Hi Ben,
We could reproduce this with the latest version 2.6.1. When we compiled the
code, we removed -O2 from CFLAGS. This seems to make it happen more
frequently. With the following script running, the error starts happening
within a few seconds and then continues to happen after every few
This issue happened on our in-use systems and we were trying to find a way
to move forward avoiding this issue so that we do not have to upgrade OVS
on thousands of our hypervisors causing down time. Our debugging did help
us avoid the issue for now by installing an explicit rule to to drop
Hi Ben,
Here i is more debug information related to this incident (still using version
2.5.0):
Summary :
We think that there is some race condition involved in processing OF Controller
connections and Packet miss processing in ovs-vswitchd.
Reasoning :
Please consider the following GDB
