[
https://issues.apache.org/jira/browse/HBASE-16264?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15451276#comment-15451276
]
stack commented on HBASE-16264:
-------------------------------
Digging in more, the above NoClassDefFoundError is a linkage time error (parent
class is LinkageError). Reading the JVM Spec.[2], we could be getting this
exception at linking time or probably later, at resolve when we are trying
replace the Class constant_pool 'logical' references with actual locations (the
discussion about late vs eager resolve doesn't seem to apply here; our
violation must be triggering eager error reporting). In the Class file, it has
methods which have com.google.protobuf.* parameters (for Coprocessor
Endpoints). We're messing stuff up when Classes are uniquely identified by the
name+classloader.
In the above, we are just trying to get the public methods on the class looking
for the main method, one of the first tasks getting a program off the ground. I
tried moving the HRegionServer main to its own class and starting the program
from there but just got a different variation on the above; it too has to load
HRegionServer class eventually with similar repercussions:
1408 RCL loading relocated com.google.protobuf.RpcCallback as
"org.apache.hadoop.hbase.shaded.com.google.protobuf.RpcCallback"
1409 [Loaded org.apache.hadoop.hbase.shaded.com.google.protobuf.RpcCallback
from
file:/Users/stack/checkouts/hbase.git/hbase-shaded-libs/target/hbase-shaded-libs-2.0.0-SNAPSHOT.jar]
1410 Exception in thread "main" [Loaded java.lang.Throwable$PrintStreamOrWriter
from
/Library/Java/JavaVirtualMachines/jdk1.8.0_65.jdk/Contents/Home/jre/lib/rt.jar]
1411 [Loaded java.lang.Throwable$WrappedPrintStream from
/Library/Java/JavaVirtualMachines/jdk1.8.0_65.jdk/Contents/Home/jre/lib/rt.jar]
1412 [Loaded java.util.IdentityHashMap$KeySet from
/Library/Java/JavaVirtualMachines/jdk1.8.0_65.jdk/Contents/Home/jre/lib/rt.jar]
1413 java.lang.NoClassDefFoundError: com/google/protobuf/RpcCallback
1414 at java.lang.Class.forName0(Native Method)
1415 at java.lang.Class.forName(Class.java:264)
1416 at
org.apache.hadoop.hbase.regionserver.HRegionServerCommandLine.main(HRegionServerCommandLine.java:107)
getDeclaredMethods0 and forName0 are native methods [3] NoClassDefFoundError is
thrown if we can't 'fix' the classname -- see end of [4] -- or it is coming up
out of JVM_FindClassFromClassLoader [5]. I kinda stopped trying to chase at
this point.
Next up I'd say would be rewriting byte codes to replace any
com.google.protobuf references with
org.apache.hadoop.hbase.shaded.com.google.protobuf using something like ASM --
it has a means for doing this on load called RemappingClassAdapter -- but I'd
have to do it for every class loaded. This seems over the top.
I'm going to go back to a modified tactic #1. I'll fill out the backing doc w/
new approach which I'll call tactic #3 where I move the embedded endpoints
after hbase-client and hbase-server....
1.
http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/8u40-b25/sun/launcher/LauncherHelper.java#LauncherHelper.validateMainClass%28java.lang.Class%29
2.https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-5.html#jvms-5.4
3.
http://code.metager.de/source/xref/openjdk/jdk8/jdk/src/share/native/java/lang/Class.c#99
4.
https://github.com/frohoff/jdk7u/blob/master/src/share/native/common/check_format.c
5.
http://hg.openjdk.java.net/jdk7/jdk7/hotspot/file/tip/src/share/vm/prims/jvm.cpp#l737
> Figure how to deal with endpoints and shaded pb
> -----------------------------------------------
>
> Key: HBASE-16264
> URL: https://issues.apache.org/jira/browse/HBASE-16264
> Project: HBase
> Issue Type: Sub-task
> Components: Coprocessors, Protobufs
> Reporter: stack
> Assignee: stack
> Priority: Critical
> Fix For: 2.0.0
>
> Attachments: 16264.tactic2.patch, HBASE-16264.master.001.patch,
> HBASE-16264.master.002.patch, HBASE-16264.master.003.patch,
> HBASE-16264.master.004.patch, HBASE-16264.master.005.patch,
> HBASE-16264.master.006.patch, HBASE-16264.master.007.patch
>
>
> Come up w/ a migration plan for coprocessor endpoints when our pb is shaded.
> Would be sweet if could make it so all just worked. At worst, come up w/ a
> prescription for how to migrate existing CPs.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
