Server error when navigating to admin image panel

[noahmanger]

Something is broken with the S3 static asset work we merged in earlier. Now going to the Images tab in the admin panel yields a 500 error. cc @patphongs @ccostino

[patphongs]

@ccostino Can you please check if each of the environments (dev, stage, and prod) have these variables set? That might be the problem.

CMS_AWS_ACCESS_KEY
CMS_AWS_SECRET_ACCESS_KEY
CMS_AWS_STORAGE_BUCKET_NAME
CMS_AWS_DEFAULT_REGION_NAME

[ccostino]

All three environments do have those set, but I will continue to take a look and see if I can reproduce this locally too.

[ccostino]

Come to think of it, is it possible that we need to rename these environment variables to not have the CMS_ in front of them? The AWS utilities (such as boto3) expect the environment variables to be named a certain way (see AWS CLI Documentation and boto3 documentation around this)

@LindsayYoung, do you happen to know if these were working as they are called now prior to the changes we were working on for the S3 buckets? Or do you know who might?

[ccostino]

I am still stuck on this. At the moment the best I've been able to do is see a 500 response being returned in the CMS logs (e.g., on Staging), but there's no actual error or stack trace being shown either.

I was unsure if it was the proxy app having an issue or the CMS, but it does appear to be the CMS itself given the log output of the one response for admin/images/, but again, I'm not sure what is actually failing, and I've not been able to reproduce the problem locally either.

[ccostino]

Thanks @LindsayYoung and @patphongs for a couple of leads to look into! I'm going to continue check connectivity between the CMS and S3, make sure environment variables are being read in and used properly, and that the underlying path is being constructed correctly.

We've ruled out the CORS setup being an issue because that has already been done in all of the S3 buckets and proven to work as demonstrated in the links in https://github.com/18F/fec-proxy/pull/11.

[ccostino]

Looks like the issue was due to a couple of mis-named environment variable references that I just didn't catch. I'm sorry about that, and a big thank you to @LindsayYoung for spotting them!

[patphongs]

major thanks to both of you @ccostino and @LindsayYoung !

[ccostino]

Final tests show this is okay, closing this issue and moving forward!

[noahmanger]

Nice work, all!