On Wed, Sep 05, 2018 at 01:55:08PM -0700, Ben Pfaff wrote: > On Wed, Sep 05, 2018 at 01:50:06PM +0200, Thomas Goirand wrote: > > On 09/04/2018 11:06 PM, Ben Pfaff wrote: > > > On Tue, Sep 04, 2018 at 09:20:45AM +0200, Thomas Goirand wrote: > > >> On 09/02/2018 03:12 AM, Justin Pettit wrote: > > >>> > > >>>> On Sep 1, 2018, at 3:52 PM, Ben Pfaff <[email protected]> wrote: > > >>>> > > >>>> On Sat, Sep 01, 2018 at 01:23:32PM -0700, Justin Pettit wrote: > > >>>>> > > >>>>>> On Sep 1, 2018, at 12:21 PM, Thomas Goirand <[email protected]> wrote: > > >>>>>> > > >>>>>> > > >>>>>> The only one failure: > > >>>>>> > > >>>>>> 2633: ovn -- ACL rate-limited logging FAILED > > >>>>>> (ovn.at:6516) > > >>>>> > > >>>>> My guess if that this is meter-related. Can you send the > > >>>>> ovs-vswitchd.log and testsuite.log so I can take a look? > > >>>> > > >>>> It probably hasn't changed from what he sent the first time around. > > >>> > > >>> Yes, "testsuite.log" was in the original message, so I don't need that. > > >>> Thomas, can you send me "ovs-vswitchd.log" and "ovn-controller.log"? > > >>> Does it consistently fail for you? > > >>> > > >>> --Justin > > >> > > >> Hi, > > >> > > >> As I blacklisted the above test, I uploaded to Sid, and now there's a > > >> number of failures on non-intel arch: > > >> > > >> https://buildd.debian.org/status/package.php?p=openvswitch > > >> https://buildd.debian.org/status/logs.php?pkg=openvswitch > > >> > > >> Ben, Justin, can you help me fix all of this? > > > > > > Thanks for passing that along. > > > > > > A lot of these failures seem to involve unexpected timeouts. I wonder > > > whether the buildds are so overloaded that some of the 10-second > > > timeouts in the testsuite are just too short. Usually, this is a > > > generous timeout interval. > > > > > > I sent a patch that should help to debug the problem by doing more > > > logging: > > > https://patchwork.ozlabs.org/patch/966087/ > > > > > > It won't help with tests that fully succeed, because the logs by default > > > are discarded, but for tests that have a sequence of waits, in which one > > > eventually fails, it will allow us to see how long the successful waits > > > took. > > > > > > Any chance you could apply that patch and try another build? Feel free > > > to wait for review, if you prefer. > > > > > > > Hi, > > > > I've just uploaded OVS with that patch. Thanks, I think it's a very good > > idea. And indeed, it looks like failing arch are the slower ones. > > I'm pretty pleased with the theory myself, but the results tend to show > that it wasn't the problem. In most of the tests that eventually > failed, the wait failure was preceded by other waits that succeeded > immediately, and the longest wait I see is 3 seconds. I'll look for > other possible causes.
Most of the test failures seem related to the "asynchronous message control" tests. I haven't yet determined the reason for the failure, but after some work I was able to reproduce it on my own system. _______________________________________________ discuss mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
