docker mount ownership fixes #2162
Conversation
There was a problem hiding this comment.
If this works for everyone needing it, I am fine with merging
What does migrating to this new provision script look like? Does vagrant provision on an existing VM just work, or is vagrant destroy needed? (Just asking preemptively because this might come up after merging)
There was a problem hiding this comment.
Did a fresh vagrant up and provision with Docker. I took the bindfs route, because OWNER was 1000 (my uid on the host). Some errors occurred, I listed them in the review comments.
After fixing them, I did another fresh run and it seems to work as expected. I get this warning on every vagrant ssh though:
fuse: mountpoint is not empty
fuse: if you are sure this is safe, use the 'nonempty' mount option
I suppose this is because we re-mount in bashrc right? If so, I am fine with leaving it
(not this PR) One thing that was annoying was that npm ci took a couple of minutes because the puppeteer chrome download failed. For future reference:
npm ERR! code 1
npm ERR! path /opt/evap/node_modules/puppeteer
npm ERR! command failed
npm ERR! command sh -c node install.js
npm ERR! ERROR: Failed to set up Chrome r115.0.5790.102! Set "PUPPETEER_SKIP_DOWNLOAD" env variable to skip download.
npm ERR! Error: read ECONNRESET
npm ERR! at TLSWrap.onStreamRead (node:internal/stream_base_commons:218:20) {
npm ERR! errno: -104,
npm ERR! code: 'ECONNRESET',
npm ERR! syscall: 'read'
npm ERR! }
This also left me without tsc and scss, so the first ./manage.py run prompted me to install those. Aborting and running npm ci again gives the same error. With PUPPETEER_SKIP_DOWNLOAD=1 npm ci, I could work again
Yes. |
|
@Kakadus are there any additional requirements for this setup to work (besides what is specified in the readme)? I could not get it to work by simply using If this is just an issue with my setup, I am also okay with that. It does not break my own dev setup using libvirt/sshfs. But if there are some additional requirements, we should document those. |
It turned out to be a SELinux problem, unrelated to bindfs and docker (rootless) |
There was a problem hiding this comment.
Works for me on Ubuntu 22.04 both with docker and virtualbox. I don't like the error message about the mounting point not being empty (wrong errors are useless, spammy, send you off in the wrong direction when debugging -- imagine someone new where the setup doesn't work trying to find the actual error when they are presented with a wrong error on every SSH), but I won't block for it.
This should finally close #2095 using bindfs to map the ownership to 1042. This basically fixes all open points from #2105.