Bug ID 1779677: Multiple docker containers can get assigned the same bridge IP during rolling upgrade

Last Modified: Jul 12, 2025

Affected Product(s):
F5OS Install/Upgrade, Velos(all modules)

Fixed In:
F5OS-C 1.8.1

Opened: Dec 23, 2024

Severity: 3-Major

Symptoms

Multiple containers can get the same bridge IP during a rolling upgrade or docker restart [root@controller-2 ~]# docker inspect controller-services-registry-2502 | grep IPAddress "SecondaryIPAddresses": null, "IPAddress": "100.64.0.2", "IPAddress": "100.64.0.2", [root@controller-2 ~]# docker inspect partition-services-registry-2202 | grep IPAddress "SecondaryIPAddresses": null, "IPAddress": "100.64.0.2", "IPAddress": "100.64.0.2",

Impact

This causes one of the two containers to answer requests depending on which container last refreshed the arp cache. The other container does not work properly.

Conditions

When multiple containers start at the same time.

Workaround

Reboot the system.

Fix Information

Docker address allocator uses bit map to manage IP address pool but it's not thread safe. Now, set/unset bitmap operations are protected by a lock.

Behavior Change

Guides & references

K10134038: F5 Bug Tracker Filter Names and Tips