[
https://issues.apache.org/jira/browse/HBASE-14548?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15362486#comment-15362486
]
li xiang commented on HBASE-14548:
----------------------------------
Hi Jerry, thanks for the review.
1. I tested FileSytem.isDirectory() by a separate program, against the path
with wildcard. It works as expected for the following inputs, for example,
"/user/hbase/*.jar" or "/user/hbase/coprocessor.*", isDirectory() returns false.
The condition I found which will fail is that suppose the jar file is put as
/user/hbase/coprocessor.jar and "/user/h*" is used as the input. As
isDirectory() returns false, the code does not append "*.jar".
FileSystem.globStatus() returns all directories and files starting with "h"
under /user, such as hbase, hive...(if there are). But JarFile followed can not
handle a directory.
So I add the logic : For each item globStatus() returned, process file only and
skip for directory. And if all items are directories, throw
FileNotFoundException: No file found matching hdfs://xxxx/xxx*
I uploaded v2 for branch 1.2.0 and master to include those changes.
2. Regarding the enhancement on ClassLoaderTestHelper, I opened a new JIRA as
https://issues.apache.org/jira/browse/HBASE-16173
> Expand how table coprocessor jar and dependency path can be specified
> ---------------------------------------------------------------------
>
> Key: HBASE-14548
> URL: https://issues.apache.org/jira/browse/HBASE-14548
> Project: HBase
> Issue Type: Improvement
> Components: Coprocessors
> Affects Versions: 1.2.0
> Reporter: Jerry He
> Assignee: li xiang
> Fix For: 2.0.0
>
> Attachments: HBASE-14548-1.2.0-v0.patch, HBASE-14548-1.2.0-v1.patch,
> HBASE-14548-1.2.0-v2.patch, HBASE-14548-master-v1.patch
>
>
> Currently you can specify the location of the coprocessor jar in the table
> coprocessor attribute.
> The problem is that it only allows you to specify one jar that implements the
> coprocessor. You will need to either bundle all the dependencies into this
> jar, or you will need to copy the dependencies into HBase lib dir.
> The first option may not be ideal sometimes. The second choice can be
> troublesome too, particularly when the hbase region sever node and dirs are
> dynamically added/created.
> There are a couple things we can expand here. We can allow the coprocessor
> attribute to specify a directory location, probably on hdfs.
> We may even allow some wildcard in there.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
