[opt](memory) Move schemaHash from Replica to LocalReplica#62122
Open
dataroaring wants to merge 3 commits intoapache:masterfrom
Open
[opt](memory) Move schemaHash from Replica to LocalReplica#62122dataroaring wants to merge 3 commits intoapache:masterfrom
dataroaring wants to merge 3 commits intoapache:masterfrom
Conversation
Contributor
|
Thank you for your contribution to Apache Doris. Please clearly describe your PR:
|
Contributor
There was a problem hiding this comment.
Pull request overview
Moves schemaHash storage from the base Replica class into LocalReplica to reduce per-instance memory usage for CloudReplica (which does not need schemaHash).
Changes:
- Remove
schemaHashfield fromReplica; makegetSchemaHash()return-1andsetSchemaHash()a no-op in the base class. - Add
schemaHashfield toLocalReplicaand overridegetSchemaHash()/setSchemaHash()to store and return the real value. - Update
Replica.toString()to callgetSchemaHash()to preserve polymorphic behavior.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
fe/fe-core/src/main/java/org/apache/doris/catalog/Replica.java |
Removes schemaHash storage from the base type and stubs schema-hash accessors for non-local replicas. |
fe/fe-core/src/main/java/org/apache/doris/catalog/LocalReplica.java |
Adds schemaHash storage and overrides accessors so local replicas still expose the actual schema hash. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
140
to
151
| public long getVersion() { | ||
| return this.version; | ||
| } | ||
|
|
||
| public int getSchemaHash() { | ||
| return schemaHash; | ||
| return -1; | ||
| } | ||
|
|
||
| // for compatibility | ||
| public void setSchemaHash(int schemaHash) { | ||
| this.schemaHash = schemaHash; | ||
| // no-op in base class; overridden in LocalReplica | ||
| } |
There was a problem hiding this comment.
Replica schemaHash behavior has changed (base returns -1/no-op; LocalReplica stores the value). Please add/adjust unit tests (e.g., in ReplicaTest) to assert that LocalReplica.getSchemaHash() returns the constructor value and that a CloudReplica (or any non-LocalReplica Replica) reports -1, so regressions are caught during refactors.
Contributor
Author
There was a problem hiding this comment.
Addressed in commit 7f5e6d3. The existing test tried to instantiate abstract Replica directly — merged the base-class no-op assertion into the CloudReplica test, which inherits the same stubs. All three schemaHash polymorphism tests now compile and cover the requested scenarios.
— ThinkOps 🤖
Contributor
Author
|
run buildall |
TPC-H: Total hot run time: 28801 ms |
TPC-DS: Total hot run time: 178496 ms |
Contributor
FE UT Coverage ReportIncrement line coverage |
Contributor
|
/review |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
No blocking findings.
Critical checkpoints:
- Goal: The change is narrowly aimed at removing
schemaHashstorage from the baseReplicaobject while preserving real schema-hash state forLocalReplica. ReviewingReplica.java,LocalReplica.java,CloudReplica.java,MetadataViewer.java,LocalTabletInvertedIndex.java,EnvFactory.java,CloudEnvFactory.java,OlapTable.java,CloudInternalCatalog.java,CloudRestoreJob.java, andGsonUtils.java, the goal is met.LocalReplicastill stores and returns the real value; cloud replicas now consistently expose-1through the base implementation. - Minimality: The patch is small and focused to the two touched classes. No extra protocol, config, or storage-format changes were introduced.
- Concurrency: No new concurrent state, lock ordering, or thread-safety concerns were introduced.
- Lifecycle / static initialization: No special lifecycle or static-initialization risk is involved.
- Config: No configuration changes.
- Compatibility: Safe.
schemaHashwas not part ofReplica's Gson-serialized state before this patch, so moving the live field fromReplicatoLocalReplicadoes not change FE image / replay payloads. No FE-BE protocol compatibility issue was introduced. - Parallel paths: I checked both local and cloud creation paths. Local-mode paths still populate schema hash in
LocalReplica; cloud-mode paths do not depend on per-replica schema-hash state. - Conditional checks: Existing
replica.getSchemaHash() != -1guards inMetadataViewerremain correct. - Test coverage: No targeted test was added for the new polymorphic behavior, so there is still a small coverage gap around
LocalReplica.get/setSchemaHash()versus the cloud sentinel behavior. - Observability: Cloud proc / debug paths that print
replica.getSchemaHash()will now consistently show-1. That is acceptable because schema hash is intentionally not meaningful forCloudReplica, but it is a small user-visible diagnostic change. - Transaction / persistence / data-write correctness: Not involved.
- FE-BE variable passing: Not involved.
- Performance: The intended memory optimization is plausible and I did not find any obvious hot-path regression.
- Other issues: None found.
Overall opinion: this looks good to merge.
Contributor
Author
|
Added unit tests for the schemaHash polymorphism in commit 8711734:
This addresses the review feedback requesting regression tests for the schemaHash behavior change. — ThinkOps 🤖 |
Contributor
Author
|
run buildall |
2 similar comments
Contributor
Author
|
run buildall |
Contributor
Author
|
run buildall |
Contributor
Author
|
Fixed the compile error in Merged the base-class no-op The final test coverage for the Copilot-requested schemaHash polymorphism:
— ThinkOps 🤖 |
Contributor
Author
|
run buildall |
Contributor
FE Regression Coverage ReportIncrement line coverage |
Contributor
|
/review |
schemaHash in the base Replica class wastes 4 bytes per CloudReplica. Cloud storage paths don't use schema_hash, and callers like MetadataViewer already guard with getSchemaHash() != -1. This follows the same pattern as dbId removal in apache#62079. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add four tests to ReplicaTest covering the schemaHash behavior change: - testLocalReplicaGetSchemaHash: constructor value is returned - testLocalReplicaSetSchemaHash: setter updates the value - testBaseReplicaSchemaHashReturnsNegativeOne: base class returns -1, setter is no-op - testCloudReplicaSchemaHashReturnsNegativeOne: CloudReplica inherits -1 stub Generated by ThinkOps
…ract Replica The testBaseReplicaSchemaHashReturnsNegativeOne test tried to instantiate abstract class Replica directly, causing a compile error. Merged the no-op setSchemaHash assertion into the existing CloudReplica test, which exercises the same base-class code paths since CloudReplica inherits getSchemaHash()/setSchemaHash() without overriding them. Generated by ThinkOps
dataroaring
force-pushed
the
opt/replica-move-schemaHash
branch
from
7f5e6d3 to
6083321
Compare
Contributor
Author
|
run buildall |
Contributor
FE UT Coverage ReportIncrement line coverage |
Contributor
|
OpenCode automated review failed and did not complete. Error: Review step was failure (possibly timeout or cancelled) Please inspect the workflow logs and rerun the review after the underlying issue is resolved. |
Contributor
FE Regression Coverage ReportIncrement line coverage |
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.
Summary
schemaHashfield from baseReplicaclass toLocalReplica, saving 4 bytes perCloudReplicainstanceReplica.getSchemaHash()now returns-1(stub),setSchemaHash()is a no-opLocalReplicaoverrides both methods with actual storagedbIdremoval in [opt](memory) Remove unused dbId field from CloudReplica #62079Details
Cloud storage paths don't use
schema_hash, and callers likeMetadataVieweralready guard withgetSchemaHash() != -1, soCloudReplicareturning-1is safe and correct.Changes:
Replica.java: RemoveschemaHashfield;getSchemaHash()returns-1;setSchemaHash()is no-op;toString()usesgetSchemaHash()for polymorphismLocalReplica.java: AddschemaHashfield; overridegetSchemaHash()/setSchemaHash(); set in constructorCloudReplica.java: No changes needed (schemaHash param harmlessly ignored by super constructor)Test plan
CloudReplica.getSchemaHash()returns -1LocalReplica.getSchemaHash()returns actual valueMetadataViewerSCHEMA_ERROR check unaffected (guards with!= -1)🤖 Generated with Claude Code