|
"Palmer J.D.F." wrote:
> Good holiday?
indeed. Hiking in Colorado was great. The scenery is everything they
tell you about. Haven't been at altitude for 20yrs. It was as hard
at 10,000' this time (where we started) as it was at 14,000' 20yrs ago.
I've been getting help on the Linux-HA mailing list about getting rc.lvs
working inside Linux-HA.
I have rc.lvs working inside linux-HA without any hand steering right now.
The rc.lvs script uses start|stop to install|remove VIP, DIP and the ipvsadm
tables. The director failover works but is assymetric. If the backup is active,
and
I bring down heartbeat on the backup, the master director assumes the director
role in about 7secs which is quite good. However if I have the master
director active however and bring down heartbeat on the master director,
it seems to take forever (30secs to a minute I'd guess)
for the backup director to become active. There are clues as
to what's wrong, but I haven't figured it all out yet.
> I've been doing other stuff on the whole but I've had two attempts at
> getting the ipvsadm status synching to work, not entirely successful I must
> say.
> It does seem to work sometimes if you leave it long enough. Not always
> though.
> Basically if I down eth0 on Director1 whilst the synch daemon isn't running
> then control gets passed to Director2, albeit with no client->realserver
> info, this transition happens pretty much immediately think it takes about
> 15 seconds for it to realise then another 10 seconds or so to run
> everything.
> Then on bringing the eth0 on Director1 back up the transition back is
> seamless. And all of my monitoring scripts work and send me the relevant
> emails
sounds allright.
> If I repeat with the synch daemon running heartbeat seems to make the backup
> director active, but it doesn't work, IE no web access, and my monitoring
> scripts don't work. Then on bringing eth0 back up on director1 it carries
> on working.
> Once or twice after a few minutes wait after initially downing director1
> eth0 director2 did spring to life, but after the persistence period so I
> imagine that any status info would have expired anyway.
I haven't worked out how to get the synch demon to trigger a changeover
inside Linux-HA. There's two problems
o I don't know what triggers Linux-HA to failover. I am forcing it to
fail by stopping heartbeat at the moment, the method you used in your
notes to me a few weeks ago.
I would like to do a failover by failing out something that isn't required
for LVS or network connectivity, so I can watch what's happening.
I don't know what Linux-HA is monitoring yet, to know what to attempt
to bring down.
o I don't know to connect the Linux-HA failure to the synch demon. I asked
here for info on how the synch demon works, but have got no reply. It
would seem that the synch demon can changeover the active director
tables all by itself, but as well I have to move the VIP and DIP
(which I'm doing at the moment with Linux-HA) at the same time.
Presumably I have to trigger one from the other or both from
another trigger.
> Oh as long as England don't lose to Brazil, I'll be in a miserable mood if
> they lose and prolly do very little all day.
USA is managing to keep it's composure through the whole thing. I can't
imagine a minute of lost work here no matter what happens. :-)
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
|
