AW: Piranha not doing "gratuitous lvs arps" on Failover AND is Piranha the right Solution for me ?

Hallo List,

Update :

it seems that Piranha sometimes does send_arp right. On the Internal side it
seems to work now. But External there is no "arp 10.49.140.238 is-at" in
"tcpdump -i eth0 arp"
And somehow Heartbeat now uses the Internal AND External network!? This
might bee nice, but somehow the Backup LB only takesover if Primary looses
both connections or goes completly down.
is that the way it is supposed to work ?

Has anybody got a running System where takeover occurs if only 1 NIC on the
Primary looses its connection ?

Thank you
Christian

> -----Ursprüngliche Nachricht-----
> Von: Christian Jacobsen [mailto:[email protected]]
> Gesendet: Freitag, 15. November 2002 12:17
> An: [email protected]
> Betreff: Piranha not doing "gratuitous lvs arps" on Failover AND is
> Piranha the right Solution for me ?
>
>
> Hallo Piranha-list,
>
> I have two Questions - I hope Somebody can Help me :)
>
> Configuration :
>
> Two LB (one primary and one secondary) : Red Hat Linux 7.2, Kernel 2.4.18
> 	(both Compaq DL360 G2 - bcm5700 driver downloaded from Compaq)
>
> ipvsadm-1.20-7 (from http://www.academy.rpi.edu/~yua/open_source/piranha/)
> piranha-0.7.0-3 (from
> ftp://ftp.linux.org.uk/pub/linux/piranha/7.2/piranha/)
>
> LVS-NAT is used.
>
> while testing 2 Real Server running NT 4.0 (both also Compaq DL360 G2)
>
> lvs.cf is included at the end of this mail.
>
>
> Problem :
> With the Primary LB everything is Working - so far so good.
>
> My problem is on Failover (if a Network connection is removed on
> Primary LB). When then the Backup LB starts it says that it makes
> "gratuitous lvs arps" - but nothing happens on the real Server or
> on the Router on the other side :(
> A "tcpdump -i eth0 arp" does not show anything of that kind !?
>
> What could be wrong ?
>
>
> How can I activate Debugging ?
>
> i tried to set the -v on pulse start but that only shows the Heartbeat.
> And "debug_level = MAX" does not do anything
>
> ##########################################
>
> And the Second Question. Is Piranha the right Solution for me ?
>
> I am trying to get a failure secure System running but have found
> that it is not always taking over or not shutting down services
> on the Failed LB.
>
> Further infos to my configurtaion : The LB's each have 2 NIC's
> which each are connected to a different Switch on each side -
> that means i have 4 Switches. ETH0 is on the External side and
> ETH1 on the Internal. The Real Servers are each connected to one
> of these Switches on the Internal side.
>
> Following is in my opinion not working :
> When i remove the connection on ETH0 on the Primary LB the second
> LB takes over - which is good. But the Primary LB stay up so that
> the VIP now is on ETH1 of BOTH LB's :( That should give some problems
>
> When i remove the connection on ETH1 on the Primary LB the second
> LB DOES NOT takes over because Heartbeat is still working. And
> the Primary LB can not reach the Real Severs :( So nothing is
> working anylonger.
>
> What I need is some kind of Dual Heartbeat and/or some
> intelligence that tests which LB can reach the most Real Servers
> or somthing like that.
>
> I am Sure sombody needed/did this before - but what did you use ?
>
> !!! Please do not suggest that i buy RedHat HA 2.1 because I have
> to get it running till Monday Evening :) !!!
>
> ############################################
> Here is my lvs.cf :
>
> serial_no = 143
> primary = 10.49.140.236
> primary_private = 10.49.143.1
> service = lvs
> backup_active = 1
> backup = 10.49.140.237
> backup_private = 10.49.143.2
> heartbeat = 1
> heartbeat_port = 539
> keepalive = 6
> deadtime = 18
> network = nat
> nat_router = 10.49.143.3 eth1:0
> nat_nmask = 255.255.255.0
> reservation_conflict_action = preempt
> debug_level = NONE
> virtual TEST {
>      active = 1
>      address = 10.49.140.238 eth0:1
>      vip_nmask = 255.255.255.0
>      fwmark = 3730
>      port = 3730
>      persistent = 100
>      load_monitor = none
>      scheduler = rr
>      protocol = tcp
>      timeout = 6
>      reentry = 15
>      quiesce_server = 0
>      server GW1A {
>          address = 10.49.143.11
>          active = 1
>          weight = 100
>      }
>      server GW1B {
>          address = 10.49.143.12
>          active = 1
>          weight = 100
>      }
> }
>
>
> Thank you in Advance
> Christian Jacobsen
>