[lvs-users] Garp problem on Keepalived v1.1.0 (24/07, 2003) [Kernel 2.4.18-5smp] with Cisco Switches

["Shaun Mccullagh" <Shaun.Mccullagh@xxxxx>]

Hi,

We have small problem on our oldest KL setup in that KL GARPS seem not
to be propagated by our Cisco switches when the KL Master fails.

The KL Backup system advertises all the VIPs correctly and the log
reports that GARPS are sent when it transits to Master state.

However clients on the same network as 10.0.45.0/24 and 10.0.50.5/24
cannot ping 10.0.45.254 or 10.0.50.5. I note the all the client ARP
caches still have the MAC address of the Master KL system, not the
Backup KL system.

Could I fix this problem by increasing garp_master_delay?

BTW the reason why advert_int is set so high is because we have had
problems with switch stability which lead to both KL Master and KL
Backup system going to Master state ==> Duplicate IPs

TIA

Shaun


! Configuration File for keepalived

global_defs {
   smtp_server 192.168.104.95
   smtp_connect_timeout 30
   lvs_id PO2
}

vrrp_sync_group G1 {
  group {
        VI_1
        VI_2
  }
  smtp_alert
}

vrrp_instance VI_1 {
    state BACKUP
    interface eth0.12 
    track_interface {
        eth0
    }
    virtual_router_id 100
    priority 100
    advert_int 60
    authentication {
        auth_type PASS
        auth_pass xxx
    }
    virtual_ipaddress {
       10.0.12.16 dev eth0.12
    }
    virtual_ipaddress_excluded {
       80.249.35.21/28 dev eth0.10
    }

    garp_master_delay 1
}

vrrp_instance VI_2 {
    state BACKUP
    interface eth1 
    track_interface {
        eth1
    }
    virtual_router_id 101
    priority 100
    advert_int 60
    authentication {
        auth_type PASS
        auth_pass marviq
    }
    virtual_ipaddress {
       10.0.43.254/24 dev eth1
    }
    virtual_ipaddress_excluded {
        10.0.45.254/24 dev eth1.45
      10.0.50.5/24 dev eth1.50  
    }
    garp_master_delay 1
} 





Op dit e-mailbericht is een disclaimer van toepassing, welke te vinden is op 
http://www.xb.nl/disclaimer.html