> On Oct. 18, 2017, 2:48 p.m., Stephan Erb wrote: > > src/main/java/org/apache/aurora/scheduler/HostOffer.java > > Lines 44-47 (patched) > > <https://reviews.apache.org/r/62956/diff/4/?file=1862825#file1862825line44> > > > > Please extend this comment a bit and indicate that this checks for the > > absence of revocable and non-revocable resources.
Done. > On Oct. 18, 2017, 2:48 p.m., Stephan Erb wrote: > > src/main/java/org/apache/aurora/scheduler/HostOffer.java > > Lines 62-66 (patched) > > <https://reviews.apache.org/r/62956/diff/4/?file=1862825#file1862825line62> > > > > I have had a short peek at our clusters and the number of offers > > matching this criteria would be very low. > > > > It could be increased if we'd use a higher bound like 0.1 CPU and 64mb > > memory as those would already be needed as executor overhead. > > > > I suspect for Twitter even using 0 as abound is significant though? > > Correct? > > David McLaughlin wrote: > For Twitter is is quite large, but it would be large (as a %) for anyone > who enables any type of bin-packing. I'd like to start with the zero-config patch with little room for doubt. Though i welcome the follow-up patch/discussion on a higher threshold if you think it would benefit you, Stephan! > On Oct. 18, 2017, 2:48 p.m., Stephan Erb wrote: > > src/main/java/org/apache/aurora/scheduler/state/TaskAssigner.java > > Lines 295-296 (patched) > > <https://reviews.apache.org/r/62956/diff/4/?file=1862829#file1862829line295> > > > > For consistency with the other early aborts below, write this in form > > of an `if` with a `continue` statement? > > > > (I don't feel strongly about this though. It just feels a little bit > > weird that this is handled differently here) I agree, this is poor consistency. Fixed. **Note**: i am also going to change the meaning of the `assigner_evaluated_offers` stat. It currently counts these early aborts, which i believe will complicate any kind of performance analysis related to optimizations like this. - Bill ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/62956/#review188586 ----------------------------------------------------------- On Oct. 18, 2017, 2:25 p.m., Bill Farner wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/62956/ > ----------------------------------------------------------- > > (Updated Oct. 18, 2017, 2:25 p.m.) > > > Review request for Aurora, David McLaughlin and Jordan Ly. > > > Repository: aurora > > > Description > ------- > > There's no reason for us to evaluate offers with no CPUs or memory, so reject > them early in the offer lifecycle. > > This is an incremental performance optimization, but it may net significant > improvements based on observations in some very large clusters. > > > Diffs > ----- > > src/main/java/org/apache/aurora/scheduler/HostOffer.java > bc40d0798f40003cab5bf6efe607217e4d5de9f1 > src/main/java/org/apache/aurora/scheduler/http/Utilization.java > 3c77e2983ce00f897f3d5ed106b779cd7f7f0940 > > src/main/java/org/apache/aurora/scheduler/preemptor/PreemptionVictimFilter.java > 1b1239753f40d7d46d91724def6c25037eb79f1c > src/main/java/org/apache/aurora/scheduler/resources/ResourceBag.java > d5db81b88a0369d0b26c8fbf70efab3886ad7695 > src/main/java/org/apache/aurora/scheduler/state/TaskAssigner.java > e35720fca903cdcd27fff50f12ece9cc615f5d54 > src/main/java/org/apache/aurora/scheduler/stats/TaskStatCalculator.java > b98aaaf48ae60afef19a368ee96abc897300f8fa > src/test/java/org/apache/aurora/scheduler/HostOfferTest.java PRE-CREATION > src/test/java/org/apache/aurora/scheduler/offers/OfferManagerImplTest.java > 2cfdc090ff75a63111ae146c9fe7b3542e7ac83f > src/test/java/org/apache/aurora/scheduler/offers/Offers.java > 129b4437315c6ad4ea47ca75d4ae6e28cadd7911 > src/test/java/org/apache/aurora/scheduler/resources/ResourceTestUtil.java > 765a527acb96997989c920be8b69dfa1113dc302 > > src/test/java/org/apache/aurora/scheduler/state/FirstFitTaskAssignerTest.java > a74efbdc55383cd748f52cd55e1a94551d7d2ba1 > > > Diff: https://reviews.apache.org/r/62956/diff/4/ > > > Testing > ------- > > > Thanks, > > Bill Farner > >
