Sharing Sessions and Clients within multiple goroutines

[sevberg]

Hello,

I'm curious if there are any "best practices" or "known pit falls" related to sharing Sessions and Clients built with this SDK across multiple goroutines at once?

I know this question is vague, but nevertheless I hope it is still answerable. If not, I can come up with a toy example which can showcase a basic use-case. But, basically, what I would do would be to:

• Create a Session object and Client object in main
• Then pass references to the client to many go-functions: go read_from_s3( &client, s3_key)
• Then do stuff with that client (e.g. reading some data from S3), then return...something (this part doesn't matter so much..)
• Collect all the futures in main
• End main

Are there any drawbacks to this approach? Would you suggest a better/smarter/safer way to achieve concurrent processes talking to AWS?

As a bit of background, I have extensive experience with the Python AWS SDK (Boto3) which often has many problems when used within an async or multithreaded context. So much so that I've learned to avoid combining Boto3 with this at all costs. So now that my team and I are considering porting some of our code into Go, I'm concerned that the Go SDK could suffer from similar concerns

[alexbilbie]

@sevberg did you come to a conclusion on this?