Re: [logs] tokens and layouts...

From: Chris Adams (cadamsat_private)
Date: Tue Aug 27 2002 - 10:44:57 PDT

  • Next message: Martin Sierakowiak: "[logs] question about loganalysis products"

    On Monday, Aug 26, 2002, at 19:57 US/Pacific, Russell Fulton wrote:
    > 2/ Some machines are constantly sync'ed using NTP some sync'ed on boot,
    > some are not sync'ed at all.  Having that information included with the
    > log file could be useful if at some time in the future you need to do
    > correlations with other files.  If you don't know how accurate the
    > clocks are it is dam near impossible.
    
    Unless we just prominently document the problem and otherwise punt the 
    issue, I see two choices. The best one would be including something 
    like NTP in the spec to determine the relative offsets between the 
    clients and the log server at periodic intervals. This would give 
    accuracy at the expense of adding a non-trivial amount of additional 
    work for implementors.
    
    As a compromise what do you think about having any relay or log server 
    simply add a local time element to any message which differs from its 
    clock by more than a defined period? That'd still leave network delays 
    but it would help us track down consistently wrong sources or relays. 
    We'd probably also need that element to include some info about its 
    time source and when it was last synced since it's easy to imagine 
    scenarios where a remote network's relay could lose connectivity to the 
    NTP server, too.
    
    Chris
    
    _______________________________________________
    LogAnalysis mailing list
    LogAnalysisat_private
    http://lists.shmoo.com/mailman/listinfo/loganalysis
    



    This archive was generated by hypermail 2b30 : Wed Aug 28 2002 - 10:34:13 PDT