Intermittent "invalid character '/' after top-level value" error about terraform-resource HOT 10 CLOSED

@divyabhargov I have never seen that error. Is this during a put to the terraform resource? Could you attach a screenshot?

from terraform-resource.

Here are some screenshots when the error occurs:

from terraform-resource.

@divyabhargov this is a weird one, still looking into it. Started a convo with the Concourse team here: https://concourseci.slack.com/archives/C07RY25QF/p1510072021000782. Seems like something is printing to Stdout that shouldn't be but I can't find where that would be happening.

Couple questions:

• Are you using ljfranklin/terraform-resource:latest or some other image?
• Are you using an S3-compatible storage with a custom endpoint or AWS S3?
• Any other things that are unique about your Terraform templates or terraform-resource config?

from terraform-resource.

Hello, we already had a random fail with this error also :)

We are using the latest terraform resource, with a S3 endpoint. First time we had this problem, and weren't able to reproduce it.

from terraform-resource.

Good grief, I just looked through all my old builds and this has happened to me 2 times out of 100+ builds. I'll try to add some debug logging to a custom image and run it in a loop or something.

from terraform-resource.

Still struggling to reproduce this. Latest finding is that the "invalid character '/' after top-level value" is reproducible in isolation if you try to unmarshal a log time stamp as JSON, e.g. log.Println will print 2017/11/08 16:39:08 .... Concourse tries to unmarshal everything written to Stdout as a JSON response which is where the error is getting raised. So somewhere in the code there's a log statement that is going to Stdout. But I can't find it, maybe it is buried in some dependency.

from terraform-resource.

This was fun. I copied stdout to stderr in a test image and ran the job in a loop all day. On run number 422 I finally got the log message: 2017/11/08 21:48:35 Request body type has been overwritten. May cause race conditions. Looks like this was an issue fixed on the aws SDK: aws/aws-sdk-go#868. I'll push a fix in a bit to bump the AWS SDK and nil out its logger.

from terraform-resource.

Thank you so much. This is exciting

from terraform-resource.

Thanks for taking a look at this, and tracking it all the way down.

from terraform-resource.

I just had this error and it showed up every time I ran a plan at random points of the plan. It complained about a seemingly random character (sometimes alphanumerics, sometimes slashes).

I found out I had a custom module instantiation with a parameter the module didn't define. After fixing this error, I did not get any more of these cryptic Terraform errors.

from terraform-resource.