[
https://issues.apache.org/jira/browse/HBASE-28534?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17838889#comment-17838889
]
Nikita Pande commented on HBASE-28534:
--------------------------------------
[~junegunn] I validated your patch and it works.
>>>That is because {{applyCreds}} changes the authentication method of the
>>>current user to {{{}PROXY{}}}, when it should still be {{KERBEROS}} for
>>>local mode to run correctly.
Only thing what I was mentioning is that I get error without this patch as " No
matching SASL authentication provider and supporting token found from providers
for user: **
(auth:KERBEROS) "and not auth: PROXY
> Authentication failure when running hbase-spark in local mode
> -------------------------------------------------------------
>
> Key: HBASE-28534
> URL: https://issues.apache.org/jira/browse/HBASE-28534
> Project: HBase
> Issue Type: Bug
> Components: spark
> Affects Versions: connector-1.0.0
> Reporter: Junegunn Choi
> Assignee: Junegunn Choi
> Priority: Major
>
> h2. Problem
> When running Spark in local mode, hbase-spark fails to authenticate to a
> Kerberos secured HBase cluster. The error message is:
> {quote}No matching SASL authentication provider and supporting token found
> from providers for user: x...@xxx.xxx (auth:PROXY)
> {quote}
> That is because {{applyCreds}} changes the authentication method of the
> current user to {{{}PROXY{}}}, when it should still be {{KERBEROS}} for local
> mode to run correctly.
> h2. Suggested solution
> To fix this, I propose removing {{{}applyCreds{}}}. The function is no longer
> needed and should be removed:
> 1. Because we should not change the authentication mode of the current user
> in local mode
> 2. And because the purpose of the function is no longer valid. It is not
> doing anything meaningful since the broadcasting of the user credentials was
> removed in this commit:
> [https://github.com/apache/hbase-connectors/commit/75e41365207408f5b47d5925469a49fd60078b5e]
> A pull request is on the way.
> h2. Testing
> The fix was manually tested against Kerberos secured HBase 2.4.17 + Hadoop
> 3.3.5 cluster using Spark 3.5.1, both in local mode and with Yarn master,
> running the following Python code.
> {code:java}
> df = (spark.read.format("org.apache.hadoop.hbase.spark")
> .option("hbase.columns.mapping", "key STRING :key, state
> STRING info:state")
> .option("hbase.table", "hbase:meta").load())
> df.first()
> {code}
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
