01/02/10

We saw it's too expensive to resolve IP addresses to
hostnames when logging
  - it can be done offline by a program such as lo gresolve

How can we throw lots of requests at our server
  - use a benchmarking tool such as ab

We need to move logs aside on a periodic basis or when
they get full, so they can be moved to cheaper archival
storage
  - pipe Apache's log to the rotatelogs program; it
    will rename the current log and create a new one
    when it needs to

Log watchers
  - to set up email alerts when someting of importance
    happens
  - e.g. LogfSurfer, Swatch, Logwatch

Log analyzers
  - to create stats and charts from a log
  - e.g. Analog (www.analog.cx)

What you can't find out from logs
  - can't rely on client IP address
    e.g. due to proxies
  - identity of user

What's a log for?
  - web-server administrator
    - to measure volume of traffic 
    - to detect problems (e.g. attacks)
  - marketeers
    - can't find out
      - which pages users vist most
      - how long they spend on a page
      - the order in which pages are visited

Web analytics 
  - e.g. Google Analytics
  - use invisible images and cookies to track patterns
    of access
  - much more useful to marketeers than server log