-
Notifications
You must be signed in to change notification settings - Fork 167
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Preliminary Alpine support for CI #437
Conversation
I'd keep striving to get Alpine 3.4 working tbh – it adds a bunch of very useful stuff for ppl in environments like Kubernetes and it'd be easier to deal with just one version of Alpine. Node.js definitely builds on it, I guess it's just a matter of finding out what's going on in this env. I know very little about the nodejs build environment, so it'll take some digging to figure out how things are setup |
|
||
VOLUME [ "/home/{{ server_user }}/" ] | ||
|
||
CMD su {{ server_user }} -c '\ |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
I explored the tests a while ago and found some reasons they were failing: |
Two of the failures (parallel/test-net-better-error-messages-port-hostname and parallel/test-net-connect-immediate-finish) are |
I'd also be interested to see if, instead of having to build inside of Alpine, you can just build with musl instead of glibc – technically I think that should work. (I believe this is what rust does, for example) |
(actually, that idea doesn't really address the testing-in-Alpine aspect, so it's probably of only limited value) |
Updated to Alpine 3.4 thanks to @mhart's pointer. Here's a run on Only 3 failures: parallel/test-net-better-error-messages-port-hostname |
@rvagg the |
great, procps has done it, updated Dockerfile and we're down to just the |
2366713
to
04dea0c
Compare
It's not possible to build Alpine binaries in CentOS unless you use a cross compiler. CentOS (and almost any other generic linux) uses |
So this is a continuous integration part (point 3 in #301 (comment)), right? |
I think we need bash as requirement |
I guess its worth mentioning that we now have full alpine passes through both docker instances and joyent's virtualization. |
We should probably update to Alpine 3.5. There's also a few new Ansible modules for managing (building/running) docker containers we can replace the shell commands with. I see both as optional before merging though. |
Alpine 3.6 is released, FWIW. |
alpine is in the new ansible/ stuff so I'm assuming it's good there, I haven't used it myself though |
Is there a generic issue for tracking a build of node for alpine? Or should I just watch #989? |
New version of this in #992, that's what's active in CI now with Alpine 3.4, 3.5 and 3.6 running in node-test-commit-linux. I've taken the old test-digitalocean-ubuntu1604_docker_alpine34-x64-1 host out of rotation, it was set up according to this PR and has been running our Alpine 3.4 tests till now. |
This adds a config for Ubuntu 16.04 + Docker that builds and manages an Alpine Linux 3.3 container which has the Jenkins slave running inside it.
After some digging I discovered that the preferable way to run Node in Alpine is to build it in alpine rather than reusing binaries that we build for generic Linux. Initially I started down the track of emulating our nodejs.org binaries by building in a CentOS 5 container before transferring to the Alpine container for testing. I scrapped that and went for everything in Alpine, assuming that we're going to ship images with binaries created inside them specifically for Alpine.
I'm using 3.3 because I had trouble getting 3.4 to run. It doesn't seem to like something about the combination of the mount point and the
CMD
and gives a weirdsh
error (su: can't execute ',,,:/home/': No such file or directory
if anyone wants to chase this up).It's temporarily in CI, you can see an example run @ https://ci.nodejs.org/job/node-test-commit-linux/nodes=ubuntu1604_docker_alpine34-64/3796 (the host is misnamed btw, it's "alpine34" but should be "alpine33").
Failures that we have are:
If we don't have a path to resolving these fairly quickly then we should remove it from CI so collaborators don't get upset at all their runs failing. ATM I don't think we have a way of marking these tests flaky just for Alpine, it's done by operating system, not distro.
Where to next?
/cc @nodejs/docker @nodejs/testing @mhart