Deployment 'sync dependencies' step never ends

[bradley-varol]

I am unable to deploy to any of my environments . I've left the deployment running for an hour and below is what I see:

Fume CLI:

✔ Deps: 314.8MB Code: 7.9MB Assets: 879.7KB Total: 323.5MB Mode: efs
⠼ Sync dependencies
  › 100.00% complete

Fume web app:

[bradley-varol]

This morning, on the same Project, my deployment made it past this step but failed with an error I've not seen before. I'm going to try deleting the Environment once more and creating again.

[bradley-varol]

No luck. Back to 100% syncing dependencies unfortunately. I'd love some assistance with this issue so that I can migrate to Fume ASAP.

[bradley-varol]

Looks like Fume is storing my entire node_modules directory in S3. Is this expected behaviour?

When I deploy another project this doesn't happen. Nothing is uploaded to S3 during deployment. Is this something to do with npm vs yarn?

[bradley-varol]

Each time I run the deployment, some more dependencies are created in the node_modules S3 folder. This leads me to believe that something is timing out in the deployment before all the files can be created.

I deleted the S3 Fume bucket deps folder, ran a deploy and had 181 deps installed before the deploy script froze at 100% syncing. I've just ran it again and it hung at 85.20% with 896 deps. I'm going to keep going - I'm curious to see what happens when all deps are installed.

Edit:
I ran the deploy script three times. Each time I ran it, it synced some more dependencies. On the third time, it made it past the sync deps step.

Unfortunately the deployment was still unsuccessful. I have the subnet error from above and I'm getting it on all projects. I think perhaps something in AWS that was created when I connected Fume to AWS has been deleted (possibly when I deleted a project) and isn't being recreated.

[acidjazz]

@bradley-varol What's happening is Fume is attempting an EFS mode since your dependencies are too large to live in Lambda natively. The EFS support with Lambda is slow and not ideal.

I am currently working on a smoother faster solution involving Lambdas new Docker support that should solve this issue.

[bradley-varol]

Excellent, looking forward to that.