[racktables-users] Re: How do you fill your racks? (Was: Developers!!!!)

  • From: Michael C Tiernan <michael.tiernan@xxxxxxxxx>
  • To: racktables-users@xxxxxxxxxxxxx
  • Date: Tue, 8 Mar 2011 08:29:59 -0500 (EST)

----- Original Message -----
> From: "Denis Ovsienko" <infrastation@xxxxxxxxx>
> the rack is already divided into front, back and interior projections

I would like to point out that this has got to be the biggest sticking point 
that I find, how to wrap your brain around this projection.

To me, it is the one big problem I'm having. I need to know how to take the 
existing world that we have in the next room and map it to the RatTab view of 
things.

Maybe this is a good time for the discussion to take place? If we can collect 
some responses from people about how things are done by others, maybe some help 
docs can be created to enlighten those who are coming to RackTables?

Here's the world I see:
   I have a bunch (~90) of almost perfectly identical racks.

Each rack:
   Is X units high with a water cooling door (w/fans) on the back.
   Has a low speed switch (maint net) inside it at the top, only accessible 
from inside.
   Has a patch panel inside that terminates at the master switch (prod net)
   Has two PDUs going down each side for power.

Each server in the rack
   Is a long (~30") box usually 1 or 2 U high.
   Has two power supplies (although some have one now)
   Has three network connections. 1 direct, 2 bridged.

Almost every server is the same type with only minor variations.

I have one rack of patch panels (the terminus of the rack panels) next to a big 
Cisco switch that connects the production network to the outside world.

My progress at implementing RatTab is slow because I'm keeping three sets of 
docs about what goes where and how. (RatTab will eliminate all three and leave 
one new one.)

So, currently, I build my rack layout quickly and easily and then build an 
inventory of servers. This grows quickly. (Which is a feature enhancement I'd 
like to suggest, being able to break down a massive list of uninstalled objects 
(or installed ones) into logical chunks allowing an easier way to find things. 
Maybe tabbed pages?)

I then begin to install the objects into the racks. When I go to document the 
cables and connections, it seems that no matter how I think I should do it I'm 
doing it wrong. I use a server entry that uses all three entries, front, 
internal, and back but it doesn't feel right. The connecting of the power from 
the servers to the PDUs also falls flat for me.

One of the things I'm working on is creating a way to collect the system data 
into a file and then batch it up into the DB as the raw uninstalled objects to 
then be placed into the racks.

How do other people do *their* layouts?
-- 
    << MCT >>   Michael C Tiernan.
    Is God a performance artist?
    http://www.linkedin.com/in/mtiernan

Other related posts: