Re: [logs] reinventing syslog [was: Secure Central Log Host]

From: Florin Andrei (florinat_private)
Date: Thu Dec 05 2002 - 15:48:21 PST

  • Next message: Tom Perrine: "Re: [logs] SDSC Secure Syslog"

    On Thu, 2002-12-05 at 14:07, Marcus J. Ranum wrote:
    > 
    > >> I think I'd really like to have both timestamps in the final
    > >> destination log file.  The one the client put on, AND the one put on
    > >> by the final "write-to-file" daemone.
    > 
    > What's the point?
    
    Forensics.
    
    Yeah, i know, in an ideal world, every system would run NTP, and the
    wolves and the sheeps would peacefully sip Coke together from the same
    bottle. However, in the _real_ world, sometimes you ask yourself: "what
    was that system's idea of time?"
    Having a timestamp from the syslog server is useful to correlate between
    logs from different machines.
    Having the original timestamp of the machine that send the datagram is
    useful to correlate between events that took place on that machine
    alone.
    
    And it's not that big a deal anyway, it's just a timestamp, a few bytes
    or so.
    
    > For most purposes, both timestamps would be
    > inaccurate enough that you're not really storing any useful information.
    
    Right, but you're not aiming for absolute time precision, but most
    usually for corellation between events. Time refferences on different
    machines not always being perfectly in sync, the more info you have the
    better.
    
    > I believe that recording the messages with the originating system's idea
    > of the time, and the receiving system's idea of the _sequence_ is probably
    > sufficient. As long as you're keeping things in the correct sequence on
    > the server, you're OK.
    
    I agree, provided that you give a unique sequence to each datagram.
    
    For example, let's admit you're logging to a database (yeah, i know) and
    you rely on a self-incrementing column to keep track of the sequence.
    But then, if you log different (groups of) hosts to different tables,
    you loose corellation between those tables.
    
    Overall, it's more simple to use the local timestamp. Simple, intuitive,
    works everywhere.
    
    > In other words: syslog is bad enough that you're not going to make it
    > much better - may as well leave it alone.
    
    Heck, no. What i want to see is a better syslogd implementation, and a
    better syslog protocol.
    Improvements _are_ possible.
    
    -- 
    Florin Andrei
    
    "If you play the WinXP CD backwards, you get a satanic message."
    "That's nothing, if you play it forward, it installs WinXP."
    
    _______________________________________________
    LogAnalysis mailing list
    LogAnalysisat_private
    http://lists.shmoo.com/mailman/listinfo/loganalysis
    



    This archive was generated by hypermail 2b30 : Thu Dec 05 2002 - 16:07:32 PST