Chris Bowzer created AVRO-1896:
----------------------------------
Summary: Leaking Deflater Objects
Key: AVRO-1896
URL: https://issues.apache.org/jira/browse/AVRO-1896
Project: Avro
Issue Type: Improvement
Components: java
Reporter: Chris Bowzer
Priority: Minor
In the java version, with normal creation and use of a DeflateCodec there is a
reference to a Deflater object that never has the end method called on it.
While this item still eventually gets garbage collected and release it's c-heap
resources, I have run into the issue of a build up of these objects (and the
resources they are holding) when using many instances of these objects in
parallel. This has caused a hard to diagnose memory issue where the java heap
keeps getting smaller while all resources we have access to are being closed.
While we have solved this issued by recycling DataFileWriters so that there is
only ever a limited amount of DeflateCodec instances to be gc'd, this is not an
ideal solution. It would be great if there was some way of closing the
DeflateCodec instance held inside a DataFileWriter so that deflater.end() can
be called explicitly rather than waiting for the finalizer to run.
Thanks!
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
