Hello Simone,
you should be referring to the 'stamp_updated' field. It tells you when
the record has been "touched" for the last time. Such minor delays could
appear for a variety of reasons, including:

* It is a cumulative delay while dealing with a remote SQL server.
* The old tale of 'lazy' refresh times. The story behind it: once in
  the times (and they were really OLD times, Wim Kerkhoff was still
  using MySQL !!) pmacct was used to handle refresh time deadlines
  the 'hard' way: by setting timers via syscall and getting kicked by
  the kernel (via SIGALRM) each expiration. Disturbing the kernel too
  much is never a good rule, expecially when it's assumed to be already
  busy delivering you the amount of packets captured from the network !
  So, we switched to 'soft' (or 'lazy') refresh times: captured packets
  make pmacct rolling; why not being savy and take advantage of such
  rolling to handle (set and check) refresh time deadlines ? 
  The method is then 'protected' by a timeout value which is equal to
  the refresh time value which make it working smoothly even on 'mute'
  networks (a quite common situation in lab/test environments). 

At this propo, remember also my last email: a correct value for refresh
time should not be too little. As a general rule, it should be kept >=
30 secs. 


Cheers,
Paolo

Reply via email to