|
Joseph Mack wrote:
>
> Jon Molin wrote:
>
> > yeah, but that requires two rc.lvs scripts, right? One for the
> > director/server being director and one for it being only server.
>
> I was assuming that the realservers looked the same from each director.
> I hadn't considered your localnode situation. Do you want the realserver
> to always stay on the primary director, ie be localnode when the primary
> is running and a regular realserver. If so you are expecting a realserver
> to be functioning on a director box that is down. Otherwise if you
> had localnode on both of the directors, then the service would look
> local no matter which box was the director.
>
what i want is when i start the two localnodes one of them will assume
the role as both director and realserver while the second localnode is
only a realserver. Which one's which doesn't really matter to me this is
what i hoped to use linux-ha for.
If the localnode being director crashes I want the second localnode (the
one running only as realserver) taking over the role as director, but
not forwarding anything to the crashed (i am looking into mon right
now). Then when the former director comes back up i want it to become a
realserver and the director adds it since it now works.
I'm not sure i explained very good so i'll give it another shot:
i start server 1, since server 2 is down server 1 becomes
direcotor/realserver but doesn't forward to anyone but itself
i start server 2, this becomes realserver since server 1 allready is the
director, server 1 adds server 2 to it's forwarding pool
server 1 goes down for some reason, server 2 sees this and takes over
the role as director/realserver only forwarding to itself
server 1 goes back up, becomes realserver and server 2 adds server 1 to
forwarding pool
ie a cluster. I realise i'll have to combine linux-ha, mon and lvs to
get this working but surely someone else has tried this before? I've
used turboclusterd and it works this way but it's getting far to
expensive if i someday will add a third server and need their 3-10
server license.
> > I'm very new to this but it feels dangerous if heartbeat does something
> > wrong and sets both as directors.
>
> you're totally hosed if this happens and you can expect the HA people
> to have done their homework here.
>
> > here's where things started to go wrong:
> > LVS realserver type vs-dr
> > adding route to real-server network 255.255.255.224
> > route: bogus netmask 212.75.72.63
> >
> > i've no clue where it gets that netmask...it's not in the conf file...
>
> this is your VIP line
>
> VIP=eth1:1 212.75.72.15 255.255.255.192 212.75.72.63
>
> for vs-dr it should be
>
> VIP=eth1:1 212.75.72.15 255.255.255.255 212.75.72.15
>
> it does seem that something else has gone wrong. It
> should setup on your incorrect netmask without
> complaint and just not work.
>
I've got that part working now.
there was also an error with my DIP=, i had only 3 not the 4 required
fields filled in. I realise this adds nothing to this thread but maybe
if someone reads the archives and has similar problems...
> > hmm, i add the output as attachment
>
> that isn't any real help to me. The only thing useful is the output
> when run under sh -x and I'll ask for that offline so it doesn't fill
> up the mailling list archive disk.
I guess you don't need that anymore as i've got it working with the
config script, or?
/Jon
>
> 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
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users
|
