Re: Director Failover

Yes. Perhaps it's the arp issue that result in the
telnet connection to stay for a while. I have read in
the mailing list that some piranha patch is required
for the VIP takeover by backup director to happen. Is
that true ? (i.e. remove primary director eth alias
for VIP and add backup director eth alias for VIP
during director failover)

 --- BMan <[email protected]> wrote: > First
off, piranha is nothing more than a name given
> to the "Frontend" 
> or gui.  With that said, it utilizes the same ipvs
> tables as any other 
> lvs Linux based solution.  If the other list says it
> can be done, my 
> guess is they are doing a lot more hard-core hand
> coding to handle 
> arping issues that the Piranha interface can not
> accomplish unless 
> they've done some major work on. My guess is they
> haven't sense Advance 
> Server is the new "flag ship". Of course I have no
> idea of the context 
> of which you speak of so bare with me as I try to go
> into a little more 
> detail for you.
> 
> The real key in making any lvs solution work is
> understanding how arp 
> works.  What happens to arp tables on the director
> and boxes providing 
> services when fail over occures can be monitored to
> help diagnose.  A 
> sniffer, standard arp commands,  netcat and a few
> other tools that 
> escape my brain are good diagnostic tools.  The key
> to seeing what the 
> interfaces are seeing.  In the Cisco world all of
> the arp issues are 
> taken care of at the router/director level and boxes
> behind it are saved 
> from the arp issues and storms that can happen.  At
> one point the kernel 
> team wasn't  putting patches into the kernel to
> handle lvs, I'm not sure 
> if it is back in or removed forever??\
> 
> <Mike any time you want to jump in here feel free>
> 
> There are ways of setting up a combination of  lo
> interfaces along with 
> iptables to handle some of these issues.  It is very
> cluge though and 
> hard to maintain in my opinion.  I say lo interfaces
> because even dummy 
> interfaces arp out.  :/  I'm not sure where they are
> with integrating 
> back into the mainstream kernel patches that handle
> some of this (Mike 
> ?). You may also want to look at doing that.   I do
> know that Red Hat's 
> Advanced Server handles a lot of these issues in a
> more comprehensible 
> package.  It also uses shared SCSI in a better
> manner than Piranha can. 
>  On the flip side there are a whole new set of issue
> with that solution 
> that I don't want to get into.
> 
> Bottom line here is it is almost impossible for me
> to diagnose your 
> setup via email.  What I can almost bet my life on
> is your most likely 
> running into an arping issue that the plain "out of
> the box piranha" 
> doesn't handle without minipulating the underlying
> parts that it sets on 
> top off.
> 
> Hope this sheads some light, good luck.
> 
> Regards,
> BMan
> 
> 
> 
> ong cheechye wrote:
> 
> >Hi BMan,
> >
> >I've read about questions raised on connection
> >synchronization between the primary and backup
> >director in the lvs mailing list. I suppose that
> means
> >existing connection can be maintained during
> director
> >failover. Does that means connection
> synchronization
> >is achieved in other software but not Piranha ?
> >
> >During my test with piranha, I noticed the telnet
> >session is still on shortly after the director
> >failover but went dead eventually. How do you
> explain
> >this observation ?
> >
> >Thanks
> >Chee Chye
> >
> > --- BMan <[email protected]> wrote: > The
> >existing telnet connection will be disconnected
> >  
> >
> >>because Piranha 
> >>(regardless of routing method used)  does not
> track
> >>session state.  The 
> >>same holds true for ssh, ftp and similar services.
> >>
> >>Regards,
> >>BMan
> >>
> >>ong cheechye wrote:
> >>
> >>    
> >>
> >>>Hi,
> >>>
> >>>I'm testing Piranha+Direct routing with telnet
> >>>      
> >>>
> >>service
> >>    
> >>
> >>>using a primary director and a backup director
> >>>      
> >>>
> >>which
> >>    
> >>
> >>>is also acting as a realserver (due to shortage
> of
> >>>machines). 
> >>>
> >>>I telnet from another machine to the VIP. I then
> >>>      
> >>>
> >>stop
> >>    
> >>
> >>>the pulse in the primary director and the backup
> >>>director take over. Would the existing telnet
> >>>connection be disconnected ?
> >>>
> >>>Thanks
> >>>Chee Chye
> >>> 
> >>>
>
>>>__________________________________________________
> >>>Do You Yahoo!?
> >>>Promote your business from just $5 a month!
> >>>http://sg.biztools.yahoo.com
> >>>
> >>>
> >>>
> >>>_______________________________________________
> >>>Piranha-list mailing list
> >>>[email protected]
> >>>      
> >>>
>
>>https://listman.redhat.com/mailman/listinfo/piranha-list
> >>    
> >>
> >>> 
> >>>
> >>>      
> >>>
> >>
> >>
> >>_______________________________________________
> >>Piranha-list mailing list
> >>[email protected]
> >>
> >>    
> >>
>
>https://listman.redhat.com/mailman/listinfo/piranha-list
> 
> >
> >__________________________________________________
> >Do You Yahoo!?
> >Promote your business from just $5 a month!
> >http://sg.biztools.yahoo.com
> >
> >
> >
> >_______________________________________________
> >Piranha-list mailing list
> >[email protected]
>
>https://listman.redhat.com/mailman/listinfo/piranha-list
> >
> >  
> >
> 
>  

__________________________________________________
Do You Yahoo!?
Promote your business from just $5 a month!
http://sg.biztools.yahoo.com