harryruirui opened a new pull request, #65374:
URL: https://github.com/apache/doris/pull/65374
### What problem does this PR solve?
Issue Number: N/A
Related PR: N/A
Problem Summary:
This PR fixes a tablet scheduler blind spot for temporary partitions of
colocate tables.
For colocate tables, `TabletChecker` skips the whole table because normal
colocate tablets are expected to be checked by
`ColocateTableCheckerAndBalancer`. However, `ColocateTableCheckerAndBalancer`
iterates `OlapTable#getPartitions()`, which intentionally excludes temporary
partitions. As a result, tablets in temporary partitions of colocate tables are
not enqueued by either checker.
Bug type and impact:
- **Bug type**: scheduler / repair feeder missing coverage.
- **Affected objects**: tablets that belong to temporary partitions of
colocate OLAP tables.
- **Impact scope**: all scenarios that rely on tablet health checking or
repair scheduling for these temporary-partition tablets.
- **Operational impact**: if such tablets are located on decommissioning
BEs, decommission can remain blocked because no repair or relocation task is
created for them.
- **Reliability impact**: replica-missing, version-incomplete, and placement
issues on colocate temporary partitions can stay undetected until the temporary
partition is dropped or replaced.
The executor side is already able to resolve temporary partitions via
`OlapTable#getPartition(partitionId)`, so this PR keeps the fix in the feeder
layer.
Solution:
- Keep formal partitions of colocate tables handled by
`ColocateTableCheckerAndBalancer`.
- Let `TabletChecker` additionally check temporary partitions for colocate
tables.
- Include temporary partitions in `ADMIN REPAIR TABLE` when no partition
name is specified, so the default repair target set is consistent with all
partitions owned by the table.
### Release note
Fix colocate table temporary partition tablets not being checked or
scheduled for repair, which could block BE decommission.
### Check List (For Author)
- Test
- [x] Regression test
- [x] Unit Test
- [ ] Manual test (add detailed scripts or steps below)
- [ ] No need to test or manual test. Explain why:
- [ ] This is a refactor/code format and no logic has been changed.
- [ ] Previous test can cover this change.
- [ ] No code files have been changed.
- [ ] Other reason
Tested locally:
```bash
export JAVA_HOME=/opt/work/tools/jdk17 JDK_17=/opt/work/tools/jdk17
PATH=/opt/work/tools/bin:/opt/work/tools/jdk17/bin:/opt/homebrew/opt/bison/bin:/opt/homebrew/bin:$PATH
cd fe
/opt/work/tools/bin/mvn test -Dmaven.repo.local=/opt/work/tools/m2 -pl
fe-common,fe-core -am -Dcheckstyle.skip=true -DfailIfNoTests=false
-Dmaven.build.cache.enabled=false -Dtest=org.apache.doris.clone.TabletHealthTest
```
Result:
```text
Tests run: 5, Failures: 0, Errors: 0, Skipped: 0
BUILD SUCCESS
```
- Behavior changed:
- [ ] No.
- [x] Yes. Temporary partitions of colocate tables are now covered by
`TabletChecker`, and `ADMIN REPAIR TABLE` without explicit partition names
includes temporary partitions by default.
- Does this need documentation?
- [x] No.
- [ ] Yes.
### Check List (For Reviewer who merge this PR)
- [ ] Confirm the release note
- [ ] Confirm test cases
- [ ] Confirm document
- [ ] Add branch pick label
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
