|
On Tue, Oct 27, 2009 at 01:24:58PM -0200, Raul da Silva {Sp4wn} wrote:
> I heard the following about flow of "master/bakcup" sync (piranha + dr ):
>
> Once a connection is established, the master sync daemon (on the active
> director) sends data to the backup sync daemon (on the backup director).
> The backup director is then aware of an established connection. If the
> primary director fails, the backup will take over and still forward packets
> from an established connection to the right reral server.
>
> This works fine on long lasting connections (interactive sessions like FTP,
> telnet, SSH and so on) but doesn't on short connections (http without
> keepalive for example).
>
> Is it true ? if is true I think the better way to solve this is implement
> some solution like conntrack-tools (netfilter).
Hi Raul,
there are some implementation details relating to packet counts and
connection synchronisation that can influence the non-synchronisation
of packets, but I'm not sure that they would cause the effect that you
describe.
The details are that there is a proc value, sync_threshold which can
be set to between 0 and 49 with a default of 3. Once a connection has
received this many packets it will be queued for synchronisation. It will
then be re-synchronised every 50 packets.
_______________________________________________
Please read the documentation before posting - it's available at:
http://www.linuxvirtualserver.org/
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
or go to http://lists.graemef.net/mailman/listinfo/lvs-users
|
