|
Yes it's the week view of the traffic on the director interface :
http://www.freazer.com/perso/mkerscaven/lvs/difference.png
For the moment I'm not doing cache but just domain rewriting with the squids.
Day view of the director interface :
http://www.freazer.com/perso/mkerscaven/lvs/lvs.png
Day views of the 3 real squid interfaces :
http://www.freazer.com/perso/mkerscaven/lvs/squid1.png
http://www.freazer.com/perso/mkerscaven/lvs/squid2.png
http://www.freazer.com/perso/mkerscaven/lvs/squid3.png
Regards,
Marc.
Graeme Fowler <graeme@xxxxxxxxxxx> a écrit :
On Sun, 2005-12-18 at 18:45 +0100, Kerscaven Marc wrote:
> since I've upgraded to kernel 2.6.14.2 for the director, I've noticed that
> input and output traffic is not the same like in the 2.6.8.
>
> Yo can see the difference hear :
> http://www.freazer.com/perso/mkerscaven/lvs/difference.png
> I've upgraded the 12/17.
I think it's more likely a bug in the _old_ kernel.
Considering you're running a set of squids, I'd expect the traffic to be
biased in one direction; think about this:
An empty cache has to fetch content before serving to the clients;
inbound and outbound of the cluster as a whole is likely to be around
the same.
A "full" cache (that is, one where content is starting to expire) is
likely to show far more traffic destined towards the clients than
destined towards the cache.
I find it rather surprising that prior to your new kernel the traffic
was identical in both directions.
What figures is your graph showing, exactly? Is that the traffic on the
director's interface(s)?
Graeme
---------------------------------
Nouveau : téléphonez moins cher avec Yahoo! Messenger ! Découvez les tarifs
exceptionnels pour appeler la France et l'international.Téléchargez la version
beta.
|
