You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Jin Gary (Jira)" <ji...@apache.org> on 2020/08/24 03:28:00 UTC
[jira] [Created] (YARN-10405) ResourseManager use HA mode both in
Standby
Jin Gary created YARN-10405:
-------------------------------
Summary: ResourseManager use HA mode both in Standby
Key: YARN-10405
URL: https://issues.apache.org/jira/browse/YARN-10405
Project: Hadoop YARN
Issue Type: Bug
Affects Versions: 3.1.1
Reporter: Jin Gary
our yarn use capacity scheduler, config is
{code:java}
yarn.scheduler.capacity.maximum-am-resource-percent=0.2
yarn.scheduler.capacity.maximum-applications=10000
yarn.scheduler.capacity.node-locality-delay=20
yarn.scheduler.capacity.queue-mappings=u:username:queuename,...
yarn.scheduler.capacity.queue-mappings-override.enable=true
yarn.scheduler.capacity.resource-calculator=org.apache.hadoop.yarn.util.resource.DominantResourceCalculator
yarn.scheduler.capacity.root.acl_administer_queue=*
yarn.scheduler.capacity.root.acl_submit_applications=*
yarn.scheduler.capacity.root.capacity=100
yarn.scheduler.capacity.root.default.acl_submit_applications=*
yarn.scheduler.capacity.root.default.capacity=1
yarn.scheduler.capacity.root.default.maximum-am-resource-percent=1
yarn.scheduler.capacity.root.default.maximum-applications=10
yarn.scheduler.capacity.root.default.maximum-capacity=30
yarn.scheduler.capacity.root.default.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.default.ordering-policy=fair
yarn.scheduler.capacity.root.default.ordering-policy.fair.enable-size-based-weight=false
yarn.scheduler.capacity.root.default.priority=0
yarn.scheduler.capacity.root.default.state=RUNNING
yarn.scheduler.capacity.root.default.user-limit-factor=10
yarn.scheduler.capacity.root.developer.acl_administer_queue=*
yarn.scheduler.capacity.root.developer.acl_submit_applications=*
yarn.scheduler.capacity.root.developer.capacity=7
yarn.scheduler.capacity.root.developer.developer_data.acl_administer_queue=*
yarn.scheduler.capacity.root.developer.developer_data.acl_submit_applications=*
yarn.scheduler.capacity.root.developer.developer_data.capacity=50
yarn.scheduler.capacity.root.developer.developer_data.maximum-am-resource-percent=0.8
yarn.scheduler.capacity.root.developer.developer_data.maximum-capacity=70
yarn.scheduler.capacity.root.developer.developer_data.minimum-user-limit-percent=30
yarn.scheduler.capacity.root.developer.developer_data.ordering-policy=fair
yarn.scheduler.capacity.root.developer.developer_data.ordering-policy.fair.enable-size-based-weight=false
yarn.scheduler.capacity.root.developer.developer_data.priority=20
yarn.scheduler.capacity.root.developer.developer_data.state=RUNNING
yarn.scheduler.capacity.root.developer.developer_data.user-limit-factor=5
yarn.scheduler.capacity.root.developer.developer_recsys.acl_administer_queue=*
yarn.scheduler.capacity.root.developer.developer_recsys.acl_submit_applications=*
yarn.scheduler.capacity.root.developer.developer_recsys.capacity=50
yarn.scheduler.capacity.root.developer.developer_recsys.maximum-am-resource-percent=0.8
yarn.scheduler.capacity.root.developer.developer_recsys.maximum-capacity=70
yarn.scheduler.capacity.root.developer.developer_recsys.minimum-user-limit-percent=30
yarn.scheduler.capacity.root.developer.developer_recsys.ordering-policy=fair
yarn.scheduler.capacity.root.developer.developer_recsys.ordering-policy.fair.enable-size-based-weight=false
yarn.scheduler.capacity.root.developer.developer_recsys.priority=20
yarn.scheduler.capacity.root.developer.developer_recsys.state=RUNNING
yarn.scheduler.capacity.root.developer.developer_recsys.user-limit-factor=5
yarn.scheduler.capacity.root.developer.maximum-am-resource-percent=1
yarn.scheduler.capacity.root.developer.maximum-capacity=40
yarn.scheduler.capacity.root.developer.minimum-user-limit-percent=30
yarn.scheduler.capacity.root.developer.ordering-policy=priority-utilization
yarn.scheduler.capacity.root.developer.priority=20
yarn.scheduler.capacity.root.developer.queues=developer_data,developer_recsys
yarn.scheduler.capacity.root.developer.state=RUNNING
yarn.scheduler.capacity.root.developer.user-limit-factor=3
yarn.scheduler.capacity.root.olap.acl_administer_queue=*
yarn.scheduler.capacity.root.olap.acl_submit_applications=*
yarn.scheduler.capacity.root.olap.capacity=5
yarn.scheduler.capacity.root.olap.maximum-am-resource-percent=1
yarn.scheduler.capacity.root.olap.maximum-capacity=30
yarn.scheduler.capacity.root.olap.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.olap.ordering-policy=fifo
yarn.scheduler.capacity.root.olap.priority=50
yarn.scheduler.capacity.root.olap.state=RUNNING
yarn.scheduler.capacity.root.olap.user-limit-factor=3
yarn.scheduler.capacity.root.ordering-policy=priority-utilization
yarn.scheduler.capacity.root.pipeline.acl_administer_queue=*
yarn.scheduler.capacity.root.pipeline.acl_submit_applications=*
yarn.scheduler.capacity.root.pipeline.capacity=30
yarn.scheduler.capacity.root.pipeline.maximum-allocation-mb=21504
yarn.scheduler.capacity.root.pipeline.maximum-capacity=30
yarn.scheduler.capacity.root.pipeline.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.pipeline.ordering-policy=fifo
yarn.scheduler.capacity.root.pipeline.priority=60
yarn.scheduler.capacity.root.pipeline.state=RUNNING
yarn.scheduler.capacity.root.pipeline.user-limit-factor=1
yarn.scheduler.capacity.root.priority=0
yarn.scheduler.capacity.root.queues=default,developer,pipeline,task,olap,realtime,yarn-system
yarn.scheduler.capacity.root.realtime.acl_administer_queue=*
yarn.scheduler.capacity.root.realtime.acl_submit_applications=*
yarn.scheduler.capacity.root.realtime.capacity=6
yarn.scheduler.capacity.root.realtime.maximum-capacity=6
yarn.scheduler.capacity.root.realtime.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.realtime.ordering-policy=fifo
yarn.scheduler.capacity.root.realtime.priority=60
yarn.scheduler.capacity.root.realtime.state=RUNNING
yarn.scheduler.capacity.root.realtime.user-limit-factor=1
yarn.scheduler.capacity.root.task.acl_administer_queue=*
yarn.scheduler.capacity.root.task.acl_submit_applications=*
yarn.scheduler.capacity.root.task.capacity=51
yarn.scheduler.capacity.root.task.maximum-am-resource-percent=1
yarn.scheduler.capacity.root.task.maximum-capacity=59
yarn.scheduler.capacity.root.task.minimum-user-limit-percent=60
yarn.scheduler.capacity.root.task.ordering-policy=priority-utilization
yarn.scheduler.capacity.root.task.priority=40
yarn.scheduler.capacity.root.task.queues=task_data,task_recsys
yarn.scheduler.capacity.root.task.state=RUNNING
yarn.scheduler.capacity.root.task.task_data.acl_administer_queue=*
yarn.scheduler.capacity.root.task.task_data.acl_submit_applications=*
yarn.scheduler.capacity.root.task.task_data.capacity=70
yarn.scheduler.capacity.root.task.task_data.maximum-am-resource-percent=0.9
yarn.scheduler.capacity.root.task.task_data.maximum-capacity=80
yarn.scheduler.capacity.root.task.task_data.minimum-user-limit-percent=90
yarn.scheduler.capacity.root.task.task_data.ordering-policy=fifo
yarn.scheduler.capacity.root.task.task_data.priority=40
yarn.scheduler.capacity.root.task.task_data.state=RUNNING
yarn.scheduler.capacity.root.task.task_data.user-limit-factor=3
yarn.scheduler.capacity.root.task.task_recsys.acl_administer_queue=*
yarn.scheduler.capacity.root.task.task_recsys.acl_submit_applications=*
yarn.scheduler.capacity.root.task.task_recsys.capacity=30
yarn.scheduler.capacity.root.task.task_recsys.maximum-am-resource-percent=0.9
yarn.scheduler.capacity.root.task.task_recsys.maximum-capacity=50
yarn.scheduler.capacity.root.task.task_recsys.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.task.task_recsys.ordering-policy=fifo
yarn.scheduler.capacity.root.task.task_recsys.priority=40
yarn.scheduler.capacity.root.task.task_recsys.state=RUNNING
yarn.scheduler.capacity.root.task.task_recsys.user-limit-factor=3
yarn.scheduler.capacity.root.task.user-limit-factor=2
yarn.scheduler.capacity.root.yarn-system.acl_administer_queue=yarn-ats
yarn.scheduler.capacity.root.yarn-system.acl_submit_applications=yarn-ats
yarn.scheduler.capacity.root.yarn-system.capacity=0
yarn.scheduler.capacity.root.yarn-system.default-application-lifetime=-1
yarn.scheduler.capacity.root.yarn-system.disable_preemption=true
yarn.scheduler.capacity.root.yarn-system.maximum-am-resource-percent=0.5
yarn.scheduler.capacity.root.yarn-system.maximum-application-lifetime=-1
yarn.scheduler.capacity.root.yarn-system.maximum-capacity=100
yarn.scheduler.capacity.root.yarn-system.minimum-user-limit-percent=100
yarn.scheduler.capacity.root.yarn-system.ordering-policy=fifo
yarn.scheduler.capacity.root.yarn-system.priority=32768
yarn.scheduler.capacity.root.yarn-system.state=RUNNING
yarn.scheduler.capacity.root.yarn-system.user-limit-factor=1
{code}
As you can see, preemption is enabled between the queues.
Now there are two faults:
# active rm will enter standby mode during operation, but standby rm not automatically switch to active, and two rm logs frequently print abnormal information:
{panel}
2020-08-18 22:35:19,451 WARN rmapp.RMAppImpl (RMAppImpl.java:<init>(473)) - The specific max attempts: 0 for application: 2769 is invalid, because it is out of the range [1, 3]. Use the global max attempts instead.
2020-08-18 22:35:19,451 INFO collector.TimelineCollectorManager (TimelineCollectorManager.java:putIfAbsent(149)) - the collector for application_1597677167198_2769 already exists!
2020-08-18 22:35:19,451 INFO placement.UserGroupMappingPlacementRule (UserGroupMappingPlacementRule.java:getPlacementForApp(201)) - Application application_1597677167198_2770 user kylin mapping [olap] to [org.apache.hadoop.yarn.server.resourcemanager.placement.ApplicationPlacementContext@4ad0f252] override true
2020-08-18 22:35:19,451 WARN rmapp.RMAppImpl (RMAppImpl.java:<init>(473)) - The specific max attempts: 0 for application: 2770 is invalid, because it is out of the range [1, 3]. Use the global max attempts instead.
2020-08-18 22:35:19,451 INFO collector.TimelineCollectorManager (TimelineCollectorManager.java:putIfAbsent(149)) - the collector for application_1597677167198_2770 already exists!
2020-08-18 22:35:19,452 INFO placement.UserGroupMappingPlacementRule (UserGroupMappingPlacementRule.java:getPlacementForApp(201)) - Application application_1597677167198_2771 user kylin mapping [olap] to [org.apache.hadoop.yarn.server.resourcemanager.placement.ApplicationPlacementContext@2287c6d0] override true
{panel}
After a while, there was an zookeeper timeout exception:
ERROR curator.ConnectionState (ConnectionState.java:checkTimeouts(228)) - Connection timed out for connection string (zookeeper_host1:2181,zookeeper_host2:2181,zookeeper_host3:2181,zookeeper_host4:2181,zookeeper_host5:2181) and timeout (15000) / elapsed (13677)
org.apache.curator.CuratorConnectionLossException: KeeperErrorCode = ConnectionLossorg.apache.curator.CuratorConnectionLossException: KeeperErrorCode = ConnectionLoss at org.apache.curator.ConnectionState.checkTimeouts(ConnectionState.java:225) at org.apache.curator.ConnectionState.getZooKeeper(ConnectionState.java:94) at org.apache.curator.CuratorZookeeperClient.getZooKeeper(CuratorZookeeperClient.java:117) at org.apache.curator.framework.imps.CuratorFrameworkImpl.performBackgroundOperation(CuratorFrameworkImpl.java:835) at org.apache.curator.framework.imps.CuratorFrameworkImpl.backgroundOperationsLoop(CuratorFrameworkImpl.java:809) at org.apache.curator.framework.imps.CuratorFrameworkImpl.access$300(CuratorFrameworkImpl.java:64) at org.apache.curator.framework.imps.CuratorFrameworkImpl$4.call(CuratorFrameworkImpl.java:267) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180) at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624) at java.lang.Thread.run(Thread.java:748)
2. No tasks in the queue are running, but the resource occupancy rate has not been released, this causes the newly submitted task to remain ACCEPTED !image-2020-08-24-11-07-50-899.png!
ps: We occasionally use
yarn application -movetoqueue application_1478676388082_963529 -queue task_data
to make tasks get sufficient queue resources
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org