[
https://issues.apache.org/jira/browse/HBASE-22701?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16920870#comment-16920870
]
Josh Elser commented on HBASE-22701:
------------------------------------
.002 tries to separate the config the user provided and only warn when we
failed to find a class via the system ClassLoader and we try to check the
DynamicClassLoader(DCL) because the user asked us to do so.
I added a DEBUG level log message if we wanted to try to load a class via DCL
but DCL was already disabled because we failed to initialize it. I'm worried
about the case where we repeatedly try to load classes and spam the log at WARN
or ERROR. Because using DCL is {{true}} by default, I'm worried about creating
noise.
> Better handle invalid local directory for DynamicClassLoader
> ------------------------------------------------------------
>
> Key: HBASE-22701
> URL: https://issues.apache.org/jira/browse/HBASE-22701
> Project: HBase
> Issue Type: Improvement
> Components: regionserver
> Reporter: Josh Elser
> Assignee: Josh Elser
> Priority: Major
> Fix For: 2.3.0, 2.1.7, 2.2.2
>
> Attachments: HBASE-22701.001.patch
>
>
> If you give HBase an {{hbase.local.dir}} (usually, "{{hbase.tmp.dir}}/local")
> which is not writable to it, you will get some weird errors on the scan path.
> I just saw this (again?) with Phoenix.
> Specifically, the first attempt to reference DynamicClassLoader (via
> ProtobufUtil), will result in an ExceptionInInitializationError because the
> unchecked exception coming out of DynamicClassLoader's constructor interrupts
> the loading of {{DynamicClassLoader.class}}.
> {noformat}
> 2019-07-14 06:25:34,284 ERROR
> [RpcServer.Metadata.Fifo.handler=12,queue=0,port=16020]
> coprocessor.MetaDataEndpointImpl: dropTable failed
> org.apache.hadoop.hbase.DoNotRetryIOException:
> java.lang.ExceptionInInitializerError
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.translateException(RpcRetryingCallerImpl.java:221)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithoutRetries(RpcRetryingCallerImpl.java:194)
> at
> org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:387)
> at
> org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:361)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithRetries(RpcRetryingCallerImpl.java:107)
> at
> org.apache.hadoop.hbase.client.ResultBoundedCompletionService$QueueingFuture.run(ResultBoundedCompletionService.java:80)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: java.lang.ExceptionInInitializerError
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil.toFilter(ProtobufUtil.java:1598)
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil.toScan(ProtobufUtil.java:1152)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.newRegionScanner(RSRpcServices.java:2967)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.scan(RSRpcServices.java:3301)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.openScanner(ScannerCallable.java:332)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.rpcCall(ScannerCallable.java:242)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.rpcCall(ScannerCallable.java:58)
> at
> org.apache.hadoop.hbase.client.RegionServerCallable.call(RegionServerCallable.java:127)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithoutRetries(RpcRetryingCallerImpl.java:192)
> ... 7 more
> Caused by: java.lang.RuntimeException: Failed to create local dir
> /hadoopfs/fs1/hbase/local/jars, DynamicClassLoader failed to init
> at
> org.apache.hadoop.hbase.util.DynamicClassLoader.initTempDir(DynamicClassLoader.java:110)
> at
> org.apache.hadoop.hbase.util.DynamicClassLoader.<init>(DynamicClassLoader.java:98)
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil$ClassLoaderHolder.lambda$static$0(ProtobufUtil.java:261)
> at java.security.AccessController.doPrivileged(Native Method)
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil$ClassLoaderHolder.<clinit>(ProtobufUtil.java:260)
> ... 16 more
> {noformat}
> Every subsequent call will result in a NoClassDefFoundError, because we
> already tried to load DynamicClassLoader.class once and failed.
> {noformat}
> 2019-07-14 06:25:34,380 ERROR
> [RpcServer.Metadata.Fifo.handler=2,queue=2,port=16020]
> coprocessor.MetaDataEndpointImpl: dropTable failed
> org.apache.hadoop.hbase.DoNotRetryIOException:
> java.lang.NoClassDefFoundError: Could not initialize class
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil$ClassLoaderHolder
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.translateException(RpcRetryingCallerImpl.java:221)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithoutRetries(RpcRetryingCallerImpl.java:194)
> at
> org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:387)
> at
> org.apache.hadoop.hbase.client.ScannerCallableWithReplicas$RetryingRPC.call(ScannerCallableWithReplicas.java:361)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithRetries(RpcRetryingCallerImpl.java:107)
> at
> org.apache.hadoop.hbase.client.ResultBoundedCompletionService$QueueingFuture.run(ResultBoundedCompletionService.java:80)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> Caused by: java.lang.NoClassDefFoundError: Could not initialize class
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil$ClassLoaderHolder
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil.toFilter(ProtobufUtil.java:1598)
> at
> org.apache.hadoop.hbase.shaded.protobuf.ProtobufUtil.toScan(ProtobufUtil.java:1152)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.newRegionScanner(RSRpcServices.java:2967)
> at
> org.apache.hadoop.hbase.regionserver.RSRpcServices.scan(RSRpcServices.java:3301)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.openScanner(ScannerCallable.java:332)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.rpcCall(ScannerCallable.java:242)
> at
> org.apache.hadoop.hbase.client.ScannerCallable.rpcCall(ScannerCallable.java:58)
> at
> org.apache.hadoop.hbase.client.RegionServerCallable.call(RegionServerCallable.java:127)
> at
> org.apache.hadoop.hbase.client.RpcRetryingCallerImpl.callWithoutRetries(RpcRetryingCallerImpl.java:192)
> ... 7 more
> {noformat}
> The client gets an error about this, and would presumably know that something
> is amiss, but an operator wouldn't potentially see this on their own.
> I see two options:
> # We abort the RegionServer when the DynamicClassLoader fails to run
> # We catch the exception and treat the DynamicClassLoader as disabled (same
> action as if you had set {{hbase.use.dynamic.jars=false}}).
> I want to do #1 so that we don't propagate bogus configuration, but it feels
> a bit "harsh" to do that. I think #2 is the right solution with a
> big-fat-warning.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
