wuya created HDDS-15963:
---------------------------
Summary: Fix EC degraded read NPE caused by CodecRegistry
depending on TCCL in isolated ClassLoader
Key: HDDS-15963
URL: https://issues.apache.org/jira/browse/HDDS-15963
Project: Apache Ozone
Issue Type: Bug
Components: Ozone Client
Affects Versions: 2.2.0, 2.1.0
Reporter: wuya
Assignee: wuya
When EC degraded read is triggered in an environment with an isolated plugin
class loader (e.g. Trino Hive connector), CodecRegistry fails to discover
RawErasureCoderFactory providers, causing a NullPointerException.
CodecRegistry uses ServiceLoader.load(RawErasureCoderFactory.class) which
relies on the thread context class loader (TCCL). In Trino's plugin isolation
model, the TCCL cannot see the Ozone jar's META-INF/services providers.
As a result, CodecRegistry.getCoderNames("rs") returns null, leading to:
java.lang.NullPointerException:
Cannot read the array length because "<local3>" is null
at org.apache.ozone.erasurecode.rawcoder.util.CodecUtil
.createRawDecoderWithFallback(CodecUtil.java:90)
The fix is to pass CodecRegistry's own class loader explicitly:
ServiceLoader.load(RawErasureCoderFactory.class,
CodecRegistry.class.getClassLoader());
This ensures provider discovery works regardless of the TCCL, matching the
behavior seen when not using plugin class loaders (e.g. Spark works fine).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
