Hi,
I was looking into the Agent setup and configuration today and found out
that this is quit outdated.
All the documentation is still pointing to the cloud-setup-agent tool,
but do we still want that?
On my systems this tool seems to brake more then you want.
I'm working on fixing most of the bugs, but setting up the agent isn't
that hard at all.
1. Make sure your interfaces match you traffic labels
2. Fill the agent.properties (guid, resource host, private nic, public nic)
3. Start the agent
There is however one thing I don't like. The agent is overwriting it's
agent.properties file with various own lines, mangling anything you
might have written to it.
Admins might deploy their agents with Puppet or Chef and those tools
usually go crazy when files change without them noticing it.
Do we really need to write to this file? Shouldn't the agent just start
and whenever some property is not set use a default value?
The agent for example generates a UUID for local storage and stores it
in agent.properties. Should it? Shouldn't it simply complain if that
value is not set and let this value be set by cloud-setup-agent or by
the admin manually?
I personally don't like daemons who start touching my configuration and
modifying files without me knowing it.
To sum it up:
I think setting up an agent should be able by just providing a
agent.properties and nothing more. Start the agent and go online.
No need for the cloud-setup-agent tool imho. This is a black magic box
which does all kinds of things which should actually be documented.
Wido
