Repository navigation
HDDS-16666. Flush SCM transaction in memory during apply transaction - #11409
Merged
Merged
Conversation
Contributor
There was a problem hiding this comment.
Copilot review overview
🟡 Changes recommended
The state-machine test helper repeatedly registers global SCM metrics without unregistering them, potentially causing duplicate-source test failures.
Review effort: Balanced
Findings: 1
Open (1)
What changed in this PR
Adds count-based flushing to bound SCM HA transaction-buffer memory during Ratis backlog replay.
Changes:
- Adds a configurable pending-operation flush limit with validation.
- Flushes during transaction application after recording transaction metadata.
- Adds coverage for flushing, failures, concurrency, and ordering.
| File | Description |
|---|---|
hadoop-hdds/common/src/main/java/org/apache/hadoop/hdds/scm/ScmConfigKeys.java |
Defines the new configuration key and default. |
hadoop-hdds/common/src/main/resources/ozone-default.xml |
Documents the configuration property. |
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/ha/SCMHADBTransactionBuffer.java |
Adds the conditional-flush API. |
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/ha/SCMHADBTransactionBufferImpl.java |
Implements validated count-based flushing. |
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/ha/SCMHADBTransactionBufferStub.java |
Implements the new interface method for the stub. |
hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/ha/SCMStateMachine.java |
Checks the flush threshold after applying transactions. |
hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/ha/TestSCMHADBTransactionBufferImpl.java |
Tests threshold, persistence, failure, and concurrency behavior. |
hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/ha/TestSCMHATransactionBufferMonitorTask.java |
Supplies configuration required by the updated constructor. |
hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/ha/TestSCMStateMachine.java |
Tests state-machine flush ordering and failure handling. |
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Russole
self-requested a review
October 5, 2026 23:37
Contributor
Author
|
Thanks @sumitagrawl and @yandrey321 for the review. |
errose28
added a commit
that referenced
this pull request
Oct 6, 2026
* master: (64 commits) HDDS-16716. Add description for ozone.scm.ec.pipeline.per.volume.factor (#11415) HDDS-16008. PutBlocks from Flushes also go without Raft (#11356) HDDS-16666. Flush SCM transaction in memory during apply transaction (#11409) HDDS-15749. Run specific JUnit tests if possible (#10671) HDDS-16362. GetObjectAttributes ObjectParts should return Part entries for FSO buckets (#11242). HDDS-16643. Remove CleanupTableInfo mechanism (#11365) HDDS-16721. StreamBlockInputStream.read() returns a negative value for bytes 0x80 to 0xFF (#11411) HDDS-16241. gRPC deadline kills long-lived block streams after 30 seconds and the client never recovers (#11080) HDDS-15991. Speed up deleted table scans in quota repair (#11386) HDDS-16674. Bump awssdk to 2.55.6 (#11407) HDDS-16673. Avoid redundant ListBuckets RPCs when S3 bucket listing reaches the end (#11397) HDDS-16708. Let dependabot ignore iceberg minor version upgrades (#11398) HDDS-16713. Bump develocity-maven-extension to 2.6.0 (#11405) HDDS-16300. Allow OM to dynamically reconfigure its SCM node list without a restart (#11218) HDDS-15089. Support S3 per request read consistency (#11252) HDDS-16704. ReadBlock fails with IllegalStateException when a response is shorter than responseDataSize (#11402) HDDS-16631. Fix chooseRandom for rack names with common prefixes (#11401) HDDS-16654. Replace usage of deprecated finalize() in OM (#11376) HDDS-16658. Reuse source key details when opening input stream in S3 CopyObject (#11396) HDDS-16711. Bump moment to 2.31.0 (#11373) ... Conflicts: hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/node/HealthyReadOnlyNodeHandler.java hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/node/NodeStateManager.java hadoop-hdds/server-scm/src/main/java/org/apache/hadoop/hdds/scm/node/SCMNodeManager.java hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/ha/TestSCMStateMachine.java hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/node/TestDeadNodeHandler.java hadoop-hdds/server-scm/src/test/java/org/apache/hadoop/hdds/scm/node/TestNodeStateManager.java hadoop-ozone/client/src/test/java/org/apache/hadoop/ozone/client/rpc/TestRpcClient.java hadoop-ozone/common/src/main/java/org/apache/hadoop/ozone/OmUtils.java hadoop-ozone/common/src/test/java/org/apache/hadoop/ozone/om/ha/TestHadoopRpcOMFollowerReadFailoverProxyProvider.java hadoop-ozone/integration-test/src/test/java/org/apache/hadoop/ozone/shell/TestOzoneShellHA.java hadoop-ozone/interface-client/src/main/proto/OmClientProtocol.proto hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/ratis/OzoneManagerStateMachine.java hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/response/upgrade/OMCancelPrepareResponse.java hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/response/upgrade/OMCompleteFinalizeUpgradeResponse.java hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/om/response/upgrade/OMPrepareResponse.java hadoop-ozone/ozone-manager/src/main/java/org/apache/hadoop/ozone/protocolPB/OzoneManagerRequestHandler.java hadoop-ozone/ozone-manager/src/test/java/org/apache/hadoop/ozone/om/ratis/TestOzoneManagerStateMachine.java
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.

What changes were proposed in this pull request?
When an SCM follower restarts after being down for a long time, Ratis syncs the log to disk quickly, but
applyTransactionlags far behind, socommitIndexcan exceedappliedIndexby millions of entries. All the DB updates produced while applying that backlog accumulate in the single in-memoryBatchOperationheld bySCMHADBTransactionBufferImpl, causing heavy memory pressure.This change adds a count-based flush on the apply path:
ozone.scm.ha.dbtransactionbuffer.flush.pending.limit(default10000): the number of buffered DB operations after which the buffer is flushed. Values<= 0are invalid and fall back to the default with a warning.SCMHADBTransactionBuffer#flushIfPendingLimitReached()SCMStateMachine#applyTransactioncalls it right afterupdateLatestTrxInfo(...)andupdateLastAppliedTermIndex(...), still inside thebeginApplyingTransaction()/endApplyingTransaction()window. Flushing after the index is recorded keeps the persistedTransactionInfoconsistent with the committed batch. The existing interval-based flush is unchanged.What is the link to the Apache JIRA
https://issues.apache.org/jira/browse/HDDS-16666
How was this patch tested?
Added unit tests in
TestSCMHADBTransactionBufferImplandTestSCMStateMachine.Manually tested on a SCM HA docker cluster with the flush pending limit set to 5:
(Note: Changed log level to
INFOfor testing purpose)Follower caught up with leader.