|
Hi,
On Sun, 21 Nov 2004, Sarfaraz Munir wrote:
Hi All,
I have successfully setup LVS DR comprising 2 LVS Box
having ipvs-1.0.9 installed and 2 Real servers serving
HTTP, DNS, FTP and Telnet. Besides this heartbeat
+ldirectord also running fine. But when applying IPVS
connection synchronization i am limited to the problem
of Master/salve i.e. Active (Master) LVS syncronizes
connection in a fine way exchaging all of the
established connection information and on failover my
slave(backup) server takes over and the established
connection is continued with the slave LVS server.
But when again Master LVS comes back online it takes
over the control according to the option nice_failback
off in my /etc/ha.d/ha.cf file and Slave server
doesn't synchronizes the established connection with
Master LVS.
In this case, I'd like to suggest that the backup LVS server takes over
and it becomes the master and start the master sync daemon. When the
failed one comes back, it becomes the backup automatically, and start the
backup sync daemon.
If you want to switch back, at least you need some time for the current
master to synchronize most of connections to backup.
I have googled this problem and found one of the
message containing ipvs-1.0.9-syncd patch for solving
this master/salve problem.
IS there anyway to patch my current running ipvs-1.0.9
without recompiling the kernel as i dont have to
configure my LVS boxes again? If anyone direct me how
to patch my ipvs code to include the master/slave
option at a time.
I don't know if there is such a patch. I think that all you need to do to
change your monitoring scripts to handle connection synchronization daemon
correctly, there is no need to recompile the kernel.
Regards,
Wensong
|
