> On June 26, 2015, 9:44 p.m., Vinod Kone wrote: > > src/exec/exec.cpp, lines 496-497 > > <https://reviews.apache.org/r/35910/diff/1/?file=993586#file993586line496> > > > > s/create/overwrite/ > > > > s/pure clients/HTTP API/ > > > > Can you move this down to #514 where we also overwrite slave id? > > Vinod Kone wrote: > actually, a better option is to push this intelligence to the slave much > like what we did with the scheduler driver. now that the update.uuid() is > being made optional, we can do it!
Chatted offline with vinod, the driver needs the UUID for tracking unacknowledged updates, which get re-sent during reconnect. Executors using the HTTP API will also have to provide a valid UUID for the same reason. The validation comment below was to validate executor generated UUIDs as RFC-4122 compliant, or which I believe is the case for UUID::random(). - Ben ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/35910/#review89573 ----------------------------------------------------------- On June 26, 2015, 9:11 p.m., Ben Mahler wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/35910/ > ----------------------------------------------------------- > > (Updated June 26, 2015, 9:11 p.m.) > > > Review request for mesos and Vinod Kone. > > > Repository: mesos > > > Description > ------- > > Now that `TaskStatus` includes a 'uuid', the executor driver should set it > the same as the `StatusUpdate` 'uuid'. > > For pure clients, we'll have to use and validate the 'uuid' coming from the > executor. I've added a TODO for this in a subsequent patch. > > > Diffs > ----- > > src/exec/exec.cpp 930dda91068dc6ccbab848b5723ce1568f042779 > > Diff: https://reviews.apache.org/r/35910/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Ben Mahler > >
