Docker Containers Restarting (132), Illegal instruction (core dumped)

Question

I'm consistently running into a show-stopper with alphanet and docker on ubuntu. I suspect it's something I consistently overlook when setting up. My docker containers only run for a second or two and then restart (132).

Some poking around leads me to believe they restart from panic just about right away.

I have tried Ubuntu 16.04. 18.04, virtualBox MV and vmWare and even AMD and Intel. 4-6 processors and 4-8GB RAM. The result is always the same.

Other docker containers run happily.

$ docker version
Client:
 Version:           18.09.2
 API version:       1.39
 Go version:        go1.10.6
 Git commit:        6247962
 Built:             Sun Feb 10 04:13:47 2019
 OS/Arch:           linux/amd64
 Experimental:      false

Server: Docker Engine - Community
 Engine:
  Version:          18.09.3
  API version:      1.39 (minimum version 1.12)
  Go version:       go1.10.8
  Git commit:       774a1f4
  Built:            Thu Feb 28 05:59:55 2019
  OS/Arch:          linux/amd64
  Experimental:     false

$ docker-compose version
docker-compose version 1.17.1, build unknown
docker-py version: 2.5.1
CPython version: 2.7.15rc1
OpenSSL version: OpenSSL 1.1.0g  2 Nov 2017

Here's an image in case it helps.

Here's a representative docker log of a container:

tezos@rob-VirtualBox:~$ docker logs alphanet_node_1
Current public chain: 2018-11-30T15:30:56Z-alphanet.
Local chain data: 2018-11-30T15:30:56Z-alphanet.
Configuring the node...
Illegal instruction (core dumped)
Current public chain: 2018-11-30T15:30:56Z-alphanet.
Local chain data: 2018-11-30T15:30:56Z-alphanet.
Configuring the node...
Illegal instruction (core dumped)
Current public chain: 2018-11-30T15:30:56Z-alphanet.
Local chain data: 2018-11-30T15:30:56Z-alphanet.
Configuring the node...
Illegal instruction (core dumped)
Current public chain: 2018-11-30T15:30:56Z-alphanet.
Local chain data: 2018-11-30T15:30:56Z-alphanet.
Configuring the node...
Illegal instruction (core dumped)
Current public chain: 2018-11-30T15:30:56Z-alphanet.
Local chain data: 2018-11-30T15:30:56Z-alphanet.
Configuring the node...

I'd be very thankful for any thoughts or suggestions.

Answer 1

I believe this is because of a build issue with some OCaml module that checks hardware architecture at build time 😳 This leaks into the Docker images which throws Illegal instruction when run a different arch.

You can read more about it here and here.

The solution is to build the Tezos Docker image yourself on the infra/hardware you want to run it. You can find a working Dockerfile here.

UPDATE

Added some declarative Dockerfile's here 👍
The Dockerfile-ubuntu is the one for you, also make sure to build on the hardware you need to run. Should work the same as the official images (using entrypoint.sh etc.) 😄

Answer 2

I recall encountering the same problem in alphanet months ago. I believe it turned out that the docker image had been compiled in a way that failed on some hardware configurations. I have not been able to find gitlab issues/commits for exactly the problem, but https://gitlab.com/tezos/tezos/issues/220 is related.