Uploading snapshot always gets stuck with no info

[grosser]

cargo make ami
[cargo-make] INFO - cargo make 0.36.3
[cargo-make] INFO - Build File: Makefile.toml
[cargo-make] INFO - Task: ami
[cargo-make] INFO - Profile: development
[cargo-make] INFO - Running Task: setup
[cargo-make] INFO - Running Task: setup-build
[cargo-make] INFO - Running Task: fetch-sources
[cargo-make] INFO - Running Task: publish-tools
[cargo-make] INFO - Running Task: ami
/Users/user/Code : decoded 2147483648 bytes
/Users/user/Code : decoded 1073741824 bytes
02:56:32 [INFO] Registering 'bottlerocket-aws-k8s-1.22-x86_64-v1.11.1-104f8e0f-dirty' in us-west-2
  Uploading snapshot  [=>                                                ] 90/4096 (5s)

and it just sits there ...
on multiple tries the number is always different 90/4096 or 62/4096

Platform I'm building on:

• Mac M1
• v1.11.1

What I expected to happen:
Upload to succeed or fail with an error message.

What actually happened:
Upload gets stuck.

How to reproduce the problem:
make ami
+

# Infra.toml
[aws]
regions = ["us-west-2"]

[stmcginnis]

Hey @grosser - thanks for reporting this. I'm unable to reproduce the failure locally, so I think we'll have to try to see what might be different in your environment that could be causing this error.

Based on the output, it looks like this is sitting somewhere in the coldsnap tool. It looks like it must be in this loop:

https://github.com/awslabs/coldsnap/blob/develop/src/upload.rs#L171-L188

That's looping through each block of the snapshot and uploading to EBS.

It hasn't out right failed (yet) so I think it's in that retry back off. If I'm reading that right, if it fails uploading a block it will retry up to 12 11 times, backing off (attempt * 2) each time. The call to upload_block() probably has its own internal timeout as well, so that could increase the amount of time to takes.

It may be interesting to let is sit in this state for some time and see if it ever does completely time out and give any kind of useful error message.

Another option would be to run the coldsnap command directly. Unfortunately it doesn't look like there is anything like a --verbose arg that might tell us more, but it would at least isolate things.

You could also try to manually upload the img directly with EBS and see if that provides any additional information. It's possible that could fail too, and if so, provide a better message of what is happening that is causing this problem.

[grosser]

I'm not sure what the total time math comes out to, but some kind of feedback after >2-5min "Warning: retrying failed connection" would be nice :)
I let it run for >1.5h and nothing happened.
I'm not sure how to use coldsnap since I don't know what the path the the image is that it's trying to upload (--verbose option would be helpful there too).
... I could try uploading a random file with coldsnap, is that what you are suggesting ?

[grosser]

Ok found it ... permission error, so the bug is that it should not retry on these

Failed to put block 1551 for snapshot 'snap-0f48e9c316f6fa504': TransientError: connection closed before message completed
Failed to put block 1552 for snapshot 'snap-0f48e9c316f6fa504': AccessDeniedException: User: arn:aws:sts::589470546123:assumed-role/compute-arf/[email protected] is not authorized to perform: ebs:PutSnapshotBlock on resource: arn:aws:ec2:us-west-2::snapshot/snap-0f48e9c316f6fa504 because no identity-based policy allows the ebs:PutSnapshotBlock action

[grosser]

... and it should also not retry for 2 hours 🤦

[grosser]

some documentation for "here are all the aws permissions you need" would be nice too

[stmcginnis]

I've filed an issue in coldsnap with these details: awslabs/coldsnap#216

I'm only familiar with that code as much as it took to track down where it was failing, but I'll see if I can dig in there and make some of those changes.

I'm going to close this issue as the fix will need to be in the coldsnap tool and there's nothing in bottlerocket we change to workaround it.