Consider publishing a statically complied binary with GitHub releases

[chrishoage]

Today I deploy dozzle though the docker image in agent mode on to multiple hosts. However, this collector type service is the only of its kind as a docker image.

I already maintain komodo-periphery and telegraf though ansible using systemd as a process supervisor. This is ideal for me as there are no issues surrounding ensuring the docker socket is passed though to the image (something I try to avoid if possible) along with being able to bind directly to the hosts tailscale tunnel device giving me robust ACL permissions of what is allowed to access these services over my tailnet

I know and understand how to achieve all of this though dozzle in a docker container, however this is a lot of extra steps for the agent, which just reads the socket and forwards the information to the dozzle server (which runs in a management compose stack)

Having the ability to have an ansible role for my docker hosts which keeps dozzle agents up to date with out the need to manage the docker deployment of the agent would be ideal.

No worries if this is out of scope for this project - managing the dozzle agent deployment though docker isn't a huge problem - but being able to make it the same as the other host operators would be ideal for my workflow.

[amir20]

I feel like your use case is unusual. Long time ago I had Dozzle binaries deployed but naturally almost all people were using Docker because to use Dozzle you most likely have Docker.

I know a few people who download the container and use docker cp to pull out the binary.

No worries if this is out of scope for this project - managing the dozzle agent deployment though docker isn't a huge problem - but being able to make it the same as the other host operators would be ideal for my workflow.

I don't mind publishing the static binary as well. I do it for https://github.com/amir20/dtop. However, I would like to understand your use case better. Can you tell me more? What are the advantages of running the raw binary? Isn't it easier to check for updates with Docker?

[chrishoage]

I feel like your use case is unusual.

It certainly is!

However, I would like to understand your use case better. Can you tell me more?

Absolutely! I'm in the process of breaking up my 2 monolith docker compose file that runs 51 containers on one host and 9 containers on another host. I currently run dozzle on the large host and dozzle agent on the smaller host. This works just fine.

However this process includes switching to Proxmox and splitting the 51 container monolith compose into a few different VMs that are organized by purpose.

To power this though CI I am using komodo with komodo periphery, where the periphery is downloaded from github releases and installed though ansible. Other tools installed though ansible the same way (download from github releases, installed and set up systemd units) are telegraf and rclone.

My goal for these docker hosts in VMs is to make them homogeneous using the same ansible roles I have created for this purpose.

I desire to keep all of these "host tools" (things that need to interact with docker like komodo periphery and dozzle, or read system resources like telegrf) out of docker as these tools exist to support the docker hosts.

What are the advantages of running the raw binary?

I find that these class of tools are typically easier to maintain when they are not in a container (komodo periphery for having full access to the host filesystem, telegraf for reading /proc and friends, dozzle for accessing /var/lib/docker.sock). The container for these kinds of tools just gets in the way - the host is what we actually care about.

For the hosts that the dozzle agent needs to run, komodo doesn't need to care about "deploying" it as apart of the stacks on that particular host, just the containers for the VMs purpose.

Isn't it easier to check for updates with Docker?

For most people I would agree docker is far superior. For how I desire to set up my environment my goal is to leverage renovate to keep komodo, dozzle and telegraf up to date, with automatic ansible deployments from CI (access granted via Tailscale SSH and ACLs)

I would expect the installation method described above to be "unsupported".

I would also very much understand if producing these binaries is something that is undesirable from your point of view. In that situation I would likely use ansible to deploy the dozzle docker container directly running in agent mode. There is just benefit in having all of these "docker host support" tools to all be installed, updated and manged the same way (ansible installs the binary from github releases, uses systemd to supervise)

I do it for https://github.com/amir20/dtop.

What a neat looking project. This seems useful for the number of docker hosts I am on the path to running.

[amir20]

Thanks for the details. My only concern is that I would have compile twice. Once inside the container and once outside. There might be a tiny chance that the two result in different binaries. Ideally, I would like to use docker cp and just archive those. I'll play around with it to see how easy it would be.

But for now I would agree that it is probably unsupported.

[chrishoage]

If it's too much trouble don't worry about it. My request here is truly "if it's easy"

Thanks for considering it!

[amir20]

I tried to do this. I was some what successful but the changes became to difficult and it wasn't working. So for now, I don't think I'll do it.

[chrishoage]

Not a problem at all, thank you so much for trying! I obviously have other options available to me

Thank you very much for considering it!