|
I've looked at the HOWTO, the mini-HOWTO ;-), the docs at
http://www.linuxvirtualserver.org/docs/sync.html, the comments
in ip_vs_sync.c, the ipvs Changelog, the doc directory in ipvs, and the
ipvsadm man page. Here's what I think is true about the
director state sync demon and some questions
o The master and backup exchange information by multicast. The two
directors can do a failover as long as they have a multicast connection?
o The master transfers the ipvs connection table, which contains
the CIP<->RIP table and the tcpip connection state information, to the backup.
o when the backup assumes the master role, it uses whatever information
it has at the time, even if there have been subsequent changes in the tables on
the master node which have not been copied to the backup. The connection
information that is not correct on the new master may result in some
sort of deranged connection (true?) which will eventually timeout(?)
or hang(?).
o How does the backup know to assume the master role?
o Is it possible to trigger the exchange of backup and master machines
from the keyboard (or by pulling cables...)?
o After a failover, does the new master stay the master till the
next failure, or do the roles revert once the old master comes on line
again?
o After failover, how does the new master detect that the new backup
machine is on-line?
o How does the synch demon handle the split-brain problem (ie the
connection between the master and backup is down, but everything
else is working)?
o how do I know which synch demon is the master and which is the backup?
Can I do a command line query to tell the state of the synch demon
on each director?
Thanks Joe
--
Joseph Mack PhD, Senior Systems Engineer, Lockheed Martin
contractor to the National Environmental Supercomputer Center,
mailto:mack.joseph@xxxxxxx ph# 919-541-0007, RTP, NC, USA
|
