Home / server / Questions / 601670
In Process
tvb
Asked: 2014-06-05 00:20:44 +0800 CST

keepalived vip removed with dhcp renewal

  • 772

I am having this strange issue where my vip is being removed from my interface when I receive a new DHCP renewal request. The logs shows the following:

Jun  1 17:00:06 lb1 dhclient: DHCPREQUEST of 10.0.0.2 on eth0 to 10.0.0.3 port 67 (xid=0x6deab016)
Jun  1 17:00:06 lb1 dhclient: DHCPNAK from 10.0.0.3 (xid=0x6deab016)
Jun  1 17:00:06 lb1 dhclient: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 3 (xid=0x37e1db6a)
Jun  1 17:00:06 lb1 dhclient: DHCPREQUEST of 10.0.0.2 on eth0 to 255.255.255.255 port 67 (xid=0x37e1db6a)
Jun  1 17:00:06 lb1 dhclient: DHCPOFFER of 10.0.0.2 from 10.0.0.3
Jun  1 17:00:06 lb1 dhclient: DHCPACK of 10.0.0.2 from 10.0.0.3
Jun  1 17:00:06 lb1 dhclient: bound to 10.0.0.2 -- renewal in 38223 seconds.
Jun  1 17:00:07 lb1 ntpd[1321]: Deleting interface #8 eth0, 10.0.255.254#123, interface stats: received=7, sent=0, dropped=0, active_time=145198 secs
Jun  1 17:00:07 lb1 ntpd[1321]: peers refreshed

At this point the vip ip (10.0.255.254) is removed from the interface and is not being failed over to the other keepalived instance.

The working ip a command output:

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
   inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
    link/ether fa:16:3e:cf:35:ac brd ff:ff:ff:ff:ff:ff
    inet 10.0.0.2/16 brd 10.0.255.255 scope global eth0
       valid_lft forever preferred_lft forever
    inet 10.0.255.254/32 scope global eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fecf:35ac/64 scope link
       valid_lft forever preferred_lft forever

and when the dhcp renewal has ran:

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
   inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP group default qlen 1000
    link/ether fa:16:3e:cf:35:ac brd ff:ff:ff:ff:ff:ff
    inet 10.0.0.2/16 brd 10.0.255.255 scope global eth0
       valid_lft forever preferred_lft forever
    inet6 fe80::f816:3eff:fecf:35ac/64 scope link
       valid_lft forever preferred_lft forever

the interfaces file:

# This file describes the network interfaces available on your system
# and how to activate them. For more information, see interfaces(5).

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto eth0
iface eth0 inet dhcp

When I restart keepalived on the instance the vip is added back to the interface again. So the question is why is dhcp removing the vip?

dhcp

1 Answers

  1. I have also recently encountered this same issue. I'm running two OpenStack instances with keepalived. I noticed that when the instance renewed its IP address sometimes keepalived thought the IP address was removed and then re-added it moments later. That was enough for it to release the VIPs, but did not re-add them to the primary keepalived instance nor was it enough for the standby instance to take over the VIPs.

    After looking over keepalived's docs I found an option called dont_track_primary that can be added to the vrrp instance. I added that option to all of my vrrp instances and haven't had the issue since.

    • 2