Layers with size of around 2GB cause rce/Docker to exit with an out of memory error while updating.

[agherzan]

Can be reproduced by creating a image which contains a big file generated with dd (from /dev/random) via a RUN command.

[agherzan]

@imrehg is trying to reproduce this.

[imrehg]

@agherzan currently testing this through the regular resin infrastructure does not work do to an open registry issue: balena-io/open-balena-registry#8

As an alternate test: trying resin local push, I've successfully deployed a 4000MB layer fine.

root@cdffa2c:~# balena images
REPOSITORY                                                               TAG                 IMAGE ID            CREATED             SIZE
local-app                                                                latest              f132293eec96        4 minutes ago       4.24GB

Thus I would say:

• local push suggests that this might not be an issue anymore
• if it's still an issue, then it would be through deploying git push or resin deploy, and the registry's issue prevents such large images to reach the device at the moment anyways (ie. releases with such big layers cannot be created at the moment).

[agherzan]

Thanks @imrehg for this investigation. I will close in that case this issue for now.

[imrehg]

Just for reference, tried one more test based on comments by @shaunmulligan: push an image with a 2000MB layer to Docker HUB and then balena pull onto the device. That should be even closer to the regular deploy. Seemed to have gone through fine.