Layer not working on Amazon Linux 2

[markusahlstrand]

Thanks for publishing this layer!

I tried using it in a node lambda, but they run on the newer VM (amazon linux 2) and it doesn't seem to work.. I'm no good with python, but as far as I understand the X11 packages are not available on the newer VM's. Would it be possible to add the missing packages to the layer?

Thanks!

[gipsh]

When you execute the binary /opt/ebook-convert what error is throwing ?

I need to try it myself on the new vms.

[markusahlstrand]

This is the error I get when I try to convert a file from epub -> mobi

Command failed: ebook-convert \"/tmp/1599059003173.epub\" \"/tmp/1599059003173.epub.azw3\"
Traceback (most recent call last):
 File \"site.py\", line 77, in main
 File \"site-packages/calibre/ebooks/conversion/cli.py\", line 401, in main
 File \"site-packages/calibre/ebooks/conversion/plumber.py\", line 1275, in run
 File \"site-packages/calibre/ebooks/conversion/plugins/mobi_output.py\", line 312, in convert
 File \"site-packages/calibre/ebooks/mobi/writer2/resources.py\", line 12, in <module>
 File \"site-packages/calibre/ebooks/mobi/utils.py\", line 13, in <module>
 File \"site-packages/calibre/utils/img.py\", line 17, in <module>
ImportError: libGL.so.1: cannot open shared object file: No such file or directory

It seems that other people has run in to similar issues when trying to get calibre on Synology devices and to my understanding this is due to missing X11 packages.

[gipsh]

Did you try installing the amazon linux 2 vm on your host and run the layer locally?
Probably there are more missing libs.

[markusahlstrand]

I'll try settings up an image on ec2. Guessing that the Amazon Linux 2 AMI (HVM) is the same image they use in the lambdas?

[markusahlstrand]

I got an ec2 image up and running and have check the following:

• The automatic installer of calibre, the v4 tarball and the v3 tarball all fail with the same error.
• The prerequisites xdg-utils, wget, xz-utils and python ≥ 2.6 all seem available
• GLIBC and gcc versions seems fine.

The error I get is:
Traceback (most recent call last):
File "site.py", line 75, in main
File "site-packages/calibre/init.py", line 23, in
File "site-packages/calibre/startup.py", line 156, in
File "locale.py", line 545, in getdefaultlocale
File "locale.py", line 477, in _parse_localename
ValueError: unknown locale: UTF-8

[markusahlstrand]

Ok.. so it seems to simply be an issue with the default locale being wrong.

Tried what was suggested in this post and it seems to work in my VM: https://www.mobileread.com/forums/showthread.php?t=286960

I'll try adding these environment variable to my lambda code and hopefully the layer works fine.

[gipsh]

I tried myself to run the layer on amazon linux 2 vm, and got the same error for libGL.
Once installed i still throwing errors for Qt.
Is going to take some time to make it work and im not sure if could be done without hitting the 250MB limit for the layer.

[markusahlstrand]

Ok, I installed Qt as part of chrome headless when debugging (https://intoli.com/blog/installing-google-chrome-on-centos/), and maybe that fixed the first part of the issue.. I'll wipe the server and try it again.

[markusahlstrand]

Would custom runtimes be a way to get around this issue: https://aws.amazon.com/about-aws/whats-new/2020/08/aws-lambda-supports-custom-runtimes-amazon-linux-2/

[evangow]

@markusahlstrand did you ever find a workaround to these issues?

[markusahlstrand]

Nope, we are running a docker container as a lambda instead which works well