[HDDS-2218] Use OZONE_CLASSPATH instead of HADOOP_CLASSPATH

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Done
• Fix Version/s: 1.1.0
• Component/s: docker
• Labels: TriagePending newbie
• Assignee: Sandeep Nemuri
• Reporter: Marton Elek
• Created: 01/Oct/19 13:38
• Updated: 21/Jan/24 15:43
• Resolved: 01/Jun/23 20:39

Description

HADOOP_CLASSPATH is the standard way to add additional jar files to the classpath of the mapreduce/spark/.. .jobs. If something is added to the HADOOP_CLASSPATH, than it should be on the classpath of the classic hadoop daemons.

But for the Ozone components we don't need any new jar files (cloud connectors, libraries). I think it's more safe to separated HADOOP_CLASSPATH from OZONE_CLASSPATH. If something is really need on the classpath for Ozone daemons the dedicated environment variable should be used.

 

Most probably it can be fixed in

hadoop-hdds/common/src/main/bin/hadoop-functions.sh

And the hadoop-ozone/dev/src/main/compose files also should be checked (some of them contain HADOOP_CLASSPATH

Issue Links

• is fixed by: HDDS-4525 Replace Hadoop variables and functions in Ozone shell scripts with Ozone-specific ones