Qemu error when running gem install bundler on Docker with Apple Silicon M1

Question

When I run docker-compose up I'm getting the following error

qemu: uncaught target signal 6 (Aborted) - core dumped

This is happening when docker-compose is running the command bundle install

 => ERROR [5/6] RUN bundle install --jobs 4 --path .bundle/                                                                                                                  4.2s
------
 > [5/6] RUN bundle install --jobs 4 --path .bundle/:
#10 4.170 Fetching gem metadata from https://rubygems.org/.[BUG] Segmentation fault at 0x0000418f81e2b0
#10 4.177 ruby 2.3.4p301 (2017-03-30 revision 58214) [x86_64-linux]
#10 4.177
#10 4.177 SEGV received in SEGV handler
#10 4.178 -- Control frame information -----------------------------------------------
#10 4.178 SEGV received in SEGV handler
#10 4.178 SEGV received in SEGV handler
#10 4.178 qemu: uncaught target signal 6 (Aborted) - core dumped
#10 4.192 Aborted
------
executor failed running [/bin/sh -c bundle install --jobs 4 --path .bundle/]: exit code: 134

According to this github issue thread on docker's github – its a qemu related bug:

https://github.com/docker/for-mac/issues/5148#issuecomment-784992663

It mentions that qemu is the upstream component Docker use for running Intel (amd64) containers on M1 (arm64) chips. I checked the qemu repo fora related issues with no luck:

https://gitlab.com/qemu-project/qemu/-/issues?scope=all&utf8=%E2%9C%93&state=opened&search=uncaught+target+signal+6

In my docker-compose file I've passed the following config which helped me fix a couple M1 related issue on docker. But this isn't working for this specific issue.

web:
    platform: linux/amd64
    ...

This issue is happening on the following environment

• Macbook Pro M1 2020
• Mac OS Big Sur 11.2.3
• docker-compose version 1.29.1
• docker engine: 20.10.6
• docker desktop 3.3.3(64133)

I've also opened an issue on qemu repo as well: https://gitlab.com/qemu-project/qemu/-/issues/340

Any ideas on how to fix this?

Halbert

Answer 1

I found a hack. This is a way around that specific issue. I fixed it by adding the following line at the top of my Dockerfile.yml to update the image source which might have been outdated

FROM ruby:2.3.4

If you're experiencing a similar issue with another library I suggest looking for an image on Docker Hub https://hub.docker.com/ and making sure that they have support for the desired architecture: amd64 or arm64, etc...

---

For more information about the qemu error, you can follow this GitLab issue thread: https://gitlab.com/qemu-project/qemu/-/issues/340

Halbert

Answer 2

Just thought I'd add a discussion point since this is the first thing that comes up on Google.

We're using Fullstaq Ruby jemalloc as our base image and unfortunately there isn't an ARM build of Fullstaq Ruby. Changing our Dockerfile to use a Dockerhub base image appears to solve the issue.

Switching

FROM quay.io/evl.ms/fullstaq-ruby:2.7.6-jemalloc-buster-slim

to

FROM ruby:2.7.6-slim-bullseye

Seems to do the trick.

Invitation