Red Hat Bugzilla – Bug 874030
Bond mode changing through ifcfg changes has no effect.
Last modified: 2016-11-25 08:08:12 EST
Description of problem: initscripts-9.03.32-1.el6 introduce a change in behavior that makes it impossible to change the bonding mode of an existing bond. In the yum changelog of the initscripts package I can see that in Oct. 8th, entry by Lukáš Nykrýn, there is one line that states: - sets BONDING_OPTS before interface is brougth up and checks if it is not already added I have not looked at the code, but it might be that the commit that resulted in this message is the cause of the issue. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. Create an ifcfg file for a bond with mode 4, e.g., ifcfg-bond10 DEVICE=bond10 BONDING_OPTS="mode=4 miimon=500" BOOTPROTO=none ONBOOT=yes NM_CONTROLLED=no 2. ifup bond10 3. cat /sys/class/net/bond10/bonding/mode to see that mode is effectively 4 4. Change ifcfg-bond10 -BONDING_OPTS="mode=4 miimon=500" +BONDING_OPTS="mode=2 miimon=500" 5. ifdown bond10 6. ifup bond10 Actual results: cat /sys/class/net/bond10/bonding/mode will show the mode to be 4. Expected results: cat /sys/class/net/bond10/bonding/mode should show the mode to be 2. Additional info:
Can you please try following patch? diff --git a/sysconfig/network-scripts/ifup-eth b/sysconfig/network-scripts/ifup-eth index 5ef3055..7dc9fc6 100755 --- a/sysconfig/network-scripts/ifup-eth +++ b/sysconfig/network-scripts/ifup-eth @@ -120,6 +120,7 @@ fi if [ "$ISALIAS" = no ] && is_bonding_device ${DEVICE} ; then /sbin/ip link set dev ${DEVICE} up [ -n "${LINKDELAY}" ] && /bin/sleep ${LINKDELAY} + install_bonding_driver ${DEVICE} for device in $(LANG=C egrep -l "^[[:space:]]*MASTER=\"?${DEVICE}\"?[[:space:]]*$" /etc/sysconfig/network-scripts/ifcfg-*) ; do is_ignored_file "$device" && continue /sbin/ifup ${device##*/}
Unfortunately that does not fix it.
Maybe it's nonsense, but if I apply the patch proposed in comment #3 and modify in network-functions method install_bonding_driver to remove the following line: ethtool -i $1 2>/dev/null | grep -q "driver: bonding" && return 0 Then it works. This line, in previous versions was in the context of a method that was only creating the bond by echoing +bond to bonding masters, but now it also has the responsibility of setting the options. Since, as far as I can see, the only use it had was to avoid doing the bond creation, but now there is already the following: if ! grep -sq "$1" /sys/class/net/bonding_masters; then echo "+$1" > /sys/class/net/bonding_masters 2>/dev/null fi That wasn't there before. I propose to drop the ethtool line.
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. http://rhn.redhat.com/errata/RHBA-2013-0518.html