Problem with shinyloadtest::load_runs on remote server

[watson-microexcel]

We are trying to setup a load test for a remote server and having issues loading the report.

We are able to record the sessions and run the the recording with several workers on a remote machine, but we are running into problems when we trying to load the runs using:
df <- shinyloadtest::load_runs("mytest" = "./run1")

We are receiving the following error:

Error: $ operator is invalid for atomic vectors

We are able to run this on our local dev environment with 1 worker without an issue. The only difference we see in the recording.log file is the w${WORKER} and the target types ("Shiny Server or Shiny Server Pro" vs "R/Shiny")

What is the best way to debug this? Are we missing something?

[alandipert]

Hi, thanks for the report. I haven't seen that one before. Would you please reproduce the error, then call traceback() and report back with the full stack trace? That would help us narrow it down.

[alandipert]

Oh, another thing that would be helpful for you to try would be to transfer the run1 from the remote machine to your local machine, and see if load_runs() fails with that data locally.

Also, if you're an RStudio customer, don't hesitate to send an e-mail to support@rstudio.com referencing this thread and my name. If you go that route, you could send us your run1 data confidentially and I could try to reproduce the problem.

[watson-microexcel]

Hi, thanks for the report. I haven't seen that one before. Would you please reproduce the error, then call traceback() and report back with the full stack trace? That would help us narrow it down.

Thanks for the quick reply.

Here is the traceback() results.

load_runs_traceback.txt

In response to your second message: We are recording and running (shinycannon) on our local machines. We have the recordings locally. Should we be running them on the remote machine?

[alandipert]

Thanks for the that traceback(). I'm studying it.

Re: recordings, sorry, I was confused. I was under the impression that you were running load_runs() on the remote machine, but it sounds like you have run1 locally and are running load_runs() locally. Is my understanding correct?

It would be most helpful if you were able to somehow send the contents of run1 to me so that I could reproduce myself. A copy of the recording could be helpful too. Let me know by email to alan@rstudio.com if that would be feasible.

Otherwise, I'll try to figure it out just based on the trace and let you know what I find.

[alandipert]

@schloerke studied the traceback and suggested the problem might be around JSON deserialization. It seems likely there's a JSON object in the run log with an unexpected structure, and assumptions in load_runs() that a named list exists are invalid. It's definitely odd that we haven't encountered this before in our own testing, so we're very interested in helping you get to the bottom of this.

Unfortunately there are a few places the error could be happening, and it's hard to know which one without your run1 data. So, if you could somehow provide that data to us, we'd be grateful.

[watson-microexcel]

I have emailed you a sample that has the problem we are experiencing. Thanks.