On Fri, Oct 07, 2022 at 01:37:45PM +0200, Eelco Chaudron wrote: > > > On 15 Sep 2022, at 14:03, Ilya Maximets wrote: > > > On 9/15/22 13:56, Ilya Maximets wrote: > >> On 9/15/22 13:38, Tianyu Yuan wrote: > >>> > >>> On 9/15/22 19:28, Ilya Maximets wrote:
... > >>>> This looks very strange to me. The test does send 10 packets. > >>>> Why the flow should report only one? > >>>> > >>> Thanks for your review Ilya. > >>> The test does send 10 packets but 9 of them are dropped by > >>> meter action. As we descript in commit (dd9881ed55e6), the > >>> flow stats should not report the police stats. > > > > "In previous patch, after parsing police action, the flower stats will > > be updated by dumped meter table stats" > > > > I suppose, that is the root cause. We should not mix these two > > completely different types of statistics. I didn't look at the > > code though to say why this was implemented in a way it is or > > how to fix that. > > Tianyu I might have missed this, but is there a follow on fix for this? > Asking as the test in the current master is broken. > > Ilya/Simon should we undo “dd9881ed5 tc: Fix stats dump when using same meter > table” and get a proper fix, fixing these counter issues? Thanks Eelco, there has been a bit of follow-on discussion in this thread regarding the root cause and possible fixes. We think it is fair to say that it is now clear that dd9881ed5 does not address the issue it seeks to address. Tianyu and I agree that while working on a solution it would be best to revert dd9881ed5. And we expect to post a patch to revert dd9881ed5 in the near future. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
