Caideyipi opened a new pull request, #18014:
URL: https://github.com/apache/iotdb/pull/18014
## Description
### Problem
When multiple load-tsfile datatype conversion tasks run concurrently, each
conversion can open TsFile tablet parsers and reserve Pipe tablet memory. In
the investigated failure, the receiver failed in
`LoadTreeStatementDataTypeConvertExecutionVisitor` with
`forceAllocateForTablet: failed to allocate because there's too much memory for
tablets`, so the load stayed temporarily unavailable and the sender retried.
### Fix
Limit concurrent load-tsfile datatype conversions with a shared semaphore.
The permit count is the smaller of:
- `load_tsfile_tablet_conversion_thread_count`
- an estimate derived from the Pipe tablet memory reject threshold and the
tablet parser memory reserved per conversion
If Pipe memory management is disabled, the existing configured conversion
thread count is preserved.
### Validation
- `mvn spotless:apply -pl iotdb-core/datanode`
- `mvn spotless:check -pl iotdb-core/datanode`
- `mvn -Ddevelocity.off=true compile -pl iotdb-core/datanode` was attempted;
it reached javac but failed on existing generated-source/ANTLR/dependency
issues unrelated to this converter change, e.g. generated freemarker
fill/aggregation classes and parser symbols.
<hr>
This PR has:
- [x] been self-reviewed.
- [x] concurrent write
<hr>
##### Key changed/added classes (or packages if there are too many classes)
in this PR
-
`org.apache.iotdb.db.storageengine.load.converter.LoadTsFileDataTypeConverter`
--
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]
