I’m trying this on Ubuntu 22.04 Rust’s cargo install seems to keep creating permission problems between what I have to install, compile and what gets published in the cargo “registry”, which causes issues at runtime when I run as lemmy:lemmy through systemctl.

If I run: cargo install lemmy_server --target-dir /usr/bin/ --locked --features embed-pictrs as a non-root user, I get permission denied issues with /usr/bin/.future-incompat-report.json and /usr/bin/release

If I run the build as a root user, and then manually copy the binaries to /usr/bin and chmod them to lemmy:lemmy, then try to run as lemmy:lemmy, it appears the binary is trying to access some “registry” files in /root/.cargo/registry (for which of course it does not have permissions.)

How do I fix this?

  • RoundSparrow
    link
    111 months ago

    I am stuck at this same point!

    We need some help here…

    • @KIM_JONG_JUICEBOXOP
      link
      1
      edit-2
      11 months ago

      So I am able to build from the git repo.

      Had to remove rust from apt (v 1.65)

      And install rust from rustup (v 1.70)

      That fixes the build, but now I still get this db migrations error on startup.

      Jun 09 15:47:37 ip-x systemd[1]: Started Lemmy - A link aggregator for the fediverse.
      Jun 09 15:47:37 ip-x lemmy_server[15239]: 2023-06-09T15:47:37.238544Z  INFO lemmy_db_schema::utils: Running Database migrations (This may take a long time)...
      Jun 09 15:47:37 ip-x lemmy_server[15239]: thread 'main' panicked at 'Couldn't run DB Migrations', crates/db_schema/src/utils.rs:165:25
      Jun 09 15:47:37 ip-x lemmy_server[15239]: note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
      Jun 09 15:47:37 ip-x systemd[1]: lemmy.service: Main process exited, code=exited, status=101/n/a
      
      

      I know it can connect to the db because I see it created some tables already.

      I wonder if this is related to this warning regarding schema, which I see during the build.

      warning: /root/git-clones/lemmy/crates/api_common/Cargo.toml: `default-features` is ignored for lemmy_db_schema, since `default-features` was not specified for `workspace.dependencies.lemmy_db_schema`, this could become a hard error in the future
      
      • RoundSparrow
        link
        3
        edit-2
        11 months ago

        > thread ‘main’ panicked at ‘Couldn’t run DB Migrations’, crates/db_schema/src/utils.rs:165:25

        I fixed this by issuing:

        sudo -iu postgres psql -c “ALTER USER lemmy WITH SUPERUSER;”

        Obviously this has database security implications, but at least it identified that was why DB Migrations is failing.

        EDIT: I found the developers of Lemmy seem to have identified the cause for this, it’s just a couple SQL statements: https://github.com/LemmyNet/lemmy/pull/2983/commits/29c4144e61e97e895fb7eb37d2c257c8520fd6a6

        EDIT 2: The developers are currently treating this as a documentation bug: https://github.com/LemmyNet/lemmy-docs/issues/201

        • @KIM_JONG_JUICEBOXOP
          link
          211 months ago

          Thank you!

          So the git based build instructions from @Grouchy@lemmy.grouchysysadmin.com works.

          What was missing for me was then the rust version and the full list of sql commands.

          1. Rust version
          
          Remove rust installed from Ubuntu apt repo (v 1.65)
          Install rust from rustup (v 1.70)
          
          2. Postgres commands
          
          sudo -iu postgres psql -c "CREATE USER lemmy WITH PASSWORD 'db-passwd';"
          sudo -iu postgres psql -c "CREATE DATABASE lemmy WITH OWNER lemmy;"
          sudo -iu postgres psql -c "ALTER USER lemmy WITH SUPERUSER;"
          
          
          
          • RoundSparrow
            link
            211 months ago

            The lemmy-ui steps, the next part - are pretty outdated. It has instructions to install NodeJS 12.x, which is really far back. I have no idea which version the Docker install is using…

            The latest version of Node is 20

            • @KIM_JONG_JUICEBOXOP
              link
              2
              edit-2
              11 months ago

              Before I even get there, the instructions say I should be able to run

              curl localhost:8536/api/{version}/site
              

              And it should return some json structure.

              But I see this returning 404

              lemmy_server[41511]: 2023-06-09T17:02:35.296098Z  INFO actix_web::middleware::logger: 127.0.0.1 "GET /api/0.17.3/site HTTP/1.1" 404 0 "-" "curl/7.81.0" 0.000074
              

              Is that the correct way to specify the version?

              Is there a better log than journalctl which would actually tell me what REST endpoints it is using?

              • RoundSparrow
                link
                211 months ago

                Is that the correct was to specify the version?

                No, I had to dig around too, the correct command is:

                curl localhost:8536/api/v3/site

                • @KIM_JONG_JUICEBOXOP
                  link
                  111 months ago

                  Thank you, again!

                  How was I to know {version} in the REST endpoint path was supposed to be “v3”?

                  Any luck with the lemmy-ui install?

                  • RoundSparrow
                    link
                    1
                    edit-2
                    11 months ago

                    I knew it was v3 because I saw a webapp failure on one of the Lemmy instance (lemmy.ml) I was using ;) We need to update those install documents.

                    It seems right now I can get a client to load, but websockets aren’t working. Others on Reddit reported the same issue: https://old.reddit.com/r/Lemmy/comments/142sszq/spinning_up_lemmy_websockets_dont_connect/

                    EDIT: Ok, I solved the websocket problem by using proper nginx config that the instructions said. I was confused about the domain names with the certification generation as I already had some nginx certs for other subdomains in place.