Red Hat Bugzilla – Bug 127825
Latest Errata Breaks Bonded Interfaces Without IPs
Last modified: 2014-03-16 22:46:42 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7)
Description of problem:
The latest errata breaks bonded interfaces without IP addresses. If a
bonded interface doesn't have an IP address, none of its slave
interfaces will be attached on boot (it requires a manual ifup of the
At line 221 of /sbin/ifup, you explicitly check for DHCP on the bonded
interfaces and then only up the slaves if present. At line 223, you
check for whether or not the bonded interface has an IP. If not, you
'ip up' it and then immmediately dump the post-config routine. At
line 393, you do the reset of the work for handling bonded interfaces,
including up the slaves. Therefore, without an IP address, the slaves
never get upped.
I'm not sure why all the extra effort is put in to handling DHCP as a
special case. In the old /sbin/ifup, upping the slaves before
assigning the IP address caused no repercussions in our configs.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. config bonded interface with two slaves and no IP address
2. ifup bonded interface
Actual Results: Slave interfaces are not upped, and bonded interface
does not work.
Expected Results: Slave interfaces are upped, and bonded interface
Created attachment 101915 [details]
Patch to fix this
The old way of doing generated large amounts of kernel driver errors with
bonding + DHCP; it's sort of a sideeffect of how bonding works (the kernel
driver was obviously not written with dhcp in mind.)
Please try the attached.
That works fine. The workaround I'm using until an errata is released
is to just set the IPADDR on the bonding config to 0.0.0.0.
This is in current code.