[HDFS-1057] Concurrent readers hit ChecksumExceptions if following a writer to very end of file

• Type: Sub-task
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 0.20-append, 0.21.0, 0.22.0
• Fix Version/s: 0.20-append, 0.20.205.0, 0.21.0, 0.22.0
• Component/s: datanode
• Hadoop Flags: Reviewed
• Assignee: sam rash
• Reporter: Todd Lipcon
• Created: 22/Mar/10 00:11
• Updated: 14/Aug/12 00:59
• Resolved: 01/Jul/10 08:38

Description

In BlockReceiver.receivePacket, it calls replicaInfo.setBytesOnDisk before calling flush(). Therefore, if there is a concurrent reader, it's possible to race here - the reader will see the new length while those bytes are still in the buffers of BlockReceiver. Thus the client will potentially see checksum errors or EOFs. Additionally, the last checksum chunk of the file is made accessible to readers even though it is not stable.

Issue Links

• is related to: HDFS-1401 TestFileConcurrentReader test case is still timing out / failing
• is related to: HDFS-1103 Replica recovery doesn't distinguish between flushed-but-corrupted last chunk and unflushed last chunk
• is related to: HDFS-1679 TestFileConcurrentReader fails intermittently
• is related to: HDFS-1885 Recurring failures in TestFileConcurrentReader for > 12 days
• is related to: HDFS-1310 TestFileConcurrentReader fails
• relates to: HDFS-3719 Re-enable append-related tests in TestFileConcurrentReader
• relates to: HADOOP-7146 RPC server leaks file descriptors