Skip to content

Conversation

@SaketaChalamchala
Copy link
Contributor

@SaketaChalamchala SaketaChalamchala commented May 6, 2025

What changes were proposed in this pull request?

Right now s3 client can create keys/files with only RATIS replication type using the header x-amz-storage-class=STANDARD(RATIS THREE) or REDUCED_REDUNDANCY(RATIS ONE).

Proposed solution allows creation of EC keys/files by supporting x-amz-storage-class=STANDARD_IA(EC/RS-3-2-1024k by default) along with a custom meta header to configure EC replication strings like x-amz-meta-storage-config=rs-6-3-1024k.
Also, changed the replication config resolution order to be consistent with ozone cli which is

  1. Replication config provided by S3 client.
  2. Replication config provided at client side config.
  3. Bucket replication.
  4. Server side default replication config.

What is the link to the Apache JIRA

https://issues.apache.org/jira/browse/HDDS-10979

How was this patch tested?

Unit Tests
Acceptance test

@ivandika3 ivandika3 added the s3 S3 Gateway label May 6, 2025
@jojochuang jojochuang requested a review from chungen0126 May 8, 2025 00:17
public static S3StorageType fromReplicationConfig(ReplicationConfig config) {
if (config instanceof ECReplicationConfig) {
return S3StorageType.STANDARD;
return STANDARD_IA;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

map EC type to STANDARD_IA.

Copy link
Contributor Author

@SaketaChalamchala SaketaChalamchala May 14, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is done here. STANDARD_IA is mapped to EC/RS-3-2-1024k by default.


private Set<String> excludeMetadataFields =
new HashSet<>(Arrays.asList(OzoneConsts.GDPR_FLAG));
new HashSet<>(Arrays.asList(OzoneConsts.GDPR_FLAG, STORAGE_CONFIG_HEADER));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

what does this mean

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We store any x-amz-meta-key=value for an object in KeyArgs.metadata (or ex., Etag). ^^This will exclude storing EC storage config in the metadata because it will be available in replication config. We only want to use this header to determine EC replication config when client uses STANDARD_IA storage class.


copyHeader = headers.getHeaderString(COPY_SOURCE_HEADER);
storageType = headers.getHeaderString(STORAGE_CLASS_HEADER);
storageConfig = headers.getHeaderString(CUSTOM_METADATA_HEADER_PREFIX + STORAGE_CONFIG_HEADER);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

curious: how will a client add the header to the request? is it supported by s3 cli or hadoop aws cloud connector?

i wonder where we add this information in the developer doc.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All s3 clients should support it. They would just need to add the "x-amz-metadata-storage-class=" to their requests.
This is demonstrated in the acceptance test with AWS cli which does the above underneath.

"INVALID", null, RATIS3REPLICATIONCONFIG, RATIS1REPLICATIONCONFIG));
assertThrows(OS3Exception.class, () -> S3Utils.
resolveS3ClientSideReplicationConfig(S3StorageType.STANDARD_IA.name(),
"INVALID", RATIS3REPLICATIONCONFIG, RATIS1REPLICATIONCONFIG));
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

shall we also verify for invalid bucketReplConfig?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

also invalid s3StorageTypeHeader

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How would we pass an invalid ReplicationConfig here? Wouldn't the creation of an invalid ReplicationConfig itself throw an error?

}

public static S3StorageType toS3StorageType(String storageType)
public static ReplicationConfig toReplicationConfig(String storageType, String storageConfig)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

no test

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All tests for toReplicationConfig() are covered by testing resolveS3ClientSideReplicationConfig in TestS3Utils.java

@SaketaChalamchala SaketaChalamchala marked this pull request as ready for review May 14, 2025 17:19
@SaketaChalamchala
Copy link
Contributor Author

@ivandika3 @kerneltime @adoroszlai could you also take a look?

@ivandika3 ivandika3 requested a review from xichen01 May 15, 2025 14:53
Copy link
Contributor

@chungen0126 chungen0126 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @SaketaChalamchala for the patch. Left some comments.

throws OS3Exception {
try {
return S3StorageType.valueOf(storageType);
if (S3StorageType.STANDARD_IA.name().equals(s3StorageType)) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggest doing valueOf() first, then check the enum makes the logic clearer.

try {
      S3StorageType storageType = S3StorageType.valueOf(s3StorageType);
      if (S3StorageType.STANDARD_IA.equals(storageType)) {
        return (!StringUtils.isEmpty(s3StorageConfig)) ? new ECReplicationConfig(s3StorageConfig) :
            new ECReplicationConfig(S3StorageType.STANDARD_IA.getEcReplicationString());
      } else {
        return ReplicationConfig.fromProtoTypeAndFactor(
            ReplicationType.toProto(storageType.getType()),
            ReplicationFactor.toProto(storageType.getFactor()));
      }

@chungen0126
Copy link
Contributor

Please take a look at BucketEndpoint#addKey. For objects using EC, listObjects still returns the storage class as STANDARD.

${ENDPOINT_URL} http://s3g:9878
${BUCKET} generated

*** Test Cases ***
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe add a ozone sh command to validate the replication configuration?

Copy link
Contributor

@adoroszlai adoroszlai left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @SaketaChalamchala for the patch.

@SaketaChalamchala
Copy link
Contributor Author

Thanks @chungen0126 @kerneltime and @adoroszlai. Please also review the updated patch.

Copy link
Contributor

@adoroszlai adoroszlai left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks @SaketaChalamchala for updating the patch.

@adoroszlai
Copy link
Contributor

@chungen0126 please take another look

Copy link
Contributor

@chungen0126 chungen0126 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

+1 LGTM

@adoroszlai adoroszlai merged commit a88ff71 into apache:master May 21, 2025
43 checks passed
@adoroszlai
Copy link
Contributor

Thanks @SaketaChalamchala for the patch, @chungen0126, @jojochuang, @kerneltime for the review.

Tejaskriya pushed a commit to Tejaskriya/ozone that referenced this pull request May 29, 2025
aswinshakil added a commit to aswinshakil/ozone that referenced this pull request Jun 9, 2025
…239-container-reconciliation

Commits: 80 commits
5e273a4 HDDS-12977. Fail build on dependency problems (apache#8574)
5081ba2 HDDS-13034. Refactor DirectoryDeletingService to use ReclaimableDirFilter and ReclaimableKeyFilter (apache#8546)
e936e4d HDDS-12134. Implement Snapshot Cache lock for OM Bootstrap (apache#8474)
31d13de HDDS-13165. [Docs] Python client developer guide. (apache#8556)
9e6955e HDDS-13205. Bump common-custom-user-data-maven-extension to 2.0.3 (apache#8581)
750b629 HDDS-13203. Bump Bouncy Castle to 1.81 (apache#8580)
ba5177e HDDS-13202. Bump build-helper-maven-plugin to 3.6.1 (apache#8579)
07ee5dd HDDS-13204. Bump awssdk to 2.31.59 (apache#8582)
e1964f2 HDDS-13201. Bump jersey2 to 2.47 (apache#8578)
81295a5 HDDS-13013. [Snapshot] Add metrics and tests for snapshot operations. (apache#8436)
b3d75ab HDDS-12976. Clean up unused dependencies (apache#8521)
e0f08b2 HDDS-13179. rename-generated-config fails on re-compile without clean (apache#8569)
f388317 HDDS-12554. Support callback on completed reconfiguration (apache#8391)
c13a3fe HDDS-13154 Link more Grafana dashboard json files to the Observability user doc (apache#8533)
2a761f7 HDDS-11967. [Docs]DistCP Integration in Kerberized environment. (apache#8531)
81fc4c4 HDDS-12550. Use DatanodeID instead of UUID in NodeManager CommandQueue. (apache#8560)
2360af4 HDDS-13169. Intermittent failure in testSnapshotOperationsNotBlockedDuringCompaction (apache#8553)
f19789d HDDS-13170. Reclaimable filter should always reclaim entries when buckets and volumes have already been deleted (apache#8551)
315ef20 HDDS-13175. Leftover reference to OM-specific trash implementation (apache#8563)
902e715 HDDS-13159. Refactor KeyManagerImpl for getting deleted subdirectories and deleted subFiles (apache#8538)
46a93d0 HDDS-12817. Addendum rename ecIndex to replicaIndex in chunkinfo output (apache#8552)
19b9b9c HDDS-13166. Set pipeline ID in BlockExistenceVerifier to avoid cached pipeline with different node (apache#8549)
b3ff67c HDDS-13068. Validate Container Balancer move timeout and replication timeout configs (apache#8490)
7a7b9a8 HDDS-13139. Introduce bucket layout flag in freon rk command (apache#8539)
3c25e7d HDDS-12595. Add verifier for container replica states (apache#8422)
6d59220 HDDS-13104. Move auditparser acceptance test under debug (apache#8527)
8e8c432 HDDS-13071. Documentation for Container Replica Debugger Tool (apache#8485)
0e8c8d4 HDDS-13158. Bump junit to 5.13.0 (apache#8537)
8e552b4 HDDS-13157. Bump exec-maven-plugin to 3.5.1 (apache#8534)
168f690 HDDS-13155. Bump jline to 3.30.4 (apache#8535)
cc1e4d1 HDDS-13156. Bump awssdk to 2.31.54 (apache#8536)
3bfb7af HDDS-13136. KeyDeleting Service should not run for already deep cleaned snapshots (apache#8525)
006e691 HDDS-12503. Compact snapshot DB before evicting a snapshot out of cache (apache#8141)
568b228 HDDS-13067. Container Balancer delete commands should not be sent with an expiration time in the past (apache#8491)
53673c5 HDDS-11244. OmPurgeDirectoriesRequest should clean up File and Directory tables of AOS for deleted snapshot directories (apache#8509)
07f4868 HDDS-13099. ozone admin datanode list ignores --json flag when --id filter is used (apache#8500)
08c0ab8 HDDS-13075. Fix default value in description of container placement policy configs (apache#8511)
58c87a8 HDDS-12177. Set runtime scope where missing (apache#8513)
10c470d HDDS-12817. Add EC block index in the ozone debug replicas chunk-info (apache#8515)
7027ab7 HDDS-13124. Respect config hdds.datanode.use.datanode.hostname when reading from datanode (apache#8518)
b8b226c HDDS-12928. datanode min free space configuration (apache#8388)
fd3d70c HDDS-13026. KeyDeletingService should also delete RenameEntries (apache#8447)
4c1c6cf HDDS-12714. Create acceptance test framework for debug and repair tools (apache#8510)
fff80fc HDDS-13118. Remove duplicate mockito-core dependency from hdds-test-utils (apache#8508)
10d5555 HDDS-13115. Bump awssdk to 2.31.50 (apache#8505)
360d139 HDDS-13017. Fix warnings due to non-test scoped test dependencies (apache#8479)
1db1cca HDDS-13116. Bump jline to 3.30.3 (apache#8504)
322ca93 HDDS-13025. Refactor KeyDeletingService to use ReclaimableKeyFilter (apache#8450)
988b447 HDDS-5287. Document S3 ACL classes (apache#8501)
64bb29d HDDS-12777. Use module-specific name for generated config files (apache#8475)
54ed115 HDDS-9210. Update snapshot chain restore test to incorporate snapshot delete. (apache#8484)
87dfa5a HDDS-13014. Improve PrometheusMetricsSink#normalizeName performance (apache#8438)
7cdc865 HDDS-13100. ozone admin datanode list --json should output a newline at the end (apache#8499)
9cc4194 HDDS-13089. [snapshot] Add an integration test to verify snapshotted data can be read by S3 SDK client (apache#8495)
cb9867b HDDS-13065. Refactor SnapshotCache to return AutoCloseSupplier instead of ReferenceCounted (apache#8473)
a88ff71 HDDS-10979. Support STANDARD_IA S3 storage class to accept EC replication config (apache#8399)
6ec8f85 HDDS-13080. Improve delete metrics to show number of timeout DN command from SCM (apache#8497)
3bb8858 HDDS-12378. Change default hdds.scm.safemode.min.datanode to 3 (apache#8331)
0171bef HDDS-13073. Set pipeline ID in checksums verifier to avoid cached pipeline with different node (apache#8480)
5c7726a HDDS-11539. OzoneClientCache `@PreDestroy` is never called (apache#8493)
a8ed19b HDDS-13031. Implement a Flat Lock resource in OzoneManagerLock (apache#8446)
e9e8b30 HDDS-12935. Support unsigned chunked upload with STREAMING-UNSIGNED-PAYLOAD-TRAILER (apache#8366)
7590268 HDDS-13079. Improve logging in DN for delete operation. (apache#8489)
435fe7e HDDS-12870. Fix listObjects corner cases (apache#8307)
eb5dabd HDDS-12926. Remove *.tmp.* exclusion in DU (apache#8486)
eeb98c7 HDDS-13030. Snapshot Purge should unset deep cleaning flag for next 2 snapshots in the chain (apache#8451)
6bf121e HDDS-13032. Support proper S3OwnerId representation (apache#8478)
5d1b43d HDDS-13076. Refactor OzoneManagerLock class to rename Resource class to LeveledResource (apache#8482)
bafe6d9 HDDS-13064. [snapshot] Add test coverage for SnapshotUtils.isBlockLocationInfoSame() (apache#8476)
7035846 HDDS-13040. Add user doc highlighting the difference between Ozone ACL and S3 ACL. (apache#8457)
1825cdf HDDS-13049. Deprecate VolumeName & BucketName in OmKeyPurgeRequest and prevent Key version purge on Block Deletion Failure (apache#8463)
211c76c HDDS-13060. Change NodeManager.addDatanodeCommand(..) to use DatanodeID (apache#8471)
f410238 HDDS-13061. Add test for key ACL operations without permission (apache#8472)
d1a2f48 HDDS-13057. Increment block delete processed transaction counts regardless of log level (apache#8466)
0cc6fcc HDDS-13043. Replace != with assertNotEquals in TestSCMContainerPlacementRackAware (apache#8470)
e1c779a HDDS-13051. Use DatanodeID in server-scm. (apache#8465)
35e1126 HDDS-13042. [snapshot] Add future proofing test cases for unsupported file system API (apache#8458)
619c05d HDDS-13008. Exclude same SST files when calculating full snapdiff (apache#8423)
21b49d3 HDDS-12965. Fix warnings about "used undeclared" dependencies (apache#8468)
8136119 HDDS-13048. Create new module for Recon integration tests (apache#8464)

Conflicts:
	hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/node/NodeManager.java
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

s3 S3 Gateway

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants