Using with AWS Lambda

[outerlook]

Hey, can someone confirm I can use this with AWS Lambda?

I've been trying to use GAds API with lambda but my functions keep timing out. I've tried with 1 min timeout/ 1024 mb memory (although it only reaches 360mb max).

Locally it works well with the same config and function. My lambda is not in a VPC. I've tried different operations (query/mutate) with no success, even the simplest.

Thanks in advance for any experience shared.

[outerlook]

To better define my environment if someone is getting the same behavior, I use:

• AWS-CDK (so it builds actually with esbuild to upload the zip)
• Node v14

The hardest part is that it doesn't error anywhere. I don't truly know if the error is on which side. It just hangs even if I try to list customers for a refresh key (that works locally)

[rosslavery]

Not much more help but should be noted Opteo is aware of some of the challenges of running inside Lamba environments:

https://github.com/Opteo/google-ads-node/issues/17

[outerlook]

Thanks, @rosslavery ! A lot clearer to know there is WIP or even just not fully explored yet. Then I know I am not missing something obvious. :)

[Zikoel]

I have another experience. This was with google cloud run that is slightly different but maybe AWS lambda works in a similar manner. What I have experimented* is that the resource attibution for a single run follow this steps:

1. An externat trigger fire the execution (http requesto, scheduling etc etc)
2. Resources was assigned to the runner (ram, cpu) etc etc
3. Call to your code
4. wait for response
5. Resource deallocation
6. Wait the runner gracefully shut down

Now if in my code I do this:

1. handle the request
2. do the hard work
3. respond to the http request
4. gracefull shutdown

all go as expected... BUT if I do this:

1. handle the requeste
2. respond to the http request
3. do the hard work
4. gracefull shutdown

here the job take a very long time and reach the timeout because after point 2 and before point 3 google remove all the allocated resources and leave only something very neglettable waiting your complete shutdown.

So the poin is: are you sure that you do the hard memory consuming work on the right place ? Maybe this is something that happen only in google cloud but keep me almos an entire day before debug the problem... so maybe a double check can be usefull.

* The experiment was done some months ago so maybe in the meantime things are changed and actually I don't find the docs where this behaviour was documented.