* Emit aggregate segment processing metrics per sink instead of firehydrant
* add docs
* minor change
* checkstyle
* Fix DefaultQueryMetricsTest
* Minor changes in SinkMetricsEmittingQueryRunner
* spotbugs
* Address review comments
* Use ImmutableSet and ImmutableMap
* Create a helper class for saving state of StubServiceEmitter
* Add SinkQuerySegmentWalkerBenchmark
* Create SegmentMetrics class for tracking segment metrics
---------
Co-authored-by: Akshat Jain <akjn11@gmail.com>
* topn with granularity regression fixes
changes:
* fix issue where topN with query granularity other than ALL would use the heap algorithm when it was actual able to use the pooled algorithm, and incorrectly used the pool algorithm in cases where it must use the heap algorithm, a regression from #16533
* fix issue where topN with query granularity other than ALL could incorrectly process values in the wrong time bucket, another regression from #16533
* move defensive check outside of loop
* more test
* extra layer of safety
* move check outside of loop
* fix spelling
* add query context parameter to allow using pooled algorithm for topN when multi-passes is required even wihen query granularity is not all
* add comment, revert IT context changes and add new context flag
Changes
--------
- Simplify the arguments of IndexerMetadataStorageCoordinator.allocatePendingSegment
- Remove field SegmentCreateRequest.upgradedFromSegmentId as it was always null
- Miscellaneous cleanup
* snapshot column capabilities for realtime cursors
changes:
* adds `CursorBuildSpec.getPhysicalColumns()` to allow specifying the set of required physical columns from a segment. if null, all columns are assumed to be required (e.g. full scan)
* `IncrementalIndexCursorFactory`/`IncrementalIndexCursorHolder` uses the physical columns from the cursor build spec to know which set of dimensions to 'snapshot' the capabilities for, allowing expression selectors on realtime queries to no longer be required to treat selectors from `StringDimensionIndexer` as multi-valued unless they truly are multi-valued. this fixes several bugs with expressions on realtime queries that change a value from `StringDimensionIndexer` to some type other than string, which would often result in a single element array from the column being handled as multi-valued
* `StringDimensionIndexer.setSparseIndexed()` now adds the default value to the dictionary when set
* `StringDimensionIndexer` column value selectors now always report that they are dictionary encoded, and that name lookup is possible in advance on their selectors (since set sparse adds the null value so the cardinality is correct)
* fixed a mistake that expression selectors for realtime queries with no null values could not use dictionary encoded selectors
* hmm
* test changes
* cleanup
* add test coverage
* fix test
* fixes
* cleanup
This PR contains non-functional / refactoring changes of the following classes in the sql module:
1. Move ExplainPlan and ExplainAttributes fromsql/src/main/java/org/apache/druid/sql/http to processing/src/main/java/org/apache/druid/query/explain
2. Move sql/src/main/java/org/apache/druid/sql/SqlTaskStatus.java -> processing/src/main/java/org/apache/druid/query/http/SqlTaskStatus.java
3. Add a new class processing/src/main/java/org/apache/druid/query/http/ClientSqlQuery.java that is effectively a thin POJO version of SqlQuery in the sql module but without any of the Calcite functionality and business logic.
4. Move BrokerClient, BrokerClientImpl and Broker classes from sql/src/main/java/org/apache/druid/sql/client to server/src/main/java/org/apache/druid/client/broker.
5. Remove BrokerServiceModule that provided the BrokerClient. The functionality is now contained in ServiceClientModule in the server package itself which provides all the clients as well.
This is done so that we can reuse the said classes in #17353 without brining in Calcite and other dependencies to the Overlord.
* plan consistently with either UnionDataSource or UnionQuery for decoupled mode
* expose errors
* move decoupled related setting from PlannerConfig to QueryContexts
This change is to emit following metrics as part of GroupByStatsMonitor monitor,
mergeBuffer/used -> Number of merge buffers used.
mergeBuffer/acquisitionTimeNs -> Total time required to acquire merge buffer.
mergeBuffer/acquisition -> Number of queries that acquired a batch of merge buffers.
groupBy/spilledQueries -> Number of queries that spilled onto the disk.
groupBy/spilledBytes-> Spilled bytes on the disk.
groupBy/mergeDictionarySize -> Size of the merging dictionary.
All JDK 8 based CI checks have been removed.
Images used in Dockerfile(s) have been updated to Java 17 based images.
Documentation has been updated accordingly.
changes:
* fix issue when merging projections from multiple-incremental persists which was hoping that some 'dim conversion' buffers were not closed, but they already were (by the merging iterator). fix involves selectively persisting these conversion buffers to temp files in the segment write out directory and mapping them and tying them to the segment level closer so that they are available after the lifetime of the parent merger
* modify auto column serializers to use segment write out directory for temp files instead of java.io.tmpdir
* fix queryable index projection to not put the time-like column as a dimension, instead only adding it as __time
* use smoosh for temp files so can safely write any Serializer to a temp smoosh
* ScanQuery: equals/hashCode/toString
* DruidQuery: changes of Align ScanQuery column order with its desired signature #17457
* ScanQueryTest: add equalsverifer test
This patch re-uses timeBoundaryInspector for each cursor holder, which
enables caching of minDataTimestamp and maxDataTimestamp.
Fixes a performance regression introduced in #16533, where these fields
stopped being cached across cursors. Prior to that patch, they were
cached in the QueryableIndexStorageAdapter.
* introduces `UnionQuery`
* some changes to enable a `UnionQuery` to have multiple input datasources
* `UnionQuery` execution is driven by the `QueryLogic` - which could later enable to reduce some complexity in `ClientQuerySegmentWalker`
* to run the subqueries of `UnionQuery` there was a need to access the `conglomerate` from the `Runner`; to enable that some refactors were done
* renamed `UnionQueryRunner` to `UnionDataSourceQueryRunner`
* `QueryRunnerFactoryConglomerate` have taken the place of `QueryToolChestWarehouse` which shaves of some unnecessary things here and there
* small cleanup/refactors
Currently, durable storage and export both require configuring a temporary directory to be used using druid.export.storage.<connectorType>.tempLocalDir and druid.msq.intermediate.storage.tempDir.
Tasks on middle manager already have a configured temporary directory. This PR aims to reduce the configuration required by using the task directory as a default if it is not explicitly configured, thus reducing the number of configs that a user has to set.
Please note that preference would be given to the user configured, druid.*.storage.temp*Dir, on the tasks. If that is not configured, we then use the configured temporary directory.
Overlord and brokers also require storage connector configurations (for the durableStorageCleanerOverlordDuty and to fetch results of async queries respectively), but do not have a default temporary task directory. The configuration is still required for these services.
* Update errorprone, mockito, jacoco, checkerframework.
This patch updates various build and test dependencies, to see if they
cause unit tests on JDK 21 to behave more reliably.
* Update licenses, tests.
* Remove assertEquals.
* Repair two tests.
* Update some more tests.
Calling toString on newConfig is unnecessary, because it will be done
automatically by the logger. This saves some effort under log levels
higher than DEBUG.
changes:
* adds `SqlBenchmarkDatasets` which contains commonly used benchmark data generator schemas
* adds `SqlBaseBenchmark` which contains common benchmark segment generation methods for any benchmark using `SqlBenchmarkDatasets`
* adds `SqlBaseQueryBenchmark` and `SqlBasePlanBenchmark` for benchmarks measuring queries and planning respectively
* migrate all existing SQL jmh benchmarks to extend `SqlBaseQueryBenchmark`, quite dramatically reducing the boilerplate needed to create benchmarks, and allowing the use of multiple datasources within a benchmark file
* adjustments to data generator stuff to allow passing in an ObjectMapper so that the same mapper can be used for both benchmark queries and segment generation, avoiding the need to register stuff with both mappers for benchmarks
* adds `SqlProjectionsBenchmark` and `SqlComplexMetricsColumnsBenchmark` for measuring projections and measuring complex metric compression respectively
* GlueingPartitioningOperator: It continuously receives data, and outputs batches of partitioned RACs. It maintains a last-partitioning-boundary of the last-pushed-RAC, and attempts to glue it with the next RAC it receives, ensuring that partitions are handled correctly, even across multiple RACs. You can check GlueingPartitioningOperatorTest for some good examples of the "glueing" work.
* PartitionSortOperator: It sorts rows inside partitioned RACs, on the sort columns. The input RACs it receives are expected to be "complete / separate" partitions of data.
Follow up to #17214, adds implementations for substituteCombiningFactory so that more
datasketches aggs can match projections, along with some projections tests for datasketches.
* Logger: Log context of DruidExceptions.
There is often interesting and unique information available in the
"context" of a DruidException. This information is additive to both
the message and the cause, and was missed when we log. This patch adds
the DruidException context to log messages whenever stack traces are
enabled.
* Only log nonempty contexts.
The javadoc for SegmentDescriptor discusses differences between it and
SegmentId, but misses the most important difference: SegmentDescriptor
can have a narrower interval than the segment being referenced.
Refactors a few things.
- Adds SemanticUtils maps to columns.
- Add some addAll functions to reduce duplication, and for future reuse.
- Refactor VariantColumnAndIndexSupplier to only take a SmooshedFileMapper instead.
- Refactor LongColumnSerializerV2 to have separate functions for serializing a value and null.
changes:
adds ExpressionProcessing.allowVectorizeFallback() and ExpressionProcessingConfig.allowVectorizeFallback(), defaulting to false until few remaining bugs can be fixed (mostly complex types and some odd interactions with mixed types)
add cannotVectorizeUnlessFallback functions to make it easy to toggle the default of this config, and easy to know what to delete when we remove it in the future
Akshat Jain
Rishabh Singh
Clint Wylie
Karan Kumar
Adarsh Sanjeev
Kashif Faraz
Zoltan Haindrich
Abhishek Radhakrishnan
Virushade
Vivek Dhiman
Gian Merlino
Suraj Goel
Laksh Singla
Hardik Bajaj
Shivam Garg