The metrics container (prometheus_wireguard_exporter) fails to run on arm64

[codestation]

Describe the bug

The metrics container currently fails on arm64 so the pod cannot start.

The container log shows the following:

thread 'main' panicked at library/alloc/src/raw_vec.rs:534:5:
capacity overflow
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

And passing the required env variable to the pod:

thread 'main' panicked at library/alloc/src/raw_vec.rs:534:5:
capacity overflow
stack backtrace:
   0: rust_begin_unwind
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/std/src/panicking.rs:597:5
   1: core::panicking::panic_fmt
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/core/src/panicking.rs:72:14
   2: alloc::raw_vec::capacity_overflow
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/alloc/src/raw_vec.rs:534:5
   3: alloc::raw_vec::RawVec<T,A>::allocate_in
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/alloc/src/raw_vec.rs:177:27
   4: alloc::raw_vec::RawVec<T,A>::with_capacity_in
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/alloc/src/raw_vec.rs:130:9
   5: alloc::vec::Vec<T,A>::with_capacity_in
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/alloc/src/vec/mod.rs:670:20
   6: alloc::vec::Vec<T>::with_capacity
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/alloc/src/vec/mod.rs:479:9
   7: std::sys::unix::args::imp::clone
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/std/src/sys/unix/args.rs:146:28
   8: std::sys::unix::args::imp::args
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/std/src/sys/unix/args.rs:131:22
   9: std::sys::unix::args::args
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/std/src/sys/unix/args.rs:19:5
  10: std::env::args_os
             at ./rustc/79e9716c980570bfd1f666e3b16ac583f0168962/library/std/src/env.rs:794:21
  11: <core::pin::Pin<P> as core::future::future::Future>::poll
  12: tokio::runtime::scheduler::current_thread::CurrentThread::block_on
  13: prometheus_wireguard_exporter::main
note: Some details are omitted, run with `RUST_BACKTRACE=full` for a verbose backtrace.

I also tried to compile the binary in debug mode and failed on this crate:

https://github.com/clap-rs/clap/blob/f8e9211e38d19f060980a6cb10f3fc8a2735c5b2/src/builder/command.rs#L548

To Reproduce
Steps to reproduce the behavior:

1. kubectl apply -f https://github.com/jodevsa/wireguard-operator/releases/download/v2.0.17/release.yaml
2. Create a server and a peer.
3. The pod reports CrashLoopBackOff, with the metrics container stuck in a crash loop.

Expected behavior
The metrics container should start.

Additional context
I also ran the container on docker with the same result

docker run --rm --entrypoint /usr/local/bin/prometheus_wireguard_exporter ghcr.io/jodevsa/wireguard-operator/agent:v2.0.17

If i copy the prometheus_wireguard_exporter from the container to the host (Ubuntu 22.04 LTS) and run it, then it executes without errors, so the problem is with the alpine image. I also switched the final image from alpine:3.18 to debian:bookworm and it ran without errors.

Possible fixes:

1. Use a final image based on glibc instead of alpine (debian, for example).
2. Change the compilation targets on the metrics binary so is compatible with alpine.
3. Workaround: disable the metrics container (cannot find a way to do it without editing the deployment by hand).

[jodevsa]

Thanks @codestation. I'll try to look into this. I'm fine with removing those metrics. I'm not aware or anyone using them. The other alternative that I would prefer is to implement those metrics in the agent. so rewriting them in golang. This shouldn't be that much work.

I'm doing a language course along side working my job. So I cannot promise this will land soon. I'm also open for PRs with whatever soloution you choose from the suggestions you proposed.

Thanks,
Subhi

[codestation]

I would be using those metrics once i finish my grafana setup so please keep it, maybe add some option to disable it per wireguard server. The reimplementation in Go sounds a good idea, as i cannot read rust code at all.

I pushed a simple PR with a image change. The image gets a little bigger but the controller now starts correctly (i currently have 2 days of uptime on my ARM Ampere instance).

[jodevsa]

Thanks @codestation for fixing this. PR is merged and the functionality is available in the latest release. Please close the issue if the problem is now resolved

[codestation]

Thank you, just updated and everything is running fine with the latest agent image.