On 19 March 2014 04:50, Ben Pfaff <b...@nicira.com> wrote: > On Tue, Mar 18, 2014 at 03:45:01PM +1300, Joe Stringer wrote: > > Regarding the testsuite failures, the problem is that we wait a certain > > amount of time and check the datapath flows to ensure that they are as > > expected. Patch #8 causes most flows (anything with <8KB/s traffic) to be > > deleted when it is dumped, rather than waiting for the max_idle time. Do > > you have a sense of whether it would be better to: > > * Modify the tests to check that packets flowing through have the correct > > behaviour rather than comparing the dp flows, or > > * Add a flag to disable the threshold logic, and add that flag to the > tests > > so flows with <8KB/s traffic will not be immediately deleted in the test > > scenarios, > > * Or some other alternative > > > > If there is nothing too obvious that I may have missed from the > > xlate_actions() functionality, then that will do for the moment. > > I didn't actually run the tests to see which ones fail, so some of the > following might be badly uninformed. > > I think that modifying the tests, in general, is a good option, if it > doesn't somehow unacceptably reduce the value of the tests. > > Adding a flag seems to me undesirable, especially if the flag is only > useful for tests. I prefer tests to be similar to functionality used by > users, when it can be. (But I'm not saying "no", if this is much better > than other choices.) > > Maybe these tests only want to make sure that correct flows were added > to the datapath at some point. In that case, one could turn up the log > level for "dpif" and check for log messages showing that the flows were > added.
For completeness of discussion, two changes enabled fixing these testsuite issues: * Use dpif logging at debug level to ensure flows are installed correctly where possible. * Changed the way that "high throughput" is calculated so it only applies in situations where flow dump duration is high.
_______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
