On Thu, Jul 28, 2016 at 10:10:57PM -0500, Ryan Moats wrote: > "dev" <dev-boun...@openvswitch.org> wrote on 07/27/2016 02:03:23 AM: > > > From: Ben Pfaff <b...@ovn.org> > > To: dev@openvswitch.org > > Cc: Ben Pfaff <b...@ovn.org> > > Date: 07/27/2016 02:04 AM > > Subject: [ovs-dev] [PATCH 3/5] ovn-nbctl: Add "sync" command to wait > > for previous changes to take effect. > > Sent by: "dev" <dev-boun...@openvswitch.org> > > > > It's slow to add --wait to every ovn-nbctl command; only the last command > > needs it. But it's sometimes inconvenient to add it to the last command > > if it's in a loop, etc. This makes it possible to separately wait for > > the OVN southbound or hypervisors to catch up to the northbound. > > > > Also, modify the tests to replace "sleep" invocations that were waiting > > for hypervisors to catch up by --wait=hv or by invocations of "ovn-nbctl > > --wait=hv sync". > > > > Signed-off-by: Ben Pfaff <b...@ovn.org> > > --- > > Paradoxically, after applying this, I saw more failures at the sync check > points that I was seeing with the original code. Rechecking showed that > the tests pass, so I'm thinking that we might want to consider either > increasing the timeouts from 5 seconds or making them configurable so > that people building/testsing on slower systems (and/or running tests in > parallel) can tune them to avoid false positives. > > Honestly, all of that can be handled in a later patchset, so... > > Acked-by: Ryan Moats <rmo...@us.ibm.com>
Thanks for the review. It seems likely to me that there's still a small remaining race with sync (or equivalently with --wait=hv). I know that there's one in the case where ovs-vswitchd has already handled packets in a given flow, but most of these cases don't have that. I think there must be something else in ovn-controller itself; maybe it's not properly updating everything southbound of it before it reports that it's in sync. I think I'll sit on this one for a bit until we get that sorted. (Patch 5 in the series fixes some additional races that this patch exposes, but it doesn't fix all of them.) _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev