Bug ID 1273025: Once TACACS server-group is configured on a non-default port, on downgrade virt-handler pod gets into a crash loopbackoff state because of SELinux corruption

Last Modified: Aug 01, 2024

Affected Product(s):
F5OS None(all modules)

Fixed In:
F5OS-C 1.6.0, F5OS-A 1.7.0, F5OS-A 1.5.0

Opened: Mar 17, 2023

Severity: 2-Critical

Symptoms

Virt-handler pod is crashing upon downgrading from F5OS-A 1.4.0 to F5OS-A 1.3.2, and a tenant is stuck in pending state.

Impact

Tenant becomes stuck in pending state.

Conditions

Configure TACACS server-group on a non-default port (that is, other than port 49) and downgrade from F5OS-A 1.4.0 to F5OS-A 1.3.2.

Workaround

Two workarounds: 1. Configure standard port (49) for the TACACS server instead of a non-standard port. Using the standard port should not trigger these issues. Releases older than F5OS-A 1.4.0 do not correctly support any port other than 49 for the TACACS server. 2. Fix SELinux policy on the appliance: a. cp selinux module from /usr cp /usr/etc/selinux/targeted/active/modules/400/f5_appliance to /etc/selinux/targeted/active/modules/400/f5_appliance b. Reboot the device reboot

Fix Information

Enforce closure of non-standard port every time configuration is updated or system is shut down. This avoids leaving ports open for which SELinux may not have exceptions. Note this does not address the non-standard port in older releases when downgrading. Non-standard ports for TACACS are still not properly supported in the older release.

Behavior Change

Guides & references

K10134038: F5 Bug Tracker Filter Names and Tips