Description of problem:
A single node hostsubnet object that contains a hostip equal to its tun0 interface causing the error in dmesg:
"Dead loop on virtual device tun0, fix it urgently!"
Is resulting in stack corruption and reboot of all other nodes with even correct configuration for hostsubnet.
Upstream openshift-sdn pull request to prevent this on the openshift side is https://github.com/openshift/openshift-sdn/pull/245
https://github.com/openshift/openshift-sdn/pull/245 got merged to origin and will be present in OpenShift 3.2 and later.
Test latest OSE env, issue have been fixed.
After changing hostip in hostsubnet on master1 node1 equal to the ip of its tun0, node2 could work normally without any crash or reboot, and the pods located on node1 could transfer to node2.
(In reply to Dan Williams from comment #13)
> https://github.com/openshift/openshift-sdn/pull/245 got merged to origin and
> will be present in OpenShift 3.2 and later.
OSE 3.2 is now available: https://access.redhat.com/errata/RHSA-2016:1064
(In reply to Yan Du from comment #14)
> Test latest OSE env, issue have been fixed.
> openshift v18.104.22.168
> kubernetes v1.2.0-36-g4a3f9c5
> etcd 2.2.5
3.2 released version atomic-openshift-22.214.171.124-1.git.0.f44746c.el7 which should contain the fix.