HDDS-609. On restart, SCM does not exit chill mode as it expects DNs to report containers in ALLOCATED state. Contributed by Hanisha Koneru.
Project: http://git-wip-us.apache.org/repos/asf/hadoop/repo Commit: http://git-wip-us.apache.org/repos/asf/hadoop/commit/74db993a Tree: http://git-wip-us.apache.org/repos/asf/hadoop/tree/74db993a Diff: http://git-wip-us.apache.org/repos/asf/hadoop/diff/74db993a Branch: refs/heads/HDFS-12943 Commit: 74db993a61e201d2e8d33bb42f787e0a71388d0d Parents: c011810 Author: Bharat Viswanadham <bha...@apache.org> Authored: Thu Oct 11 21:03:01 2018 -0700 Committer: Bharat Viswanadham <bha...@apache.org> Committed: Thu Oct 11 21:03:01 2018 -0700 ---------------------------------------------------------------------- .../hdds/scm/server/SCMChillModeManager.java | 12 +++-- .../scm/server/TestSCMChillModeManager.java | 53 +++++++++++++++++++- 2 files changed, 61 insertions(+), 4 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/hadoop/blob/74db993a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMChillModeManager.java ---------------------------------------------------------------------- diff --git a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMChillModeManager.java b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMChillModeManager.java index c11a60f..44744f0 100644 --- a/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMChillModeManager.java +++ b/hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/server/SCMChillModeManager.java @@ -29,6 +29,7 @@ import java.util.concurrent.atomic.AtomicBoolean; import java.util.concurrent.atomic.AtomicLong; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.hdds.HddsConfigKeys; +import org.apache.hadoop.hdds.protocol.proto.HddsProtos; import org.apache.hadoop.hdds.protocol.proto.HddsProtos.ScmOps; import org.apache.hadoop.hdds.scm.container.common.helpers.ContainerInfo; import org.apache.hadoop.hdds.scm.events.SCMEvents; @@ -162,7 +163,7 @@ public class SCMChillModeManager implements // Required cutoff % for containers with at least 1 reported replica. private double chillModeCutoff; - // Containers read from scm db. + // Containers read from scm db (excluding containers in ALLOCATED state). private Map<Long, ContainerInfo> containerMap; private double maxContainer; @@ -174,11 +175,16 @@ public class SCMChillModeManager implements containerMap = new ConcurrentHashMap<>(); if(containers != null) { containers.forEach(c -> { - if (c != null) { + // Containers in ALLOCATED state should not be included while + // calculating the total number of containers here. They are not + // reported by DNs and hence should not affect the chill mode exit + // rule. + if (c != null && c.getState() != null && + !c.getState().equals(HddsProtos.LifeCycleState.ALLOCATED)) { containerMap.put(c.getContainerID(), c); } }); - maxContainer = containers.size(); + maxContainer = containerMap.size(); } } http://git-wip-us.apache.org/repos/asf/hadoop/blob/74db993a/hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/server/TestSCMChillModeManager.java ---------------------------------------------------------------------- diff --git a/hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/server/TestSCMChillModeManager.java b/hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/server/TestSCMChillModeManager.java index 53d76e6..41708a4 100644 --- a/hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/server/TestSCMChillModeManager.java +++ b/hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/server/TestSCMChillModeManager.java @@ -22,6 +22,7 @@ import java.util.List; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.hdds.HddsConfigKeys; import org.apache.hadoop.hdds.conf.OzoneConfiguration; +import org.apache.hadoop.hdds.protocol.proto.HddsProtos; import org.apache.hadoop.hdds.scm.HddsTestUtils; import org.apache.hadoop.hdds.scm.container.common.helpers.ContainerInfo; import org.apache.hadoop.hdds.scm.events.SCMEvents; @@ -70,6 +71,11 @@ public class TestSCMChillModeManager { private void testChillMode(int numContainers) throws Exception { containers = new ArrayList<>(); containers.addAll(HddsTestUtils.getContainerInfo(numContainers)); + // Assign open state to containers to be included in the chill mode + // container list + for (ContainerInfo container : containers) { + container.setState(HddsProtos.LifeCycleState.OPEN); + } scmChillModeManager = new SCMChillModeManager(config, containers, queue); queue.addHandler(SCMEvents.NODE_REGISTRATION_CONT_REPORT, scmChillModeManager); @@ -85,6 +91,11 @@ public class TestSCMChillModeManager { public void testChillModeExitRule() throws Exception { containers = new ArrayList<>(); containers.addAll(HddsTestUtils.getContainerInfo(25 * 4)); + // Assign open state to containers to be included in the chill mode + // container list + for (ContainerInfo container : containers) { + container.setState(HddsProtos.LifeCycleState.OPEN); + } scmChillModeManager = new SCMChillModeManager(config, containers, queue); queue.addHandler(SCMEvents.NODE_REGISTRATION_CONT_REPORT, scmChillModeManager); @@ -119,6 +130,46 @@ public class TestSCMChillModeManager { testChillModeDataNodes(5); } + /** + * Check that containers in Allocated state are not considered while + * computing percentage of containers with at least 1 reported replica in + * chill mode exit rule. + */ + @Test + public void testContainerChillModeRule() throws Exception { + containers = new ArrayList<>(); + // Add 100 containers to the list of containers in SCM + containers.addAll(HddsTestUtils.getContainerInfo(25 * 4)); + // Assign OPEN state to first 25 containers and ALLLOCATED state to rest + // of the containers + for (ContainerInfo container : containers.subList(0, 25)) { + container.setState(HddsProtos.LifeCycleState.OPEN); + } + for (ContainerInfo container : containers.subList(25, 100)) { + container.setState(HddsProtos.LifeCycleState.ALLOCATED); + } + + scmChillModeManager = new SCMChillModeManager(config, containers, queue); + queue.addHandler(SCMEvents.NODE_REGISTRATION_CONT_REPORT, + scmChillModeManager); + assertTrue(scmChillModeManager.getInChillMode()); + + // When 10 OPEN containers are reported by DNs, the computed container + // threshold should be 10/25 as there are only 25 open containers. + // Containers in ALLOCATED state should not contribute towards list of + // containers while calculating container threshold in SCMChillNodeManager + testContainerThreshold(containers.subList(0, 10), 0.4); + assertTrue(scmChillModeManager.getInChillMode()); + + // When remaining 15 OPEN containers are reported by DNs, the container + // threshold should be (10+15)/25. + testContainerThreshold(containers.subList(10, 25), 1.0); + + GenericTestUtils.waitFor(() -> { + return !scmChillModeManager.getInChillMode(); + }, 100, 1000 * 5); + } + private void testChillModeDataNodes(int numOfDns) throws Exception { OzoneConfiguration conf = new OzoneConfiguration(config); conf.setInt(HddsConfigKeys.HDDS_SCM_CHILLMODE_MIN_DATANODE, numOfDns); @@ -146,7 +197,7 @@ public class TestSCMChillModeManager { queue.fireEvent(SCMEvents.NODE_REGISTRATION_CONT_REPORT, HddsTestUtils.createNodeRegistrationContainerReport(containers)); GenericTestUtils.waitFor(() -> { - return scmChillModeManager.getInChillMode(); + return !scmChillModeManager.getInChillMode(); }, 10, 1000 * 10); } --------------------------------------------------------------------- To unsubscribe, e-mail: common-commits-unsubscr...@hadoop.apache.org For additional commands, e-mail: common-commits-h...@hadoop.apache.org
