Cause: When an IP address was re-used it would be generated with a random MAC address that would be different from the previous one.
Consequence: Any node with an ARP cache that still held the old entry for the IP would not be able to communicate with the node.
Fix: Generate the MAC address deterministically from the IP address.
Result: A re-used IP address will always have the same MAC address, so the ARP cache can not be out of sync. So the traffic will flow.
verified in atomic-openshift-3.5.5.28-1.git.0.5eb85e0.el7.x86_64, the pod MAC address generation has been changed and no other issue found.
3: eth0@if15: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue state UP
link/ether 0a:58:0a:02:02:09 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 10.2.2.9/23 scope global eth0
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHBA-2017:1640
verified in atomic-openshift-3.5.5.28-1.git.0.5eb85e0.el7.x86_64, the pod MAC address generation has been changed and no other issue found. 3: eth0@if15: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue state UP link/ether 0a:58:0a:02:02:09 brd ff:ff:ff:ff:ff:ff link-netnsid 0 inet 10.2.2.9/23 scope global eth0