NetNews Usenet Archive 1992 #30/NN_1992_30.iso/spool/comp/unix/question/14751

Internet Message Format | 1992-12-16 | 1.7 KB

Path: sparky!uunet!usc!zaphod.mps.ohio-state.edu!rphroy!ilium!gdls!mccabe
From: [e-mail] (Harold McCabe)
Newsgroups: comp.unix.questions
Subject: UNIX and Accounting
Keywords: server processes, background jobs, charge back
Message-ID: <[e-mail]>
Date: 16 Dec 92 22:45:46 GMT
Organization: Computer Sciences Corp. @ General Dynamics
Lines: 26

I manage a large number of UNIX workstations for a defense contractor.
The costs of the workstations go against the contracts that the
contractor has.  A single workstation can be used on several contracts.

To handle this, I have a "charge back" system which runs at login time
and prompts the user for the contract identifier.  The contract number
is appended to the most recent record in the wtmp file and written to a data
file.  Later, the wtmp file (via last) is correlated with the record in the
data file.  The result is a record that contains the user, the login time,
the elapsed time, and the contract id.  These resultant records are summarized
and used to determine how the workstations costs are allocated to contracts.

The problem I have is that UNIX wtmp accounting tracks only login sessions.
Background jobs, and processes that are servers are not tracked.  rsh's are
not tracked either.  This kind of work is an increasingly large % of the
load on the workstations, yet I am not tracking it.  UNIX has some
additional abilities to track processes, like MULTICS cuc (command usage
count) but I would not be able to identify the contract id for these
processes.

Does anyone have any ideas about how to track server processes, rsh's,
background jobs per contract?  Know of any off-the-shelf packages, etc?

Thanks,

Harold McCabe