On 04/01/2015 07:50 PM, Ben Pfaff wrote: > On Tue, Mar 31, 2015 at 05:00:44PM -0700, Ben Pfaff wrote: >> On Tue, Mar 31, 2015 at 04:03:04PM -0400, Russell Bryant wrote: >>> Signed-off-by: Russell Bryant <rbry...@redhat.com> >>> --- >>> >>> This is very early, but I wanted to get feedback on the general approach >>> used to >>> detect changes in the ovn and ovn-nb dbs. Does using the IDLs and seqno >>> seem >>> like a sane approach? >>> >>> I don't see an easy way to tell anything about what the changes were. The >>> TODO >>> document mentions that implementing a brute force translation of the entire >>> ovn-nb DB when a change happens may be the easiest first step and this >>> approach >>> seems to support that. More intelligent approaches will probably require >>> context about the changes as they occur. Is that (or anything else) a >>> reason I >>> should be using some another approach for detecting db changes? >> >> This is fine to start. The IDL doesn't currently provide a way to >> find out what the changes are, for a couple of reasons. First, it >> hasn't been necessary in the existing use cases. Second, I'm not sure >> yet what is the best (most convenient) form for the change >> notifications. Anyway, I think we should start with the simplest >> possible approach, then make it more sophisticated as scale warrants. > > I decided that this skeleton was an OK place to start, so I applied it > to the branch. I folded in the following change to keep it from > segfaulting on startup. (I guess that the default databases probably > shouldn't be the same but the same is still better than NULL ;-)
Crashing on startup was a feature! :-p I went ahead and posted my work on this from today. I only had minor updates to the skeleton and them implemented the easier direction (changes in OVN triggering changes to OVN_Northbound). -- Russell Bryant _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
