Details
-
Type: Bug
-
Status: Resolved
-
Priority: Blocker
-
Resolution: Not A Bug
-
Affects Version/s: v0.9.3
-
Fix Version/s: None
-
Component/s: Node
-
Labels:None
Description
We're getting errors that the master is "Unable to map logical node 'foo' to physical node 'bar'". We're using tail source and E2E. All of our machines are configured the same way and when restarted, some nodes work, some don't. The logs from the nodes that work look like:
2011-02-19 01:10:49,407 INFO com.cloudera.flume.agent.LogicalNode: Connector started: LazyOpenSource | LazyOpenDecorator
2011-02-19 01:10:49,407 INFO com.cloudera.flume.agent.LogicalNode: production-node: Connector stopped: LazyOpenSource | LazyOpenDecorator
2011-02-19 01:10:49,529 INFO org.apache.hadoop.security.UserGroupInformation: JAAS Configuration already set up for Hadoop, not re-installing.
2011-02-19 01:10:49,530 INFO com.cloudera.flume.agent.FlumeNode: Hadoop Security enabled: false
and the nodes that don't work look like:
2011-02-19 01:10:49,407 INFO com.cloudera.flume.agent.LogicalNode: Connector started: LazyOpenSource | LazyOpenDecorator
2011-02-19 01:10:49,407 INFO com.cloudera.flume.agent.LogicalNode: production-node: Connector stopped: LazyOpenSource | LazyOpenDecorator
2011-02-19 01:10:49,529 INFO org.apache.hadoop.security.UserGroupInformation: JAAS Configuration already set up for Hadoop, not re-installing.
2011-02-19 01:10:49,530 INFO com.cloudera.flume.agent.FlumeNode: Hadoop Security enabled: false
Somehow the connector is stopped for the nodes that don't work. I narrowed it down in the flume code, and it appears that "FireStop()" is being called from DirectDriver.java possibly because source.next() is returning null in the run() method.
Attachments
Issue Links
- relates to
-
FLUME-535 Make agent not connecting to master due to host name error easier to detect.
- Open