Clarify AWS credentials requirements for local DB about serverless-graphql

[TomBarta]

I had this problem today, too.

export AWS_ACCESS_KEY_ID=id_key
export AWS_SECRET_ACCESS_KEY=access_key

solved it on my OSX and Ubuntu 16.04 machines.

[Aciid]

same issue here, I've checked that the relevant profile is mentioned in ~/.aws/config and ~/.aws/credentials

serverless-starter works, maybe theres something out of date here.

[laardee]

@ismarslomic yes, I had the aws cli installed beforehand and I probably used similar instructions when I configured the credentials. I have exactly same versions of python, serverless and os x installed.

[Glavin001]

I had the same problem.

I think 1 thing that helped was downloading a ZIP of the master branch of this project and copying over all of the files -- it appears that the installation did not use the latest code -- and then I changed name field in s-project.json to your applicable project name. I now have it working. I also tried installing DynamoDB Local manually vs using docker-compose up, however now they both work.
I spent a long time trying things to get this to work so at this point I am not even confident which 1 or so things ended up fixing it. However I think updating to the latest code definitely helped, if not being the actual solution.

Hope that helps!

---

I remember what fixed it! See #31 (comment)
So yes, the fix is to download and merge in a copy of the latest code on the master branch which includes that fixes within Pull request #31.

[Aciid]

Hey @Glavin001 ,

I tried the master branch but that didn't work out any better. I also checked out the pull request, doesn't seem that relevant. Thanks for the hint tho, I haven't followed the progress of this repository that much but It's evolving quite nicely.

I'm more concerned about backend-git/back/api/lib/dynamodb.js which has the actual mentions on how to drive locally this is where the "credentials" are missing. as the sessionToken doesn't get validated by the local instance.

But as none other project's are currently using local dynamodb with sls offline capabilities I don't think theres an easy way to peek for the answer other than debugging :E, I guess it's off to using sls offline with a live table ¯_(ツ)_/¯ until this gets resolved.

[Glavin001]

Hmm I'll try to figure out what I did to fix that issue. It was a long day last week, I remember that much.

[laardee]

@jesucarr @Glavin001 @Aciid How did you install the boilerplate? With sls project install serverless-graphql or cloning the repo and running sls project init? I haven't touched the admin.env and it has my personal profile settings mapped for both stages that I've created, dev when installing the project and tests later with sls stage create

AWS_DEV_PROFILE=laardee
AWS_TESTS_PROFILE=laardee

(I have my AWS work profile as [default] and [laardee] for personal stuff)

[Aciid]

@TomBarta this worked :3 thanks!

@laardee AWS_TESTS_PROFILE had no effect and I didn't have this after init ( wasn't asked for this in the config ) weird is this used for the offline then tho?

[ismarslomic]

I had same issue as well. Exporting the two env variables as suggested by @TomBarta resolved the issue. I guess this is just an workaround, as credentials shall be retrieved from ~/.aws/credentials? If i is not an workaround, then documentation shall be updated

[laardee]

@Aciid, that "AWS_TESTS_PROFILE" profile is for "tests" stage that I used. If you haven't created a stage tests there should not be such a profile. When you create a new stage with sls stage create it appends new profile mapping to the admin.env, so I don't think you need to manually change that file at all if you use serverless framework to create/init project.

[laardee]

I just can't reproduce this issue. I created new stage that uses different credentials and everything works ok. I haven't exported either work or personal credentials with

export AWS_ACCESS_KEY_ID=id_key
export AWS_SECRET_ACCESS_KEY=access_key

I logged credentials in setup db and it prints out correct ones for each stage.

Which OS / Serverless version / AWS CLI version are you using?

[ismarslomic]

@laardee do you have aws cli installed and have you done profile configuration as described here http://docs.aws.amazon.com/cli/latest/userguide/cli-chap-getting-started.html before installing serverless and creating new project? I have currently not installed aws cli, so that might be the issue?

[ismarslomic]

Just to add another observation here: after exporting these two variables, admin.env doesn´t seem to be in use anymore. I opene admin.env and added an invalid value to AWS_DEV_PROFILE, but i still managed to run the function locally.

[Aciid]

@laardee so if you run
echo $AWS_ACCESS_KEY_ID you get blank? Sorry for being so straight, but I'm just trying verify that you are not using the credentials from "AWS-CLI" standpoint (which we have verified now working)

My serverless-graphql version is "version": "0.0.4",
AWS-CLI aws-cli/1.7.0 Python/2.7.10 Darwin/15.4.0
And yes os x.

[laardee]

I'm away from the keyboard for couple hours, I'll get back to this issue later today.

[erikerikson]

Making no commentary about this specific issue.

Given the credential loading questions here, you can review how that happens in this thread:
serverless/serverless#740

Note that in specifying a profile you can declare a general AWS_PROFILE for use in all stages or a stage specific AWS_[STAGE]_PROFILE.

A later PR added support for ~/.aws/config:
serverless/serverless#917

[ismarslomic]

I have following setup:

• Python 2.7.10
• Serverless 0.5.5
• OS X El Capitan (10.11.5)

I do not have installed AWS CLI.

[ismarslomic]

Ok, so then I guess installing AWS CLI shall be prerequisite for this project in order to run functions with local database? By the way, how did you install AWS CLI on Mac? by use of pip or bundle, or?

[laardee]

@ismarslomic with pip

[laardee]

@Aciid echo $AWS_ACCESS_KEY_ID prints out empty line. I have little bit newer aws-cli installed aws-cli/1.10.24 Python/2.7.10 Darwin/15.5.0 botocore/1.4.15, I just updated it about a week ago, but it worked with older version as well. I don't remember which was the version I had previously, but it was installed maybe a 6 months ago.

[ismarslomic]

@laardee: I have now installed AWS CLI (aws-cli/1.10.28 Python/2.7.10 Darwin/15.5.0 botocore/1.4.19), but only installing AWS CLI didnt help on this credential issue. Only after running aws configure I managed to run functions with local db successfully. However, executing echo $AWS_ACCESS_KEY_ID still prints empty line.

[laardee]

@ismarslomic Everything works ok locally if credentials are set with aws configure?

[ismarslomic]

@laardee yes. Im though not sure how selection/creation of aws profile as part of serverless project create impacts or interacts together with AWS CLI profile configuration. As mentioned previously, if I change AWS_DEV_PROFILE=222 (invalid profile name) in admin.env, Im still able to successfully run the function locally.

[laardee]

There is some magic 🎩 🐇. I can also successfully run the project if I have correct AWS_[STAGE]_PROFILE but with invalid value.

[ismarslomic]

But, if you try to deploy your function with invalid value in admin.env serverless will complain about unknown profile. So I guess that the aws profile mentioned in admin.env is used mainly in AWS and not locally. Would be great to have someone from the serverless team to confirm our assumptions here and also update README so it is more clear.

[jesucarr]

@laardee I installed the boilerplate with sls project install serverless-graphql. I have not installed the AWS CLI.

[ChaosSmith]

Was having issues reading AWS credentials on OSX after initiating a project, checked the admin.env variable and found that the AWS_DEV_PROFILE was in quotes while on my ubuntu machine it was not. Removed the quotes and suddenly credential reading worked like a charm.

[Glavin001]

There's an ongoing issue with aws-sdk-js here: aws/aws-sdk-js#692