> On May 2, 2017, 12:13 a.m., Michael Park wrote: > > src/master/http.cpp > > Lines 4884-4889 (patched) > > <https://reviews.apache.org/r/58587/diff/1/?file=1704585#file1704585line4884> > > > > I'm not quite following this example. Specifically, I don't get what "a > > persistent volume can be destroyed if it is not allocated" means.. > > Jiang Yan Xu wrote: > Ok I guess I shoud say "a persistent volume can be destroyed only if it > is not used". In this case, it's not going to be pending offers so you would > unncessarily rescind all offers. > > Michael Park wrote: > Hm... it seems like this should be mitigated by > ```cpp > // If rescinding the offer would not contribute to satisfying > // the required resources, skip it. > Resources recovered = offer->resources(); > recovered.unallocate(); > > if (required == required - recovered) { > continue; > } > ``` > Is this not true?
Hm you are right. I'll drop this comment and leave just the one above. I'll send another rr for caveats about shared persistent volumes, which is different. - Jiang Yan ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/58587/#review173561 ----------------------------------------------------------- On May 1, 2017, 2:29 p.m., Jiang Yan Xu wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/58587/ > ----------------------------------------------------------- > > (Updated May 1, 2017, 2:29 p.m.) > > > Review request for mesos, Anindya Sinha and Michael Park. > > > Repository: mesos > > > Description > ------- > > Seems like the comments about "virtually always win the race against > 'allocate'" could be made more precise in helping people understand > how it works. > > > Diffs > ----- > > src/master/http.cpp 5aae52870451d883ef1ea1fda5a27764d7f318e8 > > > Diff: https://reviews.apache.org/r/58587/diff/1/ > > > Testing > ------- > > N/A > > > Thanks, > > Jiang Yan Xu > >