[u-u] Asset tracking...

[Anthony de Boer]

Colin McGregor wrote:
> In summary track everything of ANY consequence about every server.
> Further be able query by all data fields, (ie: list all servers
> running "Quake II" or in cage #4, rack #102, etc).

I've done that sort of thing before (though the code was owned by that
employer), but in outline it's a mix of:

 - information that can be generated by a script running on each box;
 you want to deploy a reasonably-up-to-date version of it and somehow
 collect output.  I used ssh and collected stdout, Alan notes he has
 the box email it from cron, there will be other ways.  Ultimately you
 collect all the outputs, and check for stale ones or decommissioned
 machines.  Running perl over ssh on the remote host with the current
 script on stdin from your workstation is one way of not having to
 install it there first without worrying about not getting the latest
 thing.  

 - information that has to be collected manually onsite.  In my case the
 major thing was a file per site with pairs of deviceID-portID, like
 "calvin-power ups03-5" and "calvin-eth0 switch5-04", indicating that the
 relevant kind of cable linked them.  I had a script that was smart
 enough to tell them apart and generate three graphviz pictures of power,
 ethernet, and "other" wiring per site.  You may also need to record
 model and serial numbers, but we had all generic whitebox rackmounts,
 and the only serial numbers we had were the random numbers we printed
 on the labelmaker when we had to pull one from 151 Front and security
 there would only be happy if they could record serials of all equipment
 being removed.

 - information from DNS

and then it's just a matter of a perl script that sucks it all in and
regurgitates reports in whatever forms you need.  Use python or feed
all the data into SQL and generate reports from that if you want.

-- 
Anthony de Boer