This PR adds checks for verification of DataSourceCompactionConfig and CompactionTask with msq engine to ensure:
each aggregator in metricsSpec is idempotent
metricsSpec is non-null when rollup is set to true
Unit tests and existing compaction ITs have been updated accordingly.
This PR fixes query correctness issues for MSQ window functions when using more than 1 worker (that is, maxNumTasks > 2).
Currently, we were keeping the shuffle spec of the previous stage when we didn't have any partition columns for window stage. This PR changes it to override the shuffle spec of the previous stage to MixShuffleSpec (if we have a window function with empty over clause) so that the window stage gets a single partition to work on.
A test has been added for a query which returned incorrect results prior to this change when using more than 1 workers.
* enables to launch a fake broker based on test resources (druidtest uri)
* could record queries into new testfiles during usage
* instead of re-purpose Calcite's Hook migrates to use DruidHook which we can add further keys
* added a quidem-ut module which could be the place for tests which could iteract with modules/etc
This patch introduces an optional cluster configuration, druid.indexing.formats.stringMultiValueHandlingMode, allowing operators to override the default mode SORTED_SET for string dimensions. The possible values for the config are SORTED_SET, SORTED_ARRAY, or ARRAY (SORTED_SET is the default). Case insensitive values are allowed.
While this cluster property allows users to manage the multi-value handling mode for string dimension types, it's recommended to migrate to using real array types instead of MVDs.
This fixes a long-standing issue where compaction will honor the configured cluster wide property instead of rewriting it as the default SORTED_ARRAY always, even if the data was originally ingested with ARRAY or SORTED_SET.
* SQL syntax error should target USER persona
* * revert change to queryHandler and related tests, based on review comments
* * add test
* Introduce KinesisRecordEntity to support Kinesis headers in InputFormats
* * add kinesisInputFormat and Reader, and tests
* * bind KinesisInputFormat class to module
* * improve test coverage
* * remove references to kafka
* * resolve review comments
* * remove comment
* * fix grammer of comment
* * fix comment again
* * fix comment again
* * more review comments
* * add partitionKey
* * add check for same timestamp and partitionKey column name
* * fix intellij inspection
If the optional query parameter detail is supplied, then the response also includes the following:
* A stages object that summarizes information about the different stages being used for query execution, such as stage number, phase, start time, duration, input and output information, processing methods, and partitioning.
* A counters object that provides details on the rows, bytes, and files processed at various stages for each worker across different channels, along with sort progress.
* A warnings object that provides details about any warnings.
* MSQ worker: Support in-memory shuffles.
This patch is a follow-up to #16168, adding worker-side support for
in-memory shuffles. Changes include:
1) Worker-side code now respects the same context parameter "maxConcurrentStages"
that was added to the controller in #16168. The parameter remains undocumented
for now, to give us a chance to more fully develop and test this functionality.
1) WorkerImpl is broken up into WorkerImpl, RunWorkOrder, and RunWorkOrderListener
to improve readability.
2) WorkerImpl has a new StageOutputHolder + StageOutputReader concept, which
abstract over memory-based or file-based stage results.
3) RunWorkOrder is updated to create in-memory stage output channels when
instructed to.
4) ControllerResource is updated to add /doneReadingInput/, so the controller
can tell when workers that sort, but do not gather statistics, are done reading
their inputs.
5) WorkerMemoryParameters is updated to consider maxConcurrentStages.
Additionally, WorkerChatHandler is split into WorkerResource, so as to match
ControllerChatHandler and ControllerResource.
* Updates for static checks, test coverage.
* Fixes.
* Remove exception.
* Changes from review.
* Address static check.
* Changes from review.
* Improvements to docs and method names.
* Update comments, add test.
* Additional javadocs.
* Fix throws.
* Fix worker stopping in tests.
* Fix stuck test.
Follow-up to #16291, this commit enables a subset of existing native compaction ITs on the MSQ engine.
In the process, the following changes have been introduced in the MSQ compaction flow:
- Populate `metricsSpec` in `CompactionState` from `querySpec` in `MSQControllerTask` instead of `dataSchema`
- Add check for pre-rolled-up segments having `AggregatorFactory` with different input and output column names
- Fix passing missing cluster-by clause in scan queries
- Add annotation of `CompactionState` to tombstone segments
Changes:
- Add API `/druid/coordinator/v1/config/compaction/global` to update cluster level compaction config
- Add class `CompactionConfigUpdateRequest`
- Fix bug in `CoordinatorCompactionConfig` which caused compaction engine to not be persisted.
Use json field name `engine` instead of `compactionEngine` because JSON field names must align
with the getter name.
- Update MSQ validation error messages
- Complete overhaul of `CoordinatorCompactionConfigResourceTest` to remove unnecessary mocking
and add more meaningful tests.
- Add `TuningConfigBuilder` to easily build tuning configs for tests.
- Add `DatasourceCompactionConfigBuilder`
Changes the WindowFrame internals / representation a bit; introduces dedicated frametypes for rows and groups which corresponds to the implemented processing methods
* MSQ window functions: Revamp logic to create separate window stages when empty over() clause is present
* Fix tests
* Revert changes of creating separate stages for empty over clause
* Address review comments
changes:
* removes `druid.indexer.task.batchProcessingMode` in favor of always using `CLOSED_SEGMENT_SINKS` which uses `BatchAppenderator`. This was intended to become the default for native batch, but that was missed so `CLOSED_SEGMENTS` was the default (using `AppenderatorImpl`), however MSQ has been exclusively using `BatchAppenderator` with no problems so it seems safe to just roll it out as the only option for batch ingestion everywhere.
* with `batchProcessingMode` gone, there is no use for `AppenderatorImpl` so it has been removed
* implify `Appenderator` construction since there are only separate stream and batch versions now
* simplify tests since `batchProcessingMode` is gone
This PR aims to check if the complex column being queried aligns with the supported types in the aggregator and aggregator factories, and throws a user-friendly error message if they don't.
* Throw exception if DISTINCT used with window functions aggregate call
* Improve error message when unsupported aggregations are used with window functions
Fixes#16766
Change log level from INFO to DEBUG when processing an empty user map
during polling. An empty user map is a normal situation for some
authenticators (e.g. LDAP) and polling is frequent (1 minute by
default.)
changes:
* removed `Firehose` and `FirehoseFactory` and remaining implementations which were mostly no longer used after #16602
* Moved `IngestSegmentFirehose` which was still used internally by Hadoop ingestion to `DatasourceRecordReader.SegmentReader`
* Rename `SQLFirehoseFactoryDatabaseConnector` to `SQLInputSourceDatabaseConnector` and similar renames for sub-classes
* Moved anything remaining in a 'firehose' package somewhere else
* Clean up docs on firehose stuff
* #16717 defer provider instatiation
* add license header
* fix style, ignore new class in jacoco as it is still initialization code
---------
Co-authored-by: Alberto Lago Alvarado <albl@sitecore.net>
* When an ArrayList RAC creates a child RAC, the start and end offsets need to have the offset of parent's start offset
* Defaults the 2nd window bound to CURRENT ROW when only a single bound is specified
* Removes the windowingStrictValidation warning and throws a hard exception when Order By alongside RANGE clause is not provided with UNBOUNDED or CURRENT ROW as both bounds
Changes:
- No functional change
- Add class `TuningConfigBuilder` to build `IndexTuningConfig`, `CompactionTuningConfig`
- Remove old class `ParallelIndexTestingFactory.TuningConfigBuilder`
- Remove some unused fields and methods
Changes
- No functional change
- Remove unused method `IndexTuningConfig.withPartitionsSpec()`
- Remove unused method `ParallelIndexTuningConfig.withPartitionsSpec()`
- Remove redundant method `CompactTask.emitIngestionModeMetrics()`
- Remove Clock argument from `CompactionTask.createDataSchemasForInterval()` as it was only needed
for one test which was just verifying the value passed by the test itself. The code now uses a `Stopwatch`
instead and test simply verifies that the metric has been emitted.
- Other minor cleanup changes
Better fallback strategy when the broker is unable to materialize the subquery's results as frames for estimating the bytes:
a. We don't touch the subquery sequence till we know that we can materialize the result as frames
Description:
Compaction operations issued by the Coordinator currently run using the native query engine.
As majority of the advancements that we are making in batch ingestion are in MSQ, it is imperative
that we support compaction on MSQ to make Compaction more robust and possibly faster.
For instance, we have seen OOM errors in native compaction that MSQ could have handled by its
auto-calculation of tuning parameters.
This commit enables compaction on MSQ to remove the dependency on native engine.
Main changes:
* `DataSourceCompactionConfig` now has an additional field `engine` that can be one of
`[native, msq]` with `native` being the default.
* if engine is MSQ, `CompactSegments` duty assigns all available compaction task slots to the
launched `CompactionTask` to ensure full capacity is available to MSQ. This is to avoid stalling which
could happen in case a fraction of the tasks were allotted and they eventually fell short of the number
of tasks required by the MSQ engine to run the compaction.
* `ClientCompactionTaskQuery` has a new field `compactionRunner` with just one `engine` field.
* `CompactionTask` now has `CompactionRunner` interface instance with its implementations
`NativeCompactinRunner` and `MSQCompactionRunner` in the `druid-multi-stage-query` extension.
The objectmapper deserializes `ClientCompactionRunnerInfo` in `ClientCompactionTaskQuery` to the
`CompactionRunner` instance that is mapped to the specified type [`native`, `msq`].
* `CompactTask` uses the `CompactionRunner` instance it receives to create the indexing tasks.
* `CompactionTask` to `MSQControllerTask` conversion logic checks whether metrics are present in
the segment schema. If present, the task is created with a native group-by query; if not, the task is
issued with a scan query. The `storeCompactionState` flag is set in the context.
* Each created `MSQControllerTask` is launched in-place and its `TaskStatus` tracked to determine the
final status of the `CompactionTask`. The id of each of these tasks is the same as that of `CompactionTask`
since otherwise, the workers will be unable to determine the controller task's location for communication
(as they haven't been launched via the overlord).
In case of few aggregators for example BloomSqlAggregator, BaseVarianceSqlAggregator etc, the aggName is being updated from a0 to a0:agg, breaching the contract as we would expect the aggName as the name which is passed. This is causing a mismatch while creating a column accessor.
This commit aims to correct those violating sql aggregators.
Add a shuffling based on the resultShuffleSpecFactory after a limit processor depending on the query destination. LimitFrameProcessors currently do not update the partition boosting column, so we also add the boost column to the previous stage, if one is required.
* Add MSQ query context maxNumSegments.
- Default is MAX_INT (unbounded).
- When set and if a time chunk contains more number of segments than set in the
query context, the MSQ task will fail with TooManySegments fault.
* Fixup hashCode().
* Rename and checkpoint.
* Add some insert and replace happy and sad path tests.
* Update error msg.
* Commentary
* Adjust the default to be null (meaning no max bound on number of segments).
Also fix formatter.
* Fix CodeQL warnings and minor cleanup.
* Assert on maxNumSegments tuning config.
* Minor test cleanup.
* Use null default for the MultiStageQueryContext as well
* Review feedback
* Review feedback
* Move logic to common function getPartitionsByBucket shared by INSERT and REPLACE.
* Rename to validateNumSegmentsPerBucketOrThrow() for consistency.
* Add segmentGranularity to error message.
MSQ cannot process null bytes in string fields, and the current workaround is to remove them using the REPLACE function. 'removeNullBytes' context parameter has been added which sanitizes the input string fields by removing these null bytes.
* first pass
* more changes
* fix tests and formatting
* fix kinesis failing tests
* fix kafka tests
* add dimension name to float parse errors
* double and convertToType handling of dimensionName can report parse errors with dimension name
* fix checkstyle issue
* fix tests
* more cases to have better parse exception messages
* fix test
* fix tests
* partially address comments
* annotate method parameter with nullable
* address comments
* fix tests
* let float, double, long dimensionIndexer pass dimensionName down to dimensionHandlerUtils
* fix compilation error and clean up formatting
* clean up whitespace
* address feedback. undo change, pass down report parse exception for convertToType
* fix test
* Support ListBasedInputRow in Kafka ingestion with header
* Fix up buildBlendedEventMap
* Add new test for KafkaInputFormat with csv value and headers
* Do not use forbidden APIs
* Move utility method to TestUtils
index_realtime tasks were removed from the documentation in #13107. Even
at that time, they weren't really documented per se— just mentioned. They
existed solely to support Tranquility, which is an obsolete ingestion
method that predates migration of Druid to ASF and is no longer being
maintained. Tranquility docs were also de-linked from the sidebars and
the other doc pages in #11134. Only a stub remains, so people with
links to the page can see that it's no longer recommended.
index_realtime_appenderator tasks existed in the code base, but were
never documented, nor as far as I am aware were they used for any purpose.
This patch removes both task types completely, as well as removes all
supporting code that was otherwise unused. It also updates the stub
doc for Tranquility to be firmer that it is not compatible. (Previously,
the stub doc said it wasn't recommended, and pointed out that it is
built against an ancient 0.9.2 version of Druid.)
ITUnionQueryTest has been migrated to the new integration tests framework and updated to use Kafka ingestion.
Co-authored-by: Gian Merlino <gianmerlino@gmail.com>
This PR fixes a few bugs with MSQ export. The main change is calling SqlResults#coerce before writing the column. This allows sketches and json to be correctly deserialized. The format of the exported complex columns are similar to those produced by Async MSQ queries with CSV format.
Notes:
Fix printing of complex columns during export. Sketches and JSON are now correctly formatted during export.
Fix an NPE if the writer has not been initialized. Empty export queries will create an empty file at the location.
Fix a bug with counters for MSQ export, where rows were reported for only the first partition.
As part of #16481, we have started uploading the chunks in parallel.
That means that it's not necessary for the part that finished uploading last
to be less than or equal to the chunkSize (as the final part could've been uploaded earlier).
This made a test in RetryableS3OutputStreamTest flaky where we were
asserting that the final part should be smaller than chunk size.
This commit fixes the test, and also adds another test where the file size
is such that all chunk sizes would be of equal size.
* contains Make a full copy of the parser and apply our modifications to it #16503
* some minor api changes pair/entry
* some unnecessary aggregation was removed from a set of queries in `CalciteSubqueryTest`
* `AliasedOperatorConversion` was detecting `CHAR_LENGTH` as not a function ; I've removed the check
* the field it was using doesn't look maintained that much
* the `kind` is passed for the created `SqlFunction` so I don't think this check is actually needed
* some decoupled test cases become broken - will be fixed later
* some aggregate related changes: due to the fact that SUM() and COUNT() of no inputs are different
* upgrade avatica to 1.25.0
* `CalciteQueryTest#testExactCountDistinctWithFilter` is now executable
Closeapache/druid#16503
* Remove unused constants
* Refactor getBlockBlobLength
* Better link
* Upper-case log
* Mark defaultStorageAccount nullable
This is the case if you do not use Azure for deep-storage but ingest from Azure blobs.
* Do not always create a new container if it doesn't exist
Specifically, only create a container if uploading a blob or writing a blob stream
* Add lots of comments, group methods
* Revert "Mark defaultStorageAccount nullable"
* Add mockito for junit
* Add extra test
* Add comment
Thanks George.
* Pass blockSize as Long
* Test more branches...
* Optimise S3 storage writing for MSQ durable storage
* Get rid of static ConcurrentHashMap
* Fix static checks
* Fix tests
* Remove unused constructor parameter chunkValidation + relevant cleanup
* Assert etags as String instead of Integer
* Fix flaky test
* Inject executor service
* Make threadpool size dynamic based on number of cores
* Fix S3StorageDruidModuleTest
* Fix S3StorageConnectorProviderTest
* Fix injection issues
* Add S3UploadConfig to manage maximum number of concurrent chunks dynamically based on chunk size
* Address the minor review comments
* Refactor S3UploadConfig + ExecutorService into S3UploadManager
* Address review comments
* Make updateChunkSizeIfGreater() synchronized instead of recomputeMaxConcurrentNumChunks()
* Address the minor review comments
* Fix intellij-inspections check
* Refactor code to use futures for maxNumConcurrentChunks. Also use executor service with blocking queue for backpressure semantics.
* Update javadoc
* Get rid of cyclic dependency injection between S3UploadManager and S3OutputConfig
* Fix RetryableS3OutputStreamTest
* Remove unnecessary synchronization parts from RetryableS3OutputStream
* Update javadoc
* Add S3UploadManagerTest
* Revert back to S3StorageConnectorProvider extends S3OutputConfig
* Address Karan's review comments
* Address Kashif's review comments
* Change a log message to debug
* Address review comments
* Fix intellij-inspections check
* Fix checkstyle
---------
Co-authored-by: asdf2014 <asdf2014@apache.org>
* Fallback vectorization for FunctionExpr and BaseMacroFunctionExpr.
This patch adds FallbackVectorProcessor, a processor that adapts non-vectorizable
operations into vectorizable ones. It is used in FunctionExpr and BaseMacroFunctionExpr.
In addition:
- Identifiers are updated to offer getObjectVector for ARRAY and COMPLEX in addition
to STRING. ExprEvalObjectVector is updated to offer ARRAY and COMPLEX as well.
- In SQL tests, cannotVectorize now fails tests if an exception is not thrown. This makes
it easier to identify tests that can now vectorize.
- Fix a null-matcher bug in StringObjectVectorValueMatcher.
* Fix tests.
* Fixes.
* Fix tests.
* Fix test.
* Fix test.
* Fix serde for ArrayOfDoublesSketchConstantPostAggregator.
The version originally added in #13819 was missing an annotation for
the "value" property. Fixes#16539.
Line endings for ArrayOfDoublesSketchConstantPostAggregator.java are changed
from \r\n to \n.
Adds a serde test, and improves various other datasketches post-aggregator
serde tests to deserialize into PostAggregator. This verifies that the type
information is set up correctly.
* Fix excessive imports.
* Fix equals, hashCode.
* Fix task bootstrap locations.
* Remove dependency of SegmentCacheManager from SegmentLoadDropHandler.
- The load drop handler code talks to the local cache manager via
SegmentManager.
* Clean up unused imports and stuff.
* Test fixes.
* Intellij inspections and test bind.
* Clean up dependencies some more
* Extract test load spec and factory to its own class.
* Cleanup test util
* Pull SegmentForTesting out to TestSegmentUtils.
* Fix up.
* Minor changes to infoDir
* Replace server announcer mock and verify that.
* Add tests.
* Update javadocs.
* Address review comments.
* Separate methods for download and bootstrap load
* Clean up return types and exception handling.
* No callback for loadSegment().
* Minor cleanup
* Pull out the test helpers into its own static class so it can have better state control.
* LocalCacheManager stuff
* Fix build.
* Fix build.
* Address some CI warnings.
* Minor updates to javadocs and test code.
* Address some CodeQL test warnings and checkstyle fix.
* Pass a Consumer<DataSegment> instead of boolean & rename variables.
* Small updates
* Remove one test constructor.
* Remove the other constructor that wasn't initializing fully and update usages.
* Cleanup withInfoDir() builder and unnecessary test hooks.
* Remove mocks and elaborate on comments.
* Commentary
* Fix a few Intellij inspection warnings.
* Suppress corePoolSize intellij-inspect warning.
The intellij-inspect tool doesn't seem to correctly inspect
lambda usages. See ScheduledExecutors.
* Update docs and add more tests.
* Use hamcrest for asserting order on expectation.
* Shutdown bootstrap exec.
* Fix checkstyle
Changes:
- Loosely followed the steps in the migration guide at
https://junit.org/junit5/docs/current/user-guide/#migrating-from-junit4
- Updated POM to add JUnit 5 dependencies
- Updated imports to JUnit 5 packages
- Updated annotations (Lifecycle annotations like `@BeforeEach`)
- Updated exception testing (`assertThrows`)
- Updated temporary path handling (use `@TempDir` annotation)
- Various other updates (replace other `Rule` usages, make sure to use JUnit 5 assertions)
This commit aims to enable the re-ordering of window operators in order to optimise
the sort and partition operators.
Example :
```
SELECT m1, m2,
SUM(m1) OVER(PARTITION BY m2) as sum1,
SUM(m2) OVER() as sum2
from numFoo
GROUP BY m1,m2
```
In order to compute this query, we can order the operators as to first compute the operators
corresponding to sum2 and then place the operators corresponding to sum1 which would
help us in reducing one sort operator if we order our operators by sum1 and then sum2.
There are a few issues with using Jackson serialization in sending datasketches between controller and worker in MSQ. This caused a blowup due to holding multiple copies of the sketch being stored.
This PR aims to resolve this by switching to deserializing the sketch payload without Jackson.
The PR adds a new query parameter used during communication between controller and worker while fetching sketches, "sketchEncoding".
If the value of this parameter is OCTET, the sketch is returned as a binary encoding, done by ClusterByStatisticsSnapshotSerde.
If the value is not the above, the sketch is encoded by Jackson as before.
This PR updates CompactionTask to not load any lookups by default, unless transformSpec is present.
If transformSpec is present, we will make the decision based on context values, loading all lookups by default. This is done to ensure backward compatibility since transformSpec can reference lookups.
If transform spec is not present and no context value is passed, we donot load any lookup.
This behavior can be overridden by supplying lookupLoadingMode and lookupsToLoad in the task context.
MSQ sorts the columns in a highly specialized manner by byte comparisons. As such the values are serialized differently. This works well for the primitive types and primitive arrays, however complex types cannot be serialized specially.
This PR adds the support for sorting the complex columns by deserializing the value from the field and comparing it via the type strategy. This is a lot slower than the byte comparisons, however, it's the only way to support sorting on complex columns that can have arbitrary serialization not optimized for MSQ.
The primitives and the arrays are still compared via the byte comparison, therefore this doesn't affect the performance of the queries supported before the patch. If there's a sorting key with mixed complex and primitive/primitive array types, for example: longCol1 ASC, longCol2 ASC, complexCol1 DESC, complexCol2 DESC, stringCol1 DESC, longCol3 DESC, longCol4 ASC, the comparison will happen like:
longCol1, longCol2 (ASC) - Compared together via byte-comparison, since both are byte comparable and need to be sorted in ascending order
complexCol1 (DESC) - Compared via deserialization, cannot be clubbed with any other field
complexCol2 (DESC) - Compared via deserialization, cannot be clubbed with any other field, even though the prior field was a complex column with the same order
stringCol1, longCol3 (DESC) - Compared together via byte-comparison, since both are byte comparable and need to be sorted in descending order
longCol4 (ASC) - Compared via byte-comparison, couldn't be coalesced with the previous fields as the direction was different
This way, we only deserialize the field wherever required
Changes:
- Remove deprecated `markAsUnused` parameter from `KillUnusedSegmentsTask`
- Allow `kill` task to use `REPLACE` lock when `useConcurrentLocks` is true
- Use `EXCLUSIVE` lock by default
* enable quidem uri support for `druidtest:///?ComponentSupplier=Nested` and similar
* changes the way `SqlTestFrameworkConfig` is being applied; all options will have their own annotation (its kinda impossible to detect that an annotation has a set value or its the default)
* enables hierarchical processing of config annotation (was needed to enable class level supplier annotation)
* moves uri processing related string2config stuff into `SqlTestFrameworkConfig`
With this PR changes, MSQ tasks (MSQControllerTask and MSQWorkerTask) only load the required lookups during querying and ingestion, based on the value of CTX_LOOKUPS_TO_LOAD key in the query context.
Add validation for reindex with realtime sources.
With the addition of concurrent compaction, it is possible to ingest data while querying from realtime sources with MSQ into the same datasource. This could potentially lead to issues if the interval that is ingested into is replaced by an MSQ job, which has queried only some of the data from the realtime task.
This PR adds validation to check that the datasource being ingested into is not being queried from, if the query includes realtime sources.
* * add another catalog clustering columns unit test
* * dissallow clusterKeys with descending order
* * make more clear that clustering is re-written into ingest node
whether a catalog table or not
* * when partitionedBy is stored in catalog, user shouldnt need to specify
it in order to specify clustering
* * fix intellij inspection failure
update dependencies to address new batch of CVEs:
- Azure POM from 1.2.19 to 1.2.23 to update transitive dependency nimbus-jose-jwt to address: CVE-2023-52428
- commons-configuration2 from 2.8.0 to 2.10.1 to address: CVE-2024-29131 CVE-2024-29133
- bcpkix-jdk18on from 1.76 to 1.78.1 to address: CVE-2024-30172 CVE-2024-30171 CVE-2024-29857
* MSQ controller: Support in-memory shuffles; towards JVM reuse.
This patch contains two controller changes that make progress towards a
lower-latency MSQ.
First, support for in-memory shuffles. The main feature of in-memory shuffles,
as far as the controller is concerned, is that they are not fully buffered. That
means that whenever a producer stage uses in-memory output, its consumer must run
concurrently. The controller determines which stages run concurrently, and when
they start and stop.
"Leapfrogging" allows any chain of sort-based stages to use in-memory shuffles
even if we can only run two stages at once. For example, in a linear chain of
stages 0 -> 1 -> 2 where all do sort-based shuffles, we can use in-memory shuffling
for each one while only running two at once. (When stage 1 is done reading input
and about to start writing its output, we can stop 0 and start 2.)
1) New OutputChannelMode enum attached to WorkOrders that tells workers
whether stage output should be in memory (MEMORY), or use local or durable
storage.
2) New logic in the ControllerQueryKernel to determine which stages can use
in-memory shuffling (ControllerUtils#computeStageGroups) and to launch them
at the appropriate time (ControllerQueryKernel#createNewKernels).
3) New "doneReadingInput" method on Controller (passed down to the stage kernels)
which allows stages to transition to POST_READING even if they are not
gathering statistics. This is important because it enables "leapfrogging"
for HASH_LOCAL_SORT shuffles, and for GLOBAL_SORT shuffles with 1 partition.
4) Moved result-reading from ControllerContext#writeReports to new QueryListener
interface, which ControllerImpl feeds results to row-by-row while the query
is still running. Important so we can read query results from the final
stage using an in-memory channel.
5) New class ControllerQueryKernelConfig holds configs that control kernel
behavior (such as whether to pipeline, maximum number of concurrent stages,
etc). Generated by the ControllerContext.
Second, a refactor towards running workers in persistent JVMs that are able to
cache data across queries. This is helpful because I believe we'll want to reuse
JVMs and cached data for latency reasons.
1) Move creation of WorkerManager and TableInputSpecSlicer to the
ControllerContext, rather than ControllerImpl. This allows managing workers and
work assignment differently when JVMs are reusable.
2) Lift the Controller Jersey resource out from ControllerChatHandler to a
reusable resource.
3) Move memory introspection to a MemoryIntrospector interface, and introduce
ControllerMemoryParameters that uses it. This makes it easier to run MSQ in
process types other than Indexer and Peon.
Both of these areas will have follow-ups that make similar changes on the
worker side.
* Address static checks.
* Address static checks.
* Fixes.
* Report writer tests.
* Adjustments.
* Fix reports.
* Review updates.
* Adjust name.
* Small changes.
Changes:
- Add new config `lagAggregate` to `LagBasedAutoScalerConfig`
- Add field `aggregateForScaling` to `LagStats`
- Use the new field/config to determine which aggregate to use to compute lag
- Remove method `Supervisor.computeLagForAutoScaler()`
Changes:
1) Check for handoff of upgraded realtime segments.
2) Drop sink only when all associated realtime segments have been abandoned.
3) Delete pending segments upon commit to prevent unnecessary upgrades and
partition space exhaustion when a concurrent replace happens. This also prevents
potential data duplication.
4) Register pending segment upgrade only on those tasks to which the segment is associated.
I'm adding OIDC context to the AuthenticationResult returned by pac4j extension. I wanted to use this context as input in OpenPolicyAgent authorization. Since AuthenticationResult already accepts context as a parameter it felt okay to pass the profile attributes there.
Fixes a bug introduced in #16296, where the sketch might not be
initialized if get() is called without calling aggregate(). Also adds
a test for this case.
Issue: #14989
The initial step in optimizing segment metadata was to centralize the construction of datasource schema in the Coordinator (#14985). Thereafter, we addressed the problem of publishing schema for realtime segments (#15475). Subsequently, our goal is to eliminate the requirement for regularly executing queries to obtain segment schema information.
This is the final change which involves publishing segment schema for finalized segments from task and periodically polling them in the Coordinator.
Buffer aggregators can contain some cached objects within them, such as
Memory references or HLL Unions. Prior to this patch, various Grouper
implementations were not releasing this state when resetting their own
internal state, which could lead to excessive memory use.
This patch renames AggregatorAdapater#close to "reset", and updates
Grouper implementations to call this reset method whenever they reset
their internal state.
The base method on BufferAggregator and VectorAggregator remains named
"close", for compatibility with existing extensions, but the contract
is adjusted to say that the aggregator may be reused after the method
is called. All existing implementations in core already adhere to this
new contract, except for the ArrayOfDoubles build flavors, which are
updated in this patch to adhere.
Additionally, this patch harmonizes buffer sketch helpers to call their
clear method "clear" rather than a mix of "clear" and "close". (Others
were already using "clear".)
Tries to address the comments made on #16284 after merged.
Changes:
- Remove method `Supervisor.getLagMetric()`
- Add method `Supervisor.computeLagForAutoScaler()`
- Remove classes `LagMetric` and `LagMetricTest`
Changes:
- Add column `task_allocator_id` to `pendingSegments` metadata table.
- Add column `upgraded_from_segment_id` to `pendingSegments` metadata table.
- Add interface `PendingSegmentAllocatingTask` and implement it by all tasks which
can allocate pending segments.
- Use `taskAllocatorId` to identify the task (and its sub-tasks or replicas) to which
a pending segment has been allocated.
- Perform active cleanup of pending segments in `TaskLockbox` once there are no
active tasks for the corresponding task allocator id.
- When committing APPEND segments, also commit all upgraded pending segments
corresponding to that task allocator id.
- When committing REPLACE segments, upgrade all overlapping pending segments in
the same transaction.
Bug:
#15724 introduced a bug where a rolling upgrade would cause all task locations
returned by the Overlord on an older version to be unknown.
Fix:
If the new API fails, fall back to single task status API which always returns a valid task location.
Currently, export creates the files at the provided destination. The addition of the manifest file will provide a list of files created as part of the manifest. This will allow easier consumption of the data exported from Druid, especially for automated data pipelines
Follow up to #16217
Changes:
- Update `OverlordClient.getReportAsMap()` to return `TaskReport.ReportMap`
- Move the following classes to `org.apache.druid.indexer.report` in the `druid-processing` module
- `TaskReport`
- `KillTaskReport`
- `IngestionStatsAndErrorsTaskReport`
- `TaskContextReport`
- `TaskReportFileWriter`
- `SingleFileTaskReportFileWriter`
- `TaskReportSerdeTest`
- Remove `MsqOverlordResourceTestClient` as it had only one method
which is already present in `OverlordResourceTestClient` itself
Changes:
- Add `TaskContextEnricher` interface to improve task management and monitoring
- Invoke `enrichContext` in `TaskQueue.add()` whenever a new task is submitted to the Overlord
- Add `TaskContextReport` to write out task context information in reports
Vishesh Garg
Akshat Jain
Zoltan Haindrich
Laksh Singla
Abhishek Radhakrishnan
zachjsh
Gian Merlino
Kashif Faraz
Clint Wylie
Sree Charan Manamala
dave-mccowan
trompa
Adarsh Sanjeev
Tom
Andreas Maechler
Karan Kumar
Sree Charan Manamala
Codegass
Igor Berman
Alberic Liu
Jan Werner
Adithya Chakilam
Arun Ramani
AmatyaAvadhanula
Jakub Matyszewski
Rishabh Singh
Parth Agrawal
YongGang