Red Hat Bugzilla – Bug 1258115
Do not use NM for br-ex/eth0 in overcloud images
Last modified: 2016-04-18 03:12:54 EDT
Description of problem:
When deploying overcloud with three controllers and two computes we saw compute nodes occasionally becoming unreachable, adding NM_CONTROLLED=no to ifcfg-br-ex and ifcfg-eth0 solved the issue. Looks like this should be by default.
I believe this may be solved by this upstream revision:
Is this just something to backport and/or rebase to fix?
can you check what version of os-net-config is installed on your overcloud nodes?
rpm -q os-net-config
This should be in the current version of os-net-config, which is 0.1.4. It was fixed with this patch:
It is possible you are using out-of-date images, or non-updated ones. If not, then we need to investigate where this regression came from. Can you have a look at the package versions on your deployment image and the actual code and let us know what the situation is?
Waiting for your result on checking that you are using the latest