Offline Agent distribution.

[Smoke]

I need to distribute zabbix agent at a large scale on my servers (RHEL 8), and the requirement is to have all zabbix files deployed on a custom folder, /opt/zabbix, meaning all binaries, logs and config files will be there.
That doesn't seems to be a problem, I can download the pre-compiled tar.gz package and do whatever I want without problems.

My specific concern is what's the best way to deal with agent daemon and startup script, in particular, the compiled binaries do not distribute with systemctl scripts. Is there any workaround on this?
We are using RHEL 8.x and we do not need to support legacy old init.d scripts.

If on the other hand i go with the RPM installation, which is something I can do also, then I will not have the functionality to deploy everything into a custom folder, and logs/config/binaries will be installed on default OS location.

What everyone's recommendation?

Worth mentioning that the agents will not have internet access so I cannot rely on the default yum repo install neither, I have to distribute it.

EDIT: In the past, 5 years ago, I used a custom chkconfig script from vrtareg , but I doubt that still works with systemctl.

[markfree]

Is there any particular reason to use that custom folder?

My recommendation would be to create a Zabbix repository mirror in house and point your hosts to that mirror repository.
I believe Ansible is a possibility for deployment with either RPM or pre-compiled packages.

[Smoke]

There are several reasons indeed, but most importantly company policies and permissions to other users to better control config, upgrades and plugins.
Yeah , well, a mirror repo just for an agent seems excessive, even worst when you consider that not all the servers will be able to reach a single repo, so i need to create like 4 mirrors on diff networks.

How does the RPM install the systemctl script after all? I can just recreate the file/commands it performs.

[markfree]

I see what you mean, but I don't find a mirror excessive for a "large scale" deployment.

[Markku]

How does the RPM install the systemctl script after all? I can just recreate the file/commands it performs.

That's how I would do it in your situation: see what the RPM actually installs and then do the same (with necessary adjustments).

I believe there is a command to view and extract the files in .rpm file.

Markku

[markfree]

Will you install all agents by hand or use a sort of central hub to push them?

Also, you could just install the agents pre-compiled package anywhere and create a Systemd service for it.
For an existing installation, you can type systemctl cat zabbix-agentd.service. There you may find a good reference to the service parameters.