[ovs-dev] [RFC] Idea for fixing "raceful" E2E ovn tests

Ryan Moats rmoats at us.ibm.com
Thu Apr 14 13:37:27 UTC 2016


I've pretty much become fed up with the "raceful" nature of the E2E
ovn test cases and so I've set my self the goal of fixing them.

After some thought last night, I *think* I might have found a way
to do it.  Now, since I'm not 100% that my idea is the cleanest way
to fix things, I thought I'd throw the idea out on the table first
and get some feedback before I go off and write code.

I'm thinking of the following changes:

in ovn/controller/ofctrl.c, change ofctrl_put to return a boolean
(true if a flow is queue, false otherwise).

in ovn/controller/ovn-controller.c:
1. add a command line argument (--unit-test).
2. if ofctrl_put return true and the above command line argument
   is specified, dump a line to the log file saying that ovn-controller
   hasn't made any OF changes in the last loop

in the ovn E2E tests cases:
1. start ovn-controller with the --unit-test argument
2. instead of sleep 1 for waiting for ovn-northd/ovn-controller to
   quiesce, look at the tail of the ovn-controller logs for two
   consecutive loops where ovn-controller hasn't made any OF changes
   in the last loop

Thoughts?

Ryan Moats (regXboi)



More information about the dev mailing list