----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/56395/ -----------------------------------------------------------
(Updated March 13, 2017, 3:53 p.m.) Review request for Aurora, Stephan Erb and Zameer Manji. Changes ------- I moved getReource back to the ThriftBackfill and instead used the QuantityOf method to do the validations. For this, the validations had to be done against the "config" instead of the backfilled "builder". This is ok given that the resource part of the backfill will be removed soon. By doing this, I found a bug in the client. If "numGpus" was 0, then the whole task.resources would be the empty set. That was later corrected by the backfill. A small change to the client stops that from happening. Repository: aurora Description ------- The Resource validation in ConfigurationManager is now done against the Resource set instead of the NumCpus, RamMb and DiskMb fields. To do this I moved the GetResource method from the ThriftBackfill to the ResourceManager so it can be used in the ConfigurationManager without exposing it on the ThriftBackfill. Related Issue: Aurora-1707 Diffs (updated) ----- src/main/java/org/apache/aurora/scheduler/configuration/ConfigurationManager.java 80f0aebd6de0c2d7c30a58ec09702d1c6d519787 src/main/java/org/apache/aurora/scheduler/storage/log/ThriftBackfill.java c8838438ef5aa94fbcc407101e7f04abc75cd96e src/main/python/apache/aurora/config/thrift.py 3539469d243638c0acd08bf0859d0ce858d8977c src/test/java/org/apache/aurora/scheduler/configuration/ConfigurationManagerTest.java 5419d7edd83c14f155b105b87c5521d7a938e248 src/test/java/org/apache/aurora/scheduler/thrift/SchedulerThriftInterfaceTest.java 0cdd9829417bb3ef0215278a9458a3dd78a49a20 Diff: https://reviews.apache.org/r/56395/diff/2/ Changes: https://reviews.apache.org/r/56395/diff/1-2/ Testing ------- src/test/sh/org/apache/aurora/e2e/test_end_to_end.sh Thanks, Nicolás Donatucci