|
> What is the kernel version?
[root@cluster root]# uname -r
2.4.7-10
> Show us the 'ipvsadm -Ln' output please.
[root@cluster root]# ipvsadm -Ln
IP Virtual Server version 0.8.1 (size=65536)
Prot LocalAddress:Port Scheduler Flags
-> RemoteAddress:Port Forward Weight ActiveConn InActConn
TCP 172.22.48.212:80 lc
-> 172.16.0.1:80 Local 80 0 0
-> 172.16.0.102:80 Masq 1 0 0
-> 172.16.0.101:80 Masq 2 0 0
Again, VIP=172.22.48.212, NAT LAN=172.16.0.0/24, 172.16.0.1=internal
side of master director.
> Can you provide me with an output of 'ipvsadm -Ln' after 100 http
> requests? So I can see if you use persistency, what scheduler you
> use and if there is really a bug in the LVS scheduler (which I
> don't think)
I yesterday posted an update to my initial message. Actually, the
problem wasn't from LVS/piranha, but from the config of apache on the
master director which wasn't using keepalive http connections, wheres
the two other realservers were.
Problem solved !
> > My other problem is still using the main director when the backup
> > takes over. Supposing the backup redirects a connection to the
> > director's web server (using the NATed network), how will this box
> > be able to answer whereas it does not have the correct gateway
> > (which is set to the external network, not the NATed one).
>
> You use LVS_NAT for your RS and LVS_DR for your director maybe?
No no, see ipvsadm output above.
> The piranha mailing list is best suited for that.
The full story about that, with nice outputs and drawings has been
posted there yesterday.
--
Sébastien Bonnet
Ingénieur d'exploitation
Centre de contacts - Experian France
|
