Issue with interface ip and vip subnet

[subhajit-cdot]

Hi,
I am getting an issue with keepalived version 1.2.24. I have configured interface with an /24 ip address. Then I have configured vip of that interface with another /24 ip address, configuration was fine in keepalived.conf. But in the kernel vip is getting configured with /32 ip address for the vip. I am getting same behavior if interface gets ip address from dhcp ( that means if I configure dhcp in the interface not static ip).

The issue does not arise if I configure vip first and then assign ip address to the interface. Is there any dependency with the sequence of programming ip address to interface?

Thanks
Subhajit

[pqarmitage]

Please provide a full copy of your keepalived configuration file so that we can test it.

[subhajit-cdot]

Hi,
I might be wrong with the statement of sequence dependency, but /32 and /24 issue is there and below are the configurations.

`
global_defs {
enable_traps
enable_dbus
snmp_socket tcp:localhost:705:1
enable_snmp_keepalived
enable_snmp_rfc
}
vrrp_sync_group VGW {
group {
enp2s1-1
dp0p8s0.2001-3
dp0p9s0.2002-4
dp0p10s0.2003-5
}
}

vrrp_instance enp2s1-1 {
state BACKUP
interface enp2s1
virtual_router_id 1
version 2
start_delay 0
priority 100
advert_int 1
virtual_ipaddress {
192.168.237.90/24
}
}
vrrp_instance dp0p8s0.2001-3 {
state BACKUP
interface dp0p8s0.2001
virtual_router_id 3
version 2
start_delay 0
priority 100
advert_int 1
virtual_ipaddress {
16.2.101.1/16
}
notify {
/opt/sbin/notify-system-switchover
}
}
vrrp_instance dp0p9s0.2002-4 {
state BACKUP
interface dp0p9s0.2002
virtual_router_id 4
version 2
start_delay 0
priority 100
advert_int 1
virtual_ipaddress {
40.40.40.40/32
}
notify {
/opt/sbin/notify-system-switchover
}
}
vrrp_instance dp0p10s0.2003-5 {
state BACKUP
interface dp0p10s0.2003
virtual_router_id 5
version 2
start_delay 0
priority 200
advert_int 1
virtual_ipaddress {
20.20.1.1/16
}
notify {
/opt/sbin/notify-system-switchover
}
}
`

It is giving issue with enp2s1, for other interfaces it is aligned with the configuration, enp2s1 has got 192.168.237.95/24 from DHCP.

#ip l l

2: enp2s1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 4000
link/ether 52:54:00:55:b4:d0 brd ff:ff:ff:ff:ff:ff
inet 192.168.237.95/24 scope global enp2s1
valid_lft forever preferred_lft forever
inet 192.168.237.90/32 scope global enp2s1
valid_lft forever preferred_lft forever
inet6 2407:9a00:1:1ed:5054:ff:fe55:b4d0/64 scope global dynamic mngtmpaddr
valid_lft 2591850sec preferred_lft 604650sec
inet6 fe80::5054:ff:fe55:b4d0/64 scope link
valid_lft forever preferred_lft forever
11: dp0p8s0.2001@dp0p8s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 36:05:c1:0a:33:b3 brd ff:ff:ff:ff:ff:ff
inet 15.35.14.20/24 scope global dp0p8s0.2001
valid_lft forever preferred_lft forever
inet 16.2.101.1/16 scope global dp0p8s0.2001
valid_lft forever preferred_lft forever
inet6 fe80::3405:c1ff:fe0a:33b3/64 scope link
valid_lft forever preferred_lft forever
12: dp0p10s0.2003@dp0p10s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 00:c0:a8:6f:1b:10 brd ff:ff:ff:ff:ff:ff
inet 15.35.13.20/24 scope global dp0p10s0.2003
valid_lft forever preferred_lft forever
inet 20.20.1.1/16 scope global dp0p10s0.2003
valid_lft forever preferred_lft forever
inet6 fe80::2c0:a8ff:fe6f:1b10/64 scope link
valid_lft forever preferred_lft forever
13: dp0p9s0.2002@dp0p9s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 0a:70:59:33:cc:a5 brd ff:ff:ff:ff:ff:ff
inet 15.35.12.20/24 scope global dp0p9s0.2002
valid_lft forever preferred_lft forever
inet 40.40.40.40/32 scope global dp0p9s0.2002
valid_lft forever preferred_lft forever
inet6 fe80::870:59ff:fe33:cca5/64 scope link
valid_lft forever preferred_lft forever

[pqarmitage]

I am a bit surprised by your configuration since you have configured enable_dbus and start_delay but they are not supported in v1.2.24 and so that suggests you are normally running with a more recent version of keepalived. Does that version add the VIP with the wrong subnet mask?

I have tried your configuration with v1.2.24 and I do not see the problem, i.e. 192.168.237.90 has a /24 subnet mask.

You state that the problem does not exist with the VIPs on the other interfaces. I note that the VIPs either already have a /32 subnet mask, or have a /16 subnet mask which is shorted that the /24 address configured on the interface.

Could you please run kill -USR1 PID_OF_KEEPALIVED_PARENT_PROCESS. This should produce a file /tmp/keepalived.data, and if you post that here, I might be able to work out what is happening.

[pqarmitage]

Also the full logs of keepalived since it started would help