Open gabrielmoral opened 4 months ago
Thanks for raising this and apologies for not getting back earlier on this. Can you share debug level client side logs? I am specifically interested in knowing which credentials provider implementation is being used to fetch credentials.
I faced a similar issue in EKS with assumed IAM role credentials and found a workaround. I implemented my own AWS credentials rotation manager, which periodically retrieves AWS credentials by assuming an IAM role. These credentials are then passed to the AWSMSKAuthTokenGenerator.GenerateAuthTokenFromCredentialsProviderAsync
method, instead of using GenerateAuthTokenAsync
.
Thanks for raising this and apologies for not getting back earlier on this. Can you share debug level client side logs? I am specifically interested in knowing which credentials provider implementation is being used to fetch credentials.
@sankalpbhatia I'm sorry but I'm not in that space anymore, we decided to stick with Sasl for a while. Thanks for jumping anyways.
Describe the bug
I'm using
mskAuthTokenGenerator.GenerateAuthTokenAsync(region)
to get the token from the ambient credentials of an ECS (Fargate) task. It works fine when I deploy the app (new task created, new creds assigned) but after 6h~ there are kafka errors. I'm guessing that might be because the app creds are rotated at that time.Expected Behavior
I would expect that the creds rotation are handle gracefully.
Current Behavior
This the error that appears in then logs: [thrd:sasl_ssl://xyz]: sasl_ssl://xyz:9098/1: SASL authentication error: [85082b3a-a3be-460e-bbbd-15781f994aeb]: Session too short (after 567ms in state AUTH_REQ)
Reproduction Steps
Possible Solution
No response
Additional Information/Context
No response
Version used
1.0.0
Operating System and version
Docker container Alpine 3.19