Clients can hang in close if processDatanodeError throws Exception ( ex: OOME). -------------------------------------------------------------------------------
Key: HDFS-2850 URL: https://issues.apache.org/jira/browse/HDFS-2850 Project: Hadoop HDFS Issue Type: Bug Components: hdfs client Affects Versions: 1.0.1 Reporter: Uma Maheswara Rao G Assignee: Uma Maheswara Rao G I met a situation, where DataStreamer#processDatanodeError throws OOME when creating ResponseProcessor thread. Due to this Datastreamer thread died. When clinet closing the stream, it keeps waiting. Looks this is because, when clinet closes, it will enque one packet by marking that a lastpacket and wait for the ack. Here Datastreamer thread died and no one is there for processsing the packet from dataqueue. Obviously will not get any ack and it will keep wait in close. This i have seen in 20.2 version. when i verified, this problem will not be there in trunk as processDatanodeError already guarded with try/catch. This problem can be there in branch-1 -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira