----- Original Message ----- > > From: "Ben Pfaff" <[email protected]> > > To: "Lance Richardson" <[email protected]> > > Cc: [email protected], "Qi Jun Ding" <[email protected]> > > Sent: Thursday, 6 July, 2017 6:42:06 PM > > Subject: Re: [ovs-dev] ovn-nbctl and patch ports for logical routers > > > > On Fri, May 12, 2017 at 09:47:46AM -0400, Lance Richardson wrote: > > > With branch-2.6, we've noticed a difference in behavior when configuring > > > logical switch ports via "ovn-nbctl lsp-set-*" vs. using "ovn-nbctl set > > > Logical_Switch_Port". The difference seems to be a matter of setting > > > individual columns for a logical switch port in multiple transactions > > > versus setting all of them in a single transaction. > > > > This should not make a difference, but it is easy for bugs to creep in. > > > > Is this still something that causes problems? If so, let me know, and > > I'll see if I can find the problem. > > > > Hi Ben, > > This is not an issue in 2.7 and later, I believe due to this commit having > removed the usage of patch ports (and their names): > > f1a8bd0 ovn-controller: Drop most uses of OVS patch ports. > > > The issue is very repeatable on 2.6, I just reproduced it with a fresh > checkout of 2.6.1 in the sandbox environment, it was just a matter of > pasting the command sequences from the previous email. It's my understanding > that this is not just a cosmetic issue, it impacts network connectivity > (maybe Qi Jun could confirm), so users of OVN 2.6 would need to be aware > of the workaround. > > I'm not sure how important this is to fix given it doesn't exist > in 2.7 and later and there is a workaround for 2.6, although I suppose > there could be an underlying issue that would impact other areas (maybe > cases where patch ports are still being used in ovn-controller?) > > Thanks, > > Lance > Hi Ben and Lance,
Yes, when the issue happens, it also impacts the connectivity between two networks on two different logical routers. We found the issue when running the internal test and now our script has been changed to use the workaround. It doesn't impact much on our test. But for end user, I cann't tell if it's the important enough to fix it. Thanks Qijun _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
