HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers #2

apotheque · 2025-03-15T14:46:45Z

Pipeline and container filtering based on datacenter binding.

ivanzlenko · 2025-03-18T07:23:37Z

hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/PipelineRequestInformation.java

+     * @return Builder for PipelineRequestInformation
+     */
+    public Builder setDatacenters(Set<String> datacenters) {
+      this.datacenters = datacenters;


I would've made a copy of the set here.

ivanzlenko · 2025-03-18T07:29:19Z

hadoop-hdds/common/src/main/resources/ozone-default.xml

+      This setting maps each DataNode to its data center using a comma-separated list of
+      hostname:ratis_port=*dc_name* pairs. It enables Ozone to associate DataNodes with specific data centers,


Suggested change

This setting maps each DataNode to its data center using a comma-separated list of

hostname:ratis_port=*dc_name* pairs. It enables Ozone to associate DataNodes with specific data centers,

This property defines a mapping that associates each DataNode with its corresponding data center. The mapping is specified as a comma-separated list of pairs in the format `hostname:ratis_port=dc_name`. By establishing this association, Ozone can identify which data center a DataNode belongs to, enabling more efficient storage management and data placement across data centers.

ivanzlenko · 2025-03-18T07:31:57Z

...ds/framework/src/main/java/org/apache/hadoop/hdds/scm/protocol/ScmBlockLocationProtocol.java

  List<AllocatedBlock> allocateBlock(long size, int numBlocks,
      ReplicationConfig replicationConfig, String owner,
-      ExcludeList excludeList, String clientMachine) throws IOException;
+      ExcludeList excludeList, String clientMachine, String datacenters) throws IOException;


Mark it Nullable

ivanzlenko · 2025-03-18T07:32:25Z

...va/org/apache/hadoop/hdds/scm/protocolPB/ScmBlockLocationProtocolClientSideTranslatorPB.java

      ReplicationConfig replicationConfig,
      String owner, ExcludeList excludeList,
-      String clientMachine
+      String clientMachine, String datacenters


Mark it Nullable

ivanzlenko · 2025-03-18T07:32:47Z

hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/block/BlockManager.java

  AllocatedBlock allocateBlock(long size, ReplicationConfig replicationConfig,
      String owner,
-      ExcludeList excludeList) throws IOException, TimeoutException;
+      ExcludeList excludeList, String datacenters) throws IOException, TimeoutException;


Mark it Nullable

ivanzlenko · 2025-03-18T07:32:59Z

hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/block/BlockManagerImpl.java

  public AllocatedBlock allocateBlock(final long size,
      ReplicationConfig replicationConfig,
-      String owner, ExcludeList excludeList)
+      String owner, ExcludeList excludeList, String datacenters)


Mark it Nullable

ivanzlenko · 2025-03-18T07:36:12Z

...s/server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/WritableContainerFactory.java


  public ContainerInfo getContainer(final long size,
-      ReplicationConfig repConfig, String owner, ExcludeList excludeList)
+      ReplicationConfig repConfig, String owner, ExcludeList excludeList, String datacenters)


Mark as Nullable

ivanzlenko · 2025-03-18T07:36:27Z

.../server-scm/src/main/java/org/apache/hadoop/hdds/scm/pipeline/WritableContainerProvider.java

   */
  ContainerInfo getContainer(long size, T repConfig,
-      String owner, ExcludeList excludeList)
+      String owner, ExcludeList excludeList, String datacenters)


Mark as Nullable

ivanzlenko · 2025-03-18T07:39:28Z

hadoop-hdds/interface-server/src/main/proto/ScmServerProtocol.proto

  optional hadoop.hdds.ECReplicationConfig ecReplicationConfig = 8;

  optional string client = 9;
+  optional string datacenters = 10;


Why pass it as a string and not as repeatable?

This reverts commit 7e78726.

…ified dcs

Cyrill · 2025-03-19T06:30:43Z

hadoop-ozone/tools/src/main/java/org/apache/hadoop/ozone/shell/bucket/CreateBucketHandler.java

    }

+    if (datacenters != null && !datacenters.isEmpty()) {
+      bb.addMetadata(OzoneConsts.DATACENTERS, datacenters);


the value should be basically validated

Cyrill · 2025-03-19T06:35:12Z

hadoop-hdds/common/src/main/resources/ozone-default.xml

+    <tag>SCM</tag>
+    <description>
+      This setting maps each DataNode to its data center using a comma-separated list of
+      hostname:ratis_port=*dc_name* pairs. It enables Ozone to associate DataNodes with specific data centers,


Ratis port is different each time. It's not feasible to keep it in settings

…ontainers (#2)

apotheque added 14 commits March 15, 2025 11:59

UD-635. Added datacenter info to datanodes, buckets and containers

4378de2

Pipeline and container filtering based on datacenter binding.

Fix for no DC mapping provided

7215507

Fix null datacenters property in protobuf

3ea5149

Fix datacenters param mock

2bc501c

Fix reading null datacenters from protobuf

430de38

Fix some tests

62d086d

Added missing xml configuration

47177aa

Check omBucketInfo is null

2c99129

Fix reading null datacenters from protobuf Cyrill#2

80dc31a

Datacenter based policies for pipelines

a5e71d7

Fix pmd

2dc4267

Fix tests

8f11ddd

Make tests pass again

a135bec

Set 1.4 branch as base

faec0bc

apotheque added a commit to apotheque/ozone that referenced this pull request Mar 15, 2025

Fix reading null datacenters from protobuf Cyrill#2

d154073

apotheque added 2 commits March 15, 2025 20:33

Fix tests

a52ba77

Choose random pipeline

6afdba7

Cyrill changed the title ~~HDDS-10000 UD-635. Added datacenter info to datanodes, buckets and containers~~ HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers Mar 18, 2025

ivanzlenko suggested changes Mar 18, 2025

View reviewed changes

apotheque added 4 commits March 18, 2025 13:58

Disable BackgroundPipelineCreator

7e78726

Return null if there is no suitable pipeline to choose from

d7d8cf2

Revert "Disable BackgroundPipelineCreator"

6378878

This reverts commit 7e78726.

Force BackgroundPipelineCreator to create pipelines bound to all spec…

eb48107

…ified dcs

Cyrill reviewed Mar 19, 2025

View reviewed changes

apotheque added 5 commits March 19, 2025 12:11

Validate bucket datacenters option with regexp

578c7c7

Use repeated type for datacenters in AllocateScmBlockRequestProto

fd13563

Copy datacenters set in PipelineRequestInformation.Builder

960ecb6

Better description for ozone.scm.dc.datanode.mapping property

6ea6503

Fix tests

60d37de

apotheque added 2 commits March 19, 2025 21:06

Fix checkstyle

2c2ada1

Fix tests

f351fbe

Cyrill approved these changes Mar 20, 2025

View reviewed changes

Cyrill merged commit ade3f33 into Cyrill:ozone-1.4 Mar 20, 2025
37 checks passed

Cyrill pushed a commit that referenced this pull request Sep 22, 2025

HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and c…

9c7ec23

…ontainers (#2)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers #2

HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers #2

Uh oh!

apotheque commented Mar 15, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

ivanzlenko Mar 18, 2025

Uh oh!

Cyrill Mar 19, 2025

Uh oh!

Cyrill Mar 19, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		This setting maps each DataNode to its data center using a comma-separated list of
		hostname:ratis_port=dc_name pairs. It enables Ozone to associate DataNodes with specific data centers,

	This setting maps each DataNode to its data center using a comma-separated list of
	hostname:ratis_port=dc_name pairs. It enables Ozone to associate DataNodes with specific data centers,
	This property defines a mapping that associates each DataNode with its corresponding data center. The mapping is specified as a comma-separated list of pairs in the format `hostname:ratis_port=dc_name`. By establishing this association, Ozone can identify which data center a DataNode belongs to, enabling more efficient storage management and data placement across data centers.

HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers #2

HDDS-10000. UD-635. Added datacenter info to datanodes, buckets and containers #2

Uh oh!

Conversation

apotheque commented Mar 15, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants