Add dockerfile #6
Conversation
Dockerfile
Outdated
| RUN python3 ./init.py -b /nucleus/build/nucleusd -c /root/.nucleus --overwrite | ||
| COPY addrbook.json /root/.nucleus/config/addrbook.json |
There was a problem hiding this comment.
Not every use case requires init.py to be run. We should provide a plain nucleus daemon and leave the rest to the users. How about simply copying init.py (without running) into the container?
Also addrbook.json file can be moved into https://github.com/KomodoPlatform/nucleus-assets repository.
There was a problem hiding this comment.
copying init.py into container will mean needing to add some extra py related deps, and it wont work unless we also copy the app.toml template.
There was a problem hiding this comment.
Right, maybe just copy the nucleusd binary and leave rest to the users. You can easily mount files/directories inside the container anyway.
As we providing the bare container for nucleus, I would pick one of debian images (like 1.22.1-bookworm) just to make install/configure parts of custom workflows easier for people.
Dockerfile
Outdated
| COPY --from=builder /nucleus/build/nucleusd /usr/bin/nucleusd | ||
| COPY --from=builder /root/.nucleus /root/.nucleus | ||
| # COPY --from=builder /nucleus/nucleus_conf.yaml /root/.nucleus/config.yaml |
There was a problem hiding this comment.
We just need nucleusd and init.py from the builder.
Dockerfile
Outdated
| RUN apk update && apk add --no-cache make git gcc libc-dev linux-headers python3-dev | ||
| WORKDIR /nucleus | ||
| COPY . /nucleus | ||
| RUN rm -rf .git && git init |
There was a problem hiding this comment.
VCS stamping fails if there is no .git folder, this was a workaround. It is because the .git is a file, not a folder when its a submodule.
There was a problem hiding this comment.
Setting GOFLAGS=-buildvcs=false env should disable VCS.
There was a problem hiding this comment.
it does, but then later when it looks for the version stamp which is required, it fails. I think this is part of the "daemon version validation" consensus
There was a problem hiding this comment.
I think this is part of the "daemon version validation" consensus
I wonder if initializing fresh git context will cause any problem. Maybe there's a better solution instead of removing the actual git state?
Co-authored-by: Onur Özkan <[email protected]>
Co-authored-by: Onur Özkan <[email protected]>
|
I've removed the dockerfile and the addrbook.json files. These are only needed for a specific use case and can be dealt with or accommodated as required elsewhere. The other branch I am working on does not include nucleus as a submodule https://github.com/smk762/nucleus-explorer/tree/v0.46.7-local |
I am happy to merge this PR without them, did you push the changes (they still exists in the PR)? |
argparsetoinit.pyscript to allow running non-interactive.addrbook.jsonto help bootstrap nodesExplorer is currently mostly functional, though it appears to be not sunning on the same chain as the existing nodes, even with the
addrbook.jsonfile. I think this might be by design - VCS stamping possibly enforces unmodified binaries for joining the network. If so, we probably need to set an update height using whatever mechanism is used here.The explorer is otherwise functional.
If it is preferable to not update the version yet, I'll create a new branch of explorer without the nucleus container and deploy/attach on an existing nucleus server to serve in the meantime anyway.