Docker builds started failing

[joachimvh]

Description

At some point our Docker builds suddenly started failing. This seems to have coincided with when then lts-alpine image changed from Node v18 to v20: https://github.com/nodejs/docker-node/blob/18b91b947173898d92dd736bcb334ab3bc4486c0/versions.json#L59 . This seems to be a known issue: nodejs/docker-node#1798 .

Testing locally the build succeeds for me when using the v18 images, but fails when using v20 with the following error:

473.3 npm ERR! code ECONNRESET
473.3 npm ERR! errno ECONNRESET
473.3 npm ERR! network request to https://registry.npmjs.org/yocto-queue/-/yocto-queue-0.1.0.tgz failed, reason: Client network socket disconnected before secure TLS connection was established
473.3 npm ERR! network This is a problem related to network connectivity.
473.3 npm ERR! network In most cases you are behind a proxy or have bad network settings.
473.3 npm ERR! network 
473.3 npm ERR! network If you are behind a proxy, please make sure that the
473.3 npm ERR! network 'proxy' config is set properly.  See: 'npm help config'
473.9 
473.9 npm ERR! A complete log of this run can be found in: /root/.npm/_logs/2023-10-31T13_06_21_050Z-debug-0.log
--------------------
   9 |     
  10 |     # Install and build the Solid community server (prepare script cannot run in wd)
  11 | >>> RUN npm ci --unsafe-perm && npm run build
  12 |     
  13 |     
--------------------
ERROR: failed to solve: process "/bin/sh -c npm ci --unsafe-perm && npm run build" did not complete successfully: exit code: 1

Will switch the Docker image to Node v18 to fix this, but we will need to switch back at some point when this is fixed.

[joachimvh]

Docker builds have started failing again, see https://github.com/CommunitySolidServer/CommunitySolidServer/actions/runs/7125529001/job/19401847448. Error seems to be the same

------
 > [linux/arm/v7 build 4/4] RUN npm ci --unsafe-perm && npm run build:
279.4 npm ERR! code ECONNRESET
279.4 npm ERR! errno ECONNRESET
279.4 npm ERR! network request to https://registry.npmjs.org/yup/-/yup-1.3.2.tgz failed, reason: socket hang up
279.4 npm ERR! network This is a problem related to network connectivity.
279.4 npm ERR! network In most cases you are behind a proxy or have bad network settings.
279.4 npm ERR! network 
279.4 npm ERR! network If you are behind a proxy, please make sure that the
279.4 npm ERR! network 'proxy' config is set properly.  See: 'npm help config'
279.6 
279.6 npm ERR! A complete log of this run can be found in: /root/.npm/_logs/2023-12-07T08_29_42_734Z-debug-0.log
------
Dockerfile:11
--------------------
   9 |     
  10 |     # Install and build the Solid community server (prepare script cannot run in wd)
  11 | >>> RUN npm ci --unsafe-perm && npm run build
  12 |     
  13 |     
--------------------
ERROR: failed to solve: process "/bin/sh -c npm ci --unsafe-perm && npm run build" did not complete successfully: exit code: 1

[joachimvh]

Had some look into this. The general problem just seems to be the slowdown of qemu when building for arm platforms. Drastically increasing the time before npm timeouts during install does not solve the problem. Same with yarn. I did get it working with pnpm (which needs rdf-js and @types/koa to be added to the package.json). Still only on the node 18 image though, not on 20.

Might just have to disable arm builds until there is a decent solution for this.

[smessie]

Could this new release be of help? https://github.blog/changelog/2024-01-30-github-actions-introducing-the-new-m1-macos-runner-available-to-open-source/

[joachimvh]

I'll be honest, I don't know enough about Docker and/or mac to answer that question. 😅 But if this can make it so we can build Docker images for that platform directly without using qemu, it might.