Hi,
I've been working on the documentation for setting up an Hypervisor with
KVM and committed[0] some new docs about his.
My main problem with cloud-setup-agent is that it is touching all kinds
of files without notifying the system administrator about this.
It happened to me a couple of times that cloud-setup-agent broke the
network connectivity to my machine because it started touching my
network configuration.
I'm trying to document the steps you have to perform to setup a
hypervisor, basically these are the things cloud-setup-agent is doing
right now.
In my opinion cloud-setup-agent should only check if you did it right
and nothing more.
The steps for a hypervisor:
1. Install OS
2. Install packages
3. Configure packages
4. Configure network bridges
5. Run cloud-setup-agent
cloud-setup-agent then does:
1. Verify if packages are installed
2. Verify if bridges are present
3. Generate agent.properties file
I think the whole setting up packages and configuring the hypervisor is
up to the admin, he/she(!) can do this manually or use Chef/Puppet for
this task.
Also, to take it a step further, there is no need to run
cloud-setup-agent, you can generate a agent.properties yourself.
The current cloud-setup-agent is writting in Python, but this can be
done in Bash pretty easy if it's not doing anything like touching
configuration.
The requirements ofcourse are:
1. Make sure every step is documented
2. Make sure this hits the ChangeLog :)
Adding hosts through the WebUI should also go much smoother, since that
also invokes cloud-setup-agent through SSH.
How do we feel about this? If we agree on this I'll work on documenting
setting up Hypervisors next week and rewriting cloud-setup-agent.
Wido
[0]:
https://git-wip-us.apache.org/repos/asf?p=incubator-cloudstack.git;a=commit;h=b6a610e2db5798ae69cfd2cc8a9d7f85617f3926
