RFC: Centrilized managment console

Dan Shearer dan at shearer.org
Wed Jun 4 11:24:36 UTC 2008


On Tue, Jun 03, 2008 at 08:00:31PM -0500, Nicolas Valcarcel wrote:
> I have been working on the blueprint of a centralized managment console
   :
> https://blueprints.edge.launchpad.net/ubuntu/+spec/ubuntu-centralized-services-administrator

I'm not sure how best to contribute, so I'll start with a few comments
here first.

Rationale
---------

I wonder if the Rationale section is maybe looking at the right things
from the wrong starting point.  To me the deeper analysis is:

    Ubuntu Server has no awareness of itself as a product.

Yast, webmin and the rest don't address this either.  Personally I'd be
delighted to stick with existing Ubuntu Server tools for managing
services (thanks, Debian, upstreams!) and just overlay a higher order of
understanding and control. Which, at our later option, we can make as
GUI as we like, or as is required.

There's a subtle point here that was only hinted at before, I can't
remember who made it. The good thing a lot of us see in the Microsoft
admin tools is that they have this higher order of understanding to some
degree. Not so much just that there is a GUI. And that is where I think
some of the debate on this list has been like ships passing in the
night, people not realising that the others are talking about different
things. I despite a mandatory GUI as much as the next Unix person. But I
recognise value in a network-centric management view, such as delivered
nicely by some GUI tools.

Outline Sketch Implementation
-----------------------------

Following is a sketch of a commandline tool ubuntu-server-admin.py that,
if it existed, would give me confidence that a useful admin tool could
be built on top of it. My tool would be interacting with existing Linux
and Debian management facilities, and would use a database. I have a
clear idea for how the database would work but that's detail.

u-s-admin --report --overview returns an XML summary file that says:
   name = X, otherwise known as Z
   services I'm running that matter to users are A,B,C
   the locations of my vital data are D, E, F
   the network services I depend on are G, H I
   the network servers I depend on are J, K, L
   the machines to which I log messages are M and N
   the machines monitoring me are O and P

(where I say 'machine' above it is likely 'CNAME' in reality to avoid
hard coding)

u-s-admin --report --depend-network-services would return:
   DNS server details, and their current status
   KDC server details and status
    :

u-s-admin --report --depend-network-servers would return:
   Server J: rsync for backup, on port X; and current status
   Server K: SQL server for webapp we're running; and current status
   Server L: web proxy for accellerator for Apache we're running; and current status

Given this level of awareness, next we need to configure these things.
The fact of this configuration would not be kept in the database, the
database would only be for the higher-level understanding. This would be
making calls to debconf or apachectl or whatever makes sense, and these
tools just manage state the same way they always did.

-- 
Dan Shearer
dan at shearer.org




More information about the ubuntu-server mailing list