[
https://issues.apache.org/jira/browse/HAWQ-1788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17303100#comment-17303100
]
Yuan Zhou commented on HAWQ-1788:
---------------------------------
[~chiyang10000] thanks for the hints! I understand how it's passed to hdfs
session now.
my deployment is a bit different, it's using a custom libhdfs3 fork from
conda-forge (the code will be regularly synced with the libhdfs component in
HAWQ).
[https://anaconda.org/conda-forge/libhdfs3]
[https://github.com/martindurant/libhdfs3-downstream]
As the code is not a "complete" set, it seems like no easy way to setup the
correct configurations(?) except changing the default configuration in codes.
Do you think if it's OK to set the default path here? This maybe unnecessary
for HAWQ setup but can help to test with the standalone libhdfs3 project in
conda-forge
[https://github.com/apache/hawq/pull/1510/files]
> providing non-empty value for dfs.domain.socket.path in libhdfs
> ---------------------------------------------------------------
>
> Key: HAWQ-1788
> URL: https://issues.apache.org/jira/browse/HAWQ-1788
> Project: Apache HAWQ
> Issue Type: Improvement
> Components: libhdfs
> Reporter: Yuan Zhou
> Assignee: Ruilong Huo
> Priority: Major
>
> Hi HAWQ developers,
> trying to enable the short-circuit read feature here and i found the default
> value for dfs.domain.socket.path is empty string
> [https://github.com/apache/hawq/blob/master/depends/libhdfs3/src/common/SessionConfig.cpp#L151]
> it looks better to provide an aligned value with the documents from HDFS:
> [https://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-hdfs/ShortCircuitLocalReads.html]
> This would allow users to test against vanilla HDFS setup quickly.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
