[
https://issues.apache.org/jira/browse/HDDS-1115?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16777081#comment-16777081
]
Elek, Marton commented on HDDS-1115:
------------------------------------
Thanks the feedback [~ste...@apache.org], unfortunately I am not sure If I
understood well.
Let me explain the current state of the ozone build:
We have no dependencies between (./hadoop-ozone,./hadoop-hdds) and
./hadoop-common any more. Ozone/Hdds depends on the released hadoop-3.2.0.
The main reason behind this decision is to make it possible to upload ozone
artifacts to maven. Without depending on fix hadoop release we can't upload our
artifacts as they would depend on a snapshot/pseudo-released hadoop common.
The next ozone release will contain only ./hadoop-hdds and ./hadoop-ozone bits
+ hadoop 3.2. So actually no risk to "do a release of the ozone artifacts
alone", because that's the goal.
There is one big limitation of this approach: we need a new hadoop-common
release if something would be changed in the common. That's something what we
need to manager.
This approach also means that sooner or later we need an ozone specific source
release. As Ozone requires hadoop-3.2 we don't need to add
hadoop-(common|yarn|hdfs) 3.3-SNAPSHOT artifacts to the ozone release tarball.
For hadoop releases we agreed during the merge vote that all the ozone bits
will be removed from the stable hadoop releases. This is what we did with
HADOOP-15791 HADOOP-15857. We should do the same after a 3.3 branch cut and
that time this pom.ozone.xml also will be deleted from the ozone release.
Does it answer your question? Please let me know if you still have any concerns.
(ps: the only 3.3-SNAPSHOT bits which are used by ozone are the
libexec/hadoop-common shell script fragments. They are not (yet) uploaded as
artifacts so we copy them directly from the source tree...)
> Provide ozone specific top-level pom.xml
> ----------------------------------------
>
> Key: HDDS-1115
> URL: https://issues.apache.org/jira/browse/HDDS-1115
> Project: Hadoop Distributed Data Store
> Issue Type: Improvement
> Components: build
> Reporter: Elek, Marton
> Assignee: Elek, Marton
> Priority: Major
> Labels: pull-request-available
> Fix For: 0.4.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Ozone build process doesn't require the pom.xml in the top level hadoop
> directory as we use hadoop 3.2 artifacts as parents of hadoop-ozone and
> hadoop-hdds. The ./pom.xml is used only to include the
> hadoop-ozone/hadoop-hdds projects in the maven reactor.
> From command line, it's easy to build only the ozone artifacts:
> {code}
> mvn clean install -Phdds -am -pl :hadoop-ozone-dist
> -Danimal.sniffer.skip=true -Denforcer.skip=true
> {code}
> Where: '-pl' defines the build of the hadoop-ozone-dist project
> and '-am' defines to build all of the dependencies from the source tree
> (hadoop-ozone-common, hadoop-hdds-common, etc.)
> But this filtering is available only from the command line.
> With providing a lightweight pom.ozone.xml we can achieve the same:
> * We can open only hdds/ozone projects in the IDE/intellij. It makes the
> development faster as IDE doesn't need to reindex all the sources all the
> time + it's easy to execute checkstyle/findbugs plugins of the intellij to
> the whole project.
> * Longer term we should create an ozone specific source artifact (currently
> the source artifact for hadoop and ozone releases are the same) which also
> requires a simplified pom.
> In this patch I also added the .mvn directory to the .gitignore file.
> With
> {code}
> mkdir -p .mvn && echo "-f ozone.pom.xml" > .mvn/maven.config" you can persist
> the usage of the ozone.pom.xml for all the subsequent builds (in the same dir)
> How to test?
> Just do a 'mvn -f ozonze.pom.xml clean install -DskipTests'
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
