[HADOOP-14872] CryptoInputStream should implement unbuffer

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.6.4
• Fix Version/s: 3.1.0, 2.10.0, 2.9.1, 3.0.1
• Component/s: fs, security
• Target Version/s: 3.1.0
• Assignee: John Zhuge Assign to me
• Reporter: John Zhuge
• Created: 15/Sep/17 20:37
• Updated: 29/Dec/17 21:55
• Resolved: 09/Nov/17 18:27

Description

Discovered in IMPALA-5909.

Opening an encrypted HDFS file returns a chain of wrapped input streams:

HdfsDataInputStream
  CryptoInputStream
    DFSInputStream

If an application such as Impala or HBase calls HdfsDataInputStream#unbuffer, FSDataInputStream#unbuffer will be called:

    try {
      ((CanUnbuffer)in).unbuffer();
    } catch (ClassCastException e) {
      throw new UnsupportedOperationException("this stream does not " +
          "support unbuffering.");
    }

If the in class does not implement CanUnbuffer, UOE will be thrown. If the application is not careful, tons of UOEs will show up in logs.

In comparison, opening an non-encrypted HDFS file returns this chain:

HdfsDataInputStream
  DFSInputStream

DFSInputStream implements CanUnbuffer.

It is good for CryptoInputStream to implement CanUnbuffer for 2 reasons:

• Release buffer, cache, or any other resource when instructed
• Able to call its wrapped DFSInputStream unbuffer
• Avoid the UOE described above. Applications may not handle the UOE very well.

Issue Links

• depends upon: HADOOP-15012 Add readahead, dropbehind, and unbuffer to StreamCapabilities
• relates to: HADOOP-13327 Add OutputStream + Syncable to the Filesystem Specification
• relates to: HADOOP-14747 S3AInputStream to implement CanUnbuffer
• relates to: HADOOP-14748 Wasb input streams to implement CanUnbuffer
• relates to: IMPALA-5909 File handle cache causes HDFS to log excessive errors when trying to unbuffer files
• relates to: HDFS-12604 StreamCapability enums are not displayed in javadoc
• relates to: HADOOP-12805 Annotate CanUnbuffer with @InterfaceAudience.Public