-
Notifications
You must be signed in to change notification settings - Fork 587
HDDS-3858. Remove support to start Ozone and HDFS datanodes in the same JVM #1117
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
Ping everybody (especially @jnp @arp7 @xiaoyuyao @mukul1987 @nandakumar131 ) |
|
This is a double edged sword :) |
|
The option to run them in same JVM was allowed, to support in-place upgrade from HDFS to Ozone in future. It was assumed that having the two Datanodes in the same JVM will certainly simplify Hdfs to Ozone upgrade. |
|
@elek thanks for bringing this up. Given the fact that it is broken currently, it makes sense to remove the document saying it is supported. Let's open a separate JIRA to track the issues such as classpath, etc. that prevent running HDDS datanode as plugin inside HDFS datanode JVM. They need to be fixed when we revisit in-place upgrade. |
@jnp Base on the last design in-place upgrade can work with different JVM on the same host. And it's possible to run HDFS and Ozone datanodes on the same node. My problem is that maintaining this feature requires significant effort. If we would like to do it, it's a release blocker:
We can call it as a broken feature ;-) |
|
/pending "Blocking merge ability as we definitely need an agreement first. Possible topic for Community Meeting." |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Marking this issue as un-mergeable as requested.
Please use /ready comment when it's resolved.
"Blocking merge ability as we definitely need an agreement first. Possible topic for Community Meeting."
I agree. |
|
+1 for documenting that this is not supported. |
|
+1, it will make Ozone repo clearly. |
Yes, it's important that I wouldn't like to modify interfaces or code. Just remove visibility (remove docs + compose examples). We can make it work any time, or we can provide specific plugin for in-place upgrade. It seems that we have an agreement that we can hide this feature, and we can further discuss long-term strategies before any code change. As we have the consensus, I will merge this PR, soon. |
|
/ready |
Blocking review request is removed.
* upstream/master: (56 commits) HDDS-3264. Fix TestCSMMetrics.java. (apache#1120) HDDS-3858. Remove support to start Ozone and HDFS datanodes in the same JVM (apache#1117) HDDS-3704. Update all the documentation to use ozonefs-hadoop2/3 instead of legacy/current (apache#1099) HDDS-3773. Add OMDBDefinition to define structure of om.db. (apache#1076) Revert "HDDS-3263. Fix TestCloseContainerByPipeline.java. (apache#1119)" (apache#1126) HDDS-3821. Disable Ozone SPNEGO should not fall back to hadoop.http.a… (apache#1101) HDDS-3819. OzoneManager#listVolumeByUser ignores userName parameter when ACL is enabled (apache#1087) HDDS-3779. Add csi interface documents to show how to use ozone csi (apache#1059) HDDS-3857. Datanode in compose/ozonescripts can't be started (apache#1116) HDDS-3430. Enable TestWatchForCommit test cases. (apache#1114) HDDS-3263. Fix TestCloseContainerByPipeline.java. (apache#1119) HDDS-3512. s3g multi-part-upload saved incorrect content using streaming (apache#1092) HDDS-3836. Modify ContainerPlacementPolicyFactory JavaDoc (apache#1097) HDDS-3780. Replace the imagePullPolicy from always to IfNotPresent (apache#1055) HDDS-3847. Change OMNotLeaderException logging to DEBUG (apache#1118) HDDS-3745. Improve OM and SCM performance with 64% by avoid collect datanode information to s3g (apache#1031) HDDS-3286. BasicOzoneFileSystem support batchDelete. (apache#814) HDDS-3850. Update the admin document to let user know how to show the status of all rules. (apache#1109) HDDS-3848. Add ratis.thirdparty.version in main pom.xml (apache#1108) HDDS-3815. Avoid buffer copy in ContainerCommandRequestProto. (apache#1085) ...
What changes were proposed in this pull request?
With a few thousands issues ago Ozone was integral part of Hadoop/HDFS project. At that time there were two options to start datanode:
Start in a separated JVM
Start in the same JVM with the HDFS
Today only 1 is the standard way, this is tested and working. 2nd is not working any more but still documented.
I propose to drop the support of this use case as I can't see any benefit to support it anymore:
What is the link to the Apache JIRA
https://issues.apache.org/jira/browse/HDDS-3858
How was this patch tested?
Not required. One compose folder + docs are removed.