Hi Sven,

On Wed, May 03, 2006 at 10:51:08AM +0200, Sven Anderson wrote:

> But to make sure, that no packets are counted in the wrong bin, the memory
> table has to be locked as long as the port list is processed. Otherwise
> packets that arrive during processing with, for example, port 80, would
> also match the * and got reset, too. Is this handled correctly?

Actually, the memory table gets locked only to avoid inconsistencies, ie.
to serve the '-r' reset flag. What about adding to the client program an
explicit (say, '-l') lock request flag ? Ie. queries are async by default;
but they can be serviced syncronously on-request. 

The solution would be pretty quick to implement. By the end of the day a
fresh snapshot will be ready. Thoughts ?

Cheers,
Paolo


_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists

Reply via email to