunck]

Hey @jeffpearce ,
we've encountered this error on our monitoring database as well. And since we help collecting data for multiple projects it's kind of hard to give a proper overview over the number of occurring errors.
Having this wrapped in a more easily matchable way would be super helpful! ;) +1