Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
3.1.2
-
None
-
None
-
None
Description
During upgrade from 2.6 to 3.1, we encountered a problem:
2019-09-23,19:29:05,303 WARN org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeImpl: Lost container container_e35_1568719110875_6460_08_000001, status: RUNNING, execution type: null 2019-09-23,19:29:05,303 WARN org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeImpl: Lost container container_e35_1568886618758_11172_01_000062, status: RUNNING, execution type: null 2019-09-23,19:29:05,303 WARN org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeImpl: Lost container container_e35_1568886618758_11172_01_000063, status: RUNNING, execution type: null 2019-09-23,19:29:05,303 WARN org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeImpl: Lost container container_e35_1568886618758_11172_01_000064, status: RUNNING, execution type: null 2019-09-23,19:29:05,303 WARN org.apache.hadoop.yarn.server.resourcemanager.rmnode.RMNodeImpl: Lost container container_e35_1568886618758_30617_01_000006, status: RUNNING, execution type: null for (ContainerStatus remoteContainer : containerStatuses) { if (remoteContainer.getState() == ContainerState.RUNNING && remoteContainer.getExecutionType() == ExecutionType.GUARANTEED) { nodeContainers.add(remoteContainer.getContainerId()); } else { LOG.warn("Lost container " + remoteContainer.getContainerId() + ", status: " + remoteContainer.getState() + ", execution type: " + remoteContainer.getExecutionType()); } }
The cause is that we has nm with version 2.6, which do not have executionType for container status.
We should check here make the upgrade process more tranparently
Attachments
Attachments
Issue Links
- duplicates
-
YARN-9547 ContainerStatusPBImpl default execution type is not returned
- Resolved
- links to