BACK
Logfile Settings

These are the settings for the access log file. Neither the watchdog log nor the message log can be changed. Don't forget that a defined logfile in the default section will be inherited in all the virtual servers unless you explicitely define other settings for them ...
  • enable logging
    This is a flag and takes no value. Activates the logging, if missing, all other settings in the Logfile subblock will be ignored
  • disable logging
    This is a flag and takes no value. Needed to deactivate an inherited logfile, if the VS should use an own logfile, use "enable logging" instead.
  • logfile
    Defines the path and the name pattern for a logfile. As the logfile support is a bit primitiv, you have to add a "%u" somewhere in the name to prevent a rotate to overwrite the previous log. The %u is replaced by the current time stamp. The path can be absolute or relative. However, the target directory and the whole path to it must exist beforehand. Message and watchdog log are always created in the work directory with the patterns msg%u.log and watchdog%u.log respectively.
  • keep in sync
    This is a flag and takes no value. If this is found, the logfile will be updated immediately, else it's used buffered. The non-buffered use of the logfile is discouraged, as it will seriously degrade performance. Watchdog log and error/message log are always kept in sync.
  • max_size
    Defines the maximum size the logfile can have in bytes before a rotation is forced. The default value is set a bit below 2GByte, a value above 2GByte is not supported.
  • max_entries
    Defines the maximum number of accesses (lines) written to the file before a rotate is forced. Default is no max.
  • intervall
    Defines the time in seconds between rotates. However, every rotate will update the time the rotation is forced. Default is 0 (inactive)

A rotation can also be forced by sending the process a SIGHUP signal (rotates ALL logfiles). So you can (and probably should) use a cron job to rotate the logs. I am thinking about implementing something alike in hssTVS, but there's currently no time frame for that.
* The watchdog logfile can only be rotated with a SIGHUP sent to its own task pid *

A logfile entry looks like this:
1145062450 | 84.148.78.40 | http/1.0 | tvs.minxam.com | GET smilies.minxam.com/sad.gif | cache_hit/200 | 602 | 0.001332
ending-timestamp | client IP | http-version | used vhost | method hostname/request-path | cache-status/response-code | transmitted bytes | time needed

BACK