Clarify expected behavior for incr_cnt recovery and snap_index_type change after single-node restart
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 2.9k
- Forks
- 339
- Avg merge
- 1d 19h
- Merged PRs (30d)
- 52
Description
Environment
- Version: OceanBase seekdb
- Source Revision: 1-f84169fbd80d020887d038cdb0892703e28833f3
- Build Branch: HEAD
- Build Time: Aug 10 2026 10:45:34
- Build Flags: RelWithDebInfo
- Is this an RD-provided temporary version?: No
Description
This issue seeks clarification on the expected behavior of vector index state recovery after a single-node (endpoint-only) restart. Specifically, the behavior of incr_cnt and the change in snap_index_type from 1 to 9 needs to be confirmed as per design.
Steps to Reproduce / Context
The test scenario is vector_reboot_lite. The environment was running a workload, and the observer was restarted. The following points require clarification from the development team:
- Does the observer use lazy load after a restart?
- Should
incr_cntonly be restored after the first ANN query is executed? - What is the meaning of the
snap_index_typechange from1to9? - Is it expected that
all_index_mem_used=0before the first query? - Does the endpoint-only (single-node) mode support automatic pre-loading?
Work Reference: xJMcKDVRoXpa89uA
Log Information
Database Connection
obclient -h% -P 2836 -uroot@sys -A -Dtest
Current Preserved Environment
host: %
port: 2836
observer pid: 3927865
Log Directory
- Workload Log:
/data/1/seekdb-gitlab/vector_reboot_lite/runs/pipeline-189961-job-919135/scenario/singlejob/vector_reboot_lite/vector_reboot_lite.log - EasyTest Log:
/data/1/seekdb-gitlab/vector_reboot_lite/runs/pipeline-189961-job-919135/scenario/singlejob/vector_reboot_lite/logs/EasyTest_20260810145128.log - Report:
/data/1/seekdb-gitlab/vector_reboot_lite/runs/pipeline-189961-job-919135/scenario/singlejob/vector_reboot_lite/reports/report.html - Database Logs:
/data/1/seekdb-gitlab/vector_reboot_lite/log/pipeline-189961-job-919135/ - Key File:
seekdb.log.20260810153409167
Other Notes
- Reproducibility: Not specified.
- Test Changes: Not specified.
- Other Environmental Anomalies/Changes: Not specified.
- Is the scene preserved?: Yes (see connection details above).
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the vector_reboot_lite scenario and inspect the preserved observer and database logs, including seekdb.log.20260810153409167 and the reported EasyTest and HTML report files. Confirm the designed recovery behavior for lazy loading, incr_cnt, snap_index_type, all_index_mem_used, and endpoint-only pre-loading; done means the development team has resolved these questions and any required tests or documentation are identified.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100