Re: [logs] Logging - descriptive vs normative

From: Bill Hudacek (hudacekat_private)
Date: Tue Aug 20 2002 - 10:16:03 PDT

  • Next message: copelandtat_private: "RE: [logs] Logging: World Domination"

    > Paul Ebersman <list-loganalysisat_private>
    > Sent by: loganalysis-adminat_private 08/20/02 10:51 AM
    > To: loganalysisat_private
    > cc: 
    > Subject: [logs] Re: What's normal?
    > 
    > tbird> Yeah, yeah, yeah, and for decades, system administrators have
    > tbird> talked themselves out of doing the experiment with arguments
    > tbird> exactly this way.
    > 
    > Fair enough. But I'd still guess that for most newbies, the real first
    > problem is learning to drink from the firehose. Coming up with some
    > semi-standard way of chewing through megabytes of raw logs to figure
    > out your baseline would be useful even if we *never* come up with a
    > standardized configuration.
    
    I listen more than speak on mailing lists such as this one; this is 
    probably
    not even my fifth post to this particular list.
    
    However, I read with great interest not Tina's initial message, but her
    restatement of a longish list of default behaviours that she'd like to see
    logged (or treated in this discussion).
    
    
    There appears to be tension between "descriptive" and "normative" here. As 
    a
    designer/architect of applications systems and "data center environments", 
    I
    recognize the need for categorization/classification (descriptive); the 
    most
    benefit I've seen reaped in the past came from automated software that
    processed thousands of emails from applications across more than 100
    enterprise-class UNIX boxes in a data center - back in the days when 
    syslog
    was broken on most UNIX flavors.  Categorization is "low-hanging fruit".
    
    However, what could *really* be exciting would be breaking new ground - 
    yes,
    think of templates for "format" (but interpret that word as loosely as
    possible!).  Yes, if a "segregation" scheme can be devised, it would be a 
    huge
    gain for "newbies" - or for the senior admin who's just thinking about 
    moving
    to a new logging architecture and wants to be using it effectively 
    *today*.
    
    Combining these two perspectives, then, might result in the following
    solution characteristics:
    
    1.      Provide a *common*, default, out-of-the-box config that maximizes
            flexibility and that can be provided with package downloads such 
    as
            syslog-ng, and used immediately.
     
            as we all know, compiling an enterprise or system management app 
    is
            usually only 2-10% of the job.  The first time I set up "cups" it 
    took me
            *days* for my particular environment (with an unusual samba
            environment).  There just isn't an "on-ramp" in too many cases.
    
    2.      Provide a "package recommendation" for log analysis that processes 
    the
            raw data produced using the specification of (1) above.
    
    3.      (ideal) Provide functional feedback to a) logging facilities 
    software
            maintainers/developers and b) logging analysis sw
            maintainers/developers with lessons learned from (1) and (2) 
    above.
    
    I know I'd have liked to not have to process raw data in my past 
    experiences.
    Front-loading *some* of the statistical analysis, "close to" if not within 
    the
    local logging facility, would allow, for example, a "remote logging" 
    server to 
    send over 
    
      "15 DEBUG last 30 min message 'Attempted Access modem port /dev/cua0'"
    
    rather than receiving the actual 15 messages.  Admittedly it could be done
    anywhere in this architecture, and I'm simply putting forth a rather 
    simplistic
    example; but, there may be many cases where "close to the action" 
    intelligence
    makes more sense than a big honking "logging processor".
    
    I'm sure there are other dimensions to the discussion that I've omitted; 
    this
    is just my $0.02 concerning the direction the conversation is taking.
    
    Regards,
    
    Bill Hudacek
    IBM Global Services
    
    "There has grown in the minds of certain groups in this country the
    idea that just because a man or corporation has made a profit out of
    the public for a number of years, the government and the courts are
    charged with guaranteeing such profit in the future, even in the face
    of changing circumstances and contrary to public interest. This
    strange doctrine is supported by neither statue or common law. Neither
    corporations or individuals have the right to come into court and ask
    that the clock of history be stopped, or turned back."
    
    -Robert Heinlein, Life Line, 1939
    
    _______________________________________________
    LogAnalysis mailing list
    LogAnalysisat_private
    https://lists.shmoo.com/mailman/listinfo/loganalysis
    



    This archive was generated by hypermail 2b30 : Tue Aug 20 2002 - 11:32:47 PDT