503 ERROR due lambda timeout after tokens are fetched

[mishabruml]

What happened:

My cloudfront dist redirects me to the cognito login UI and I successfully auth. After that, I get directed to the cloudfront 503 error page rather than my s3 static content:

503 ERROR
The request could not be satisfied.
The Lambda function associated with the CloudFront distribution is invalid or doesn't have the required permissions. We can't connect to the server for this app or website at this time. There might be too much traffic or a configuration error. Try again later, or contact the app or website owner.
If you provide content to customers through CloudFront, you can find steps to troubleshoot and help prevent this error by reviewing the CloudFront documentation.

Generated by cloudfront (CloudFront)
Request ID: BO_FjHzTRlnA-HWDVRWznjZbcpOBBMjEAT2mwvZUxje6BITPn2bbJg==

The logs in my lambda@edge appear to show the initial invocation successfully recirecting the user to the UI page:

logs

``` { "level": 20, "time": 1705676579380, "msg": "Handling Lambda@Edge event", "event": { "Records": [ ....big blob redacted here ] } } { "level": 20, "time": 1705676579400, "msg": "Cookies weren't present in the request" } { "level": 20, "time": 1705676579400, "msg": "User isn't authenticated: Error: Cookies weren't present in the request" } { "level": 20, "time": 1705676579400, "msg": "Redirecting user to Cognito User Pool URL https://***********" } ```

Then after entering my login credentials, after 5s I am presented with the 503 screen in the browser. The lambda appears to time out after fetching the tokens.

logs

``` { "level": 20, "time": 1705676947082, "msg": "Handling Lambda@Edge event", "event": { "Records": [...redacted blob] } } { "level": 20, "time": 1705676947102, "msg": "Cookies weren't present in the request" } { "level": 20, "time": 1705676947102, "msg": "User isn't authenticated: Error: Cookies weren't present in the request" } { "level": 20, "time": 1705676947102, "msg": "Fetching tokens from grant code...", "request": { "url": "https://****.auth.****.amazoncognito.com/oauth2/token", "method": "POST", "headers": { "Content-Type": "application/x-www-form-urlencoded" }, "data": "client_id=******&code=******&grant_type=authorization_code&redirect_uri=******.cloudfront.net" }, "code": "******" } { "level": 20, "time": 1705676949484, "msg": "Fetched tokens", "tokens": { "id_token": "*****", "access_token": "****", "refresh_token": "****", "expires_in": 28800, "token_type": "Bearer" } } ```

[mishabruml]

I did some debugging afger forking the repo and adding logs everywhere, and found that the lambda was timing out here https://github.com/awslabs/cognito-at-edge/blob/e8c7e305b2fe87d0fbc47bcde560b4526a0235d2/src/index.ts#L279 verifying the JWT token. Others appeared to have had this issue https://github.com/awslabs/aws-jwt-verify/issues/72 https://github.com/awslabs/aws-jwt-verify/issues/133 so I have made a PR #88 to address this. Changing the timeout to 5000ms fixed the issue for me.

[ckifer]

+1 I have the same issue but only when authenticating with refresh tokens.

Edit: jk mine is more of an infinite redirect issue

[wesdek]

+1 I have the same issue but only when authenticating with refresh tokens.

Edit: jk mine is more of an infinite redirect issue

You need to set your config with: "cookiePath": "/" to fix your redirect issue

[ckifer]

Yep, found the other issue. Thanks!

[ckifer]

Now I'm experiencing this after resolving the other issue. Still +1

2024-01-25T02:28:37.114Z ... Task timed out after 3.06 seconds

This seems to be happening only when refresh tokens are fetched and in regions that are further out (APAC / South America) because of the axios call to cognito in the set region. Is there any way we can do this with redirects instead?

@maverick089 as the contributor of refresh tokens

[lenfree]

Not sure if these would address the 5 second time out issue:

1. increase the timeout to 5 seconds
2. enable the AWS_NODEJS_CONNECTION_REUSE_ENABLED

   process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1';

[ckifer]

Increasing the timeout worked for me... for now. Lambda@edge doesn't support timeouts > 5 seconds though. Also doesn't support env vars

[lenfree]

  process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1';

That's right, it does not support Lambda environment variables but you can manually set it with below if I am not mistaken. Cold start might also contribute to the issue.

process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1';

[ckifer]

Can give it a shot. Though I think that it's mostly because of region distance. An edge function in Singapore is making a call to Cognito in Oregon

[manu-remsense]

Hey guys, I was having the same issue, and thankfully I was able to come across this issue and resolve the problem with what @mishabruml has suggested, thanks 🙏 .

[ckifer]

Sweet, thanks for the info!

[manu-remsense]

Ops, actually it didn't help, I tested it wrong. I understand the issue on my end a bit better, and turns out I have the same problem as you @ckifer. When a token is expired it has to verify the token, then verify the refresh token and then fetch the new token - process seems to take longer than 5s time limit when cognito and cloudfront locations are far apart. I still have the timeout increased for the SimpleJsonFetcher in place but it hasn't really resolved it for me. Did you manage to find something that fixes it @ckifer ? Thanks

[ckifer]

No resolution here no. Luckily I don't have too many users that are far away, but still would like to find a better solution.

[lenfree]

I had the same issue too and so far this worked for me, or maybe I just haven't tested it properly.

[manu-remsense]

I had the same issue too and so far this worked for me, or maybe I just haven't tested it properly.

@lenfree I've put that line process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1'; outside of the handler, and it didn't seem to improve the situation. Or did I not do it correctly? Thanks

[manu-remsense]

Anyway, as a band-aid solution we decided to forgo the use of refresh tokens when a user is accessing outside of our region (australia)

Changed this line in handle(event) function if (tokens.refreshToken) To, if (tokens.refreshToken && process.env.AWS_REGION === 'ap-southeast-2') (just in case anyone else is facing the same problem and is out of ideas)

[ckifer]

That's actually a solid workaround, nice!

[lenfree]

I had the same issue too and so far this worked for me, or maybe I just haven't tested it properly.

@lenfree I've put that line process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1'; outside of the handler, and it didn't seem to improve the situation. Or did I not do it correctly? Thanks

I think it needs to be inside the handler, at least that's how I configured it.

[ckifer]

if (tokens.refreshToken && process.env.AWS_REGION === 'ap-southeast-2')

@manu-remsense Is this you modifying the cognito-at-edge code? Wish there was a config item for that...

[manu-remsense]

I had the same issue too and so far this worked for me, or maybe I just haven't tested it properly.

@lenfree I've put that line process.env['AWS_NODEJS_CONNECTION_REUSE_ENABLED'] = '1'; outside of the handler, and it didn't seem to improve the situation. Or did I not do it correctly? Thanks

I think it needs to be inside the handler, at least that's how I configured it.

Ohh I see, I should give it a try 🤔

@ckifer yes that's the only way as far as I know, I can make a PR later when I have time on the weekend (but not sure how useful it would be to people).

[ckifer]

Got it. I can just use patch-package for now. Since this is an issue I believe being able to conditionally turn off the use of refresh tokens should be useful to more than just two of us