[jira] [Created] (HBASE-29654) BinaryComponentComparator filter fails non gracefully with ArrayIndexOutOfBoundsException

["Daniel Roudnitsky (Jira)" <ji...@apache.org>]

Daniel Roudnitsky created HBASE-29654:
-----------------------------------------

             Summary: BinaryComponentComparator filter fails non gracefully with ArrayIndexOutOfBoundsException
                 Key: HBASE-29654
                 URL: https://issues.apache.org/jira/browse/HBASE-29654
             Project: HBase
          Issue Type: Bug
          Components: Filters
            Reporter: Daniel Roudnitsky
            Assignee: Daniel Roudnitsky


+Problem+
BinaryComponentComparator filter enables a user to compare against a subset of a byte array by specifying an offset into the byte array from which the comparison should start. The issue here is that if the offset provided by the user is longer than a byte array encountered by the scan query (e.g an offset of 40 is used, and there is a "short" byte array of length 38 somewhere in the table), when the scan query reaches the short byte array the query will fail in a nongraceful manner - each scan RPC fails with a mysterious remote ArrayIndexOutOfBoundsException exception, the client will continue to retry the failed RPC excessively until exhausting all retries, and the client ultimately fails with an unintuitive RetriesExhaustedException. 

+Root cause+
Serverside the scan request handler will throw an unhandled/unexpected ArrayIndexOutOfBoundsException during request processing when applying the filter and attempting to do an unchecked byte array comparison using an offset which extends beyond the length of an encountered byte array, when this happens the client will treat the remote exception from the server as an IOException (which wraps the remote exception), and because IOException is retryable, the client will proceed to exhaust all of its retries re-running the same query, all of which are guaranteed to fail, and will eventually fail in a nonobvious and nongraceful manner with a stack trace that looks like:
{code:java}
org.apache.hadoop.hbase.client.RetriesExhaustedException: Failed after attempts=2, exceptions:
2025-09-26T19:09:35.531Z, RpcRetryingCaller{globalStartTime=2025-09-26T19:09:35.412Z, pause=1000, maxAttempts=2}, java.io.IOException: java.io.IOException
	at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:451)
	at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:139)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:369)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:349)
Caused by: java.lang.ArrayIndexOutOfBoundsException

2025-09-26T19:09:36.659Z, RpcRetryingCaller{globalStartTime=2025-09-26T19:09:35.412Z, pause=1000, maxAttempts=2}, java.io.IOException: java.io.IOException
	at org.apache.hadoop.hbase.ipc.RpcServer.call(RpcServer.java:451)
	at org.apache.hadoop.hbase.ipc.CallRunner.run(CallRunner.java:139)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:369)
	at org.apache.hadoop.hbase.ipc.RpcExecutor$Handler.run(RpcExecutor.java:349)
Caused by: java.lang.ArrayIndexOutOfBoundsException


	at org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithRetries(RpcRetryingCallerImpl.java:142)
	at org.apache.hadoop.hbase.client.ResultBoundedCompletionService$QueueingFuture.run(ResultBoundedCompletionService.java:73)
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
	at java.base/java.lang.Thread.run(Thread.java:840)
 {code}
+Proposed solution+
Instead of the server throwing an unhandled/unexpected ArrayIndexOutOfBoundsException, the server should gracefully return a DoNotRetryIOException which prevents the client from making excessive retries which are guaranteed to fail, and the DoNotRetryIOException should clearly explain the root cause to the user - the user provided byte offset for the filter exceeded the length of a byte array that the scan encountered.  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)