Fixed in PR #286.
Thanks to the new version of udocker the alpine containers can be used in AWS Lambda.

from scar.

The root of the problem is two-fold:

1. In the alpine:latest Docker image, the command-line utilities (say /bin/sh, /bin/ls are symlinks to /bin/busybox)
2. When activating the F1 execution mode in uDocker this problem is exhibited.

Steps to reproduce this issue:

udocker create --name=test-alpine alpine:latest
udocker setup --execmode=F1 test-alpine
udocker run test-alpine /bin/ls
Error: command not found or has no execute bit set:  ['/bin/ls']

Two options here:

1. Raise an issue in uDocker to see if they can fix this in uDocker.
2. Handle this in scarsupervisor.

Concerning (2), rewriting /bin/ls to /bin/busybox ls is not an option since the following error appears:

 udocker run test-alpine /bin/busybox ls

Warning: running as uid 0 is not supported by this engine
Warning: non-existing user will be created

##############################################################################

STARTING 819365c8-102d-34d1-9b7e-917afe328940

##############################################################################
executing: busybox
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __snprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __vfprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __chk_fail: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: rawmemchr: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __memmove_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __rawmemchr: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __strcpy_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __fprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __sprintf_chk: symbol not found

from scar.

For the sake of reference these are the udocker execution modes tested (W means successfully running the busybox command in a container out of the alpine:latest image executed in a local Linux box)

• P1: proot accelerated mode using seccomp --> W, but not a valid mode for AWS Lambda
• P2: proot accelerated mode disabled --> W, but not a valid mode for AWS Lambda
• F1: fakeroot starting executables via direct loader invocation --> Does not work
• F2: similar to F1 with protected environment and modified ld.so --> Does not work
• F3: fakeroot patching elf headers in libraries and executables --> Does not work
• F4: similar to F3 with support for newly created executables via
  dynamic patching elf headers of libraries and executables --> Does not work
• R1: runc rootless --> Does not work (not a valid mode for AWS Lambda)

from scar.

Filed an issue in uDocker: See: indigo-dc/udocker#76

from scar.

It looks like this may be an issue with Alpine only supporting Musl instead of GLIBC.

Tested the frolvlad/alpine-glibc but the very same error is obtained.

from scar.

For the sake of reference, tested another image with the same results:

udocker run mikesir87-aws-cli /usr/bin/python aws s3 ls

Warning: running as uid 0 is not supported by this engine
Warning: non-existing user will be created

 ##############################################################################
 #                                                                            #
 #               STARTING 8f630e53-1eb1-3f01-8207-b40cd2db1c1f                #
 #                                                                            #
 ##############################################################################
 executing: python2.7
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __snprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __vfprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __chk_fail: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: rawmemchr: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __memmove_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __rawmemchr: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __strcpy_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __fprintf_chk: symbol not found
Error relocating /home/ubuntu/.udocker/lib/libfakechroot-Ubuntu-16-x86_64.so: __sprintf_chk: symbol not found

from scar.

It seems that it a problem with the libfakechroot used. Using the alpine libfakechroot lib (linked to musl) it seems to work.

from scar.

@gmolto What if the alpine image I use already has glibc installed #28 (comment)

from scar.

@frezbo Hi. This was already tested unsuccessfully with the frolvlad/alpine-glibc image as you can see in the previous comments. No progress AFAIK in uDocker in this respect.

from scar.

Hi, seeking a little help here. Is your observation that solutions like alpine-pkg-glibc won't fix this kind of issues?

Error relocating ...: __snprintf_chk: symbol not found

Related: gliderlabs/docker-alpine#149

from scar.

Looks like version 1.1.3 of udocker provides some experimental code for Alpine according to: indigo-dc/udocker#76

from scar.

How can I try scar with version 1.1.3? I tried to modify the Dockerfile (and monkey-patched some files in the repo), but I can't get past this error:

Unable to find "/scar/src/providers/aws/cloud/lambda/udocker/udocker-1.1.3.tar.gz" when adding binary and data files.

The command '/bin/sh -c pyinstaller --onefile   --add-data="src/providers/aws/cloud/lambda/scarsupervisor.py:src/providers/aws/cloud/lambda"   --add-data="src/providers/aws/cloud/lambda/clients/apigateway.py:src/providers/aws/cloud/lambda/clients"   --add-data="src/providers/aws/cloud/lambda/clients/batch.py:src/providers/aws/cloud/lambda/clients"   --add-data="src/providers/aws/cloud/lambda/clients/lambdafunction.py:src/providers/aws/cloud/lambda/clients"   --add-data="src/providers/aws/cloud/lambda/clients/s3.py:src/providers/aws/cloud/lambda/clients"   --add-data="src/providers/aws/cloud/lambda/clients/udocker.py:src/providers/aws/cloud/lambda/clients"   --add-data="src/exceptions.py:src"   --add-data="src/utils.py:src"   --add-data="src/providers/aws/cloud/lambda/udocker/udocker-1.1.3.tar.gz:src/providers/aws/cloud/lambda/udocker"   --add-binary="udockerb:src/providers/aws/cloud/lambda/udocker"   --add-binary="/usr/bin/zip:src/bin"   --hidden-import=urllib3   --hidden-import=configparser   scar.py' returned a non-zero code: 1

from scar.

I created a branch called 'udocker-update' with the latest version of udocker.
This version solves the issue with alpine containers (at least the containers that I tested).
Please check the if branch works correctly with your containers.

from scar.