#16768 added the functionality to run compaction as a supervisor on the overlord.
This patch builds on top of that to restrict MSQ engine to compaction in the supervisor-mode only.
With these changes, users can no longer add MSQ engine as part of datasource compaction config,
or as the default cluster-level compaction engine, on the Coordinator.
The patch also adds an Overlord runtime property `druid.supervisor.compaction.engine=<msq/native>`
to specify the default engine for compaction supervisors.
Since these updates require major changes to existing MSQ compaction integration tests,
this patch disables MSQ-specific compaction integration tests -- they will be taken up in a follow-up PR.
Key changed/added classes in this patch:
* CompactionSupervisor
* CompactionSupervisorSpec
* CoordinatorCompactionConfigsResource
* OverlordCompactionScheduler
Co-authored-by: Vishesh Garg <gargvishesh@gmail.com>
In the compaction config, a range type partitionsSpec supports setting one of maxRowsPerSegment and targetRowsPerSegment. When compaction is run with the native engine, while maxRowsPerSegment = x results in segments of size x, targetRowsPerSegment = y results in segments of size 1.5 * y.
MSQ only supports rowsPerSegment = x as part of its tuning config, the resulting segment size being approx. x -- which is in line with maxRowsPerSegment behaviour in native compaction.
This PR makes the following changes:
use effective maxRowsPerSegment to pass as rowsPerSegment parameter for MSQ
persist rowsPerSegment as maxRowsPerSegment in lastCompactionState for MSQ
Use effective maxRowsPerSegment-based range spec in CompactionStatus check for both Native and MSQ.
Currently compaction with MSQ engine doesn't work for rollup on multi-value dimensions (MVDs), the reason being the default behaviour of grouping on MVD dimensions to unnest the dimension values; for instance grouping on `[s1,s2]` with aggregate `a` will result in two rows: `<s1,a>` and `<s2,a>`.
This change enables rollup on MVDs (without unnest) by converting MVDs to Arrays before rollup using virtual columns, and then converting them back to MVDs using post aggregators. If segment schema is available to the compaction task (when it ends up downloading segments to get existing dimensions/metrics/granularity), it selectively does the MVD-Array conversion only for known multi-valued columns; else it conservatively performs this conversion for all `string` columns.
* Use fuzzy matchers for compaction bytes asserts.
This still enables us to test that the bytes are zero and nonzero
when they're supposed to be, without having to ge them exactly
right. The need to get bytes exactly right makes it difficult to
ensure ITs pass when making changes to default segment metadata.
* Additional fuzziness.
Fixes#13936
In cases where a supervisor is idle and the overlord is restarted for some reason, the supervisor would
start spinning tasks again. In clusters where there are many low throughput streams, this would spike
the task count unnecessarily.
This commit compares the latest stream offset with the ones in metadata during the startup of supervisor
and sets it to idle state if they match.
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.
Changes:
- Rename `CoordinatorCompactionConfig` to `DruidCompactionConfig`
- Rename `CompactionConfigUpdateRequest` to `ClusterCompactionConfig`
- Refactor methods in `DruidCompactionConfig`
- Clean up `DataSourceCompactionConfigHistory` and its tests
- Clean up tests and add new tests
- Change API path `/druid/coordinator/v1/config/global` to `/druid/coordinator/v1/config/cluster`
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:
* 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
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).
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>
* * add new catalog IT with failure to ensure that it is run in CI
* * actually add failing test referred to and fix checkstyle
* * add some tests
* * fix checkstyle
* * add test descriptions
* * add more tests
* 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.
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.
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
Changes
- No functional changes
- Add method `AbstractBatchIndexTask.buildIngestionStatsReport()` used in several batch tasks
- Add utility method `AbstractBatchIndexTask.addBuildSegmentStatsToReport()`
- Use boolean argument to represent a full report instead of the String `full`
in internal methods. (REST API remains unchanged.)
- Rename `IngestionStatsAndErrorsTaskReportData` to `IngestionStatsAndErrors`
- Clean up some of the methods
Changes:
- Add visibility into number of segments read/published by each parallel compaction
- Add new fields `segmentsRead`, `segmentsPublished` to `IngestionStatsAndErrorsTaskReportData`
- Update `ParallelIndexSupervisorTask` to populate the new stats
A more ideal/permanent fix would be to have status checks exposed by the services,
but that'll require more code changes. So temporarily bump it to unblock CI now.
Changes
- Add `log` implementation for `AuditManager` alongwith `SQLAuditManager`
- `LoggingAuditManager` simply logs the audit event. Thus, it returns empty for
all `fetchAuditHistory` calls.
- Add new config `druid.audit.manager.type` which can take values `log`, `sql` (default)
- Add new config `druid.audit.manager.logLevel` which can take values `DEBUG`, `INFO`, `WARN`.
This gets activated only if `type` is `log`.
- Remove usage of `ConfigSerde` from `AuditManager` as audit is not just limited to configs
- Add `AuditSerdeHelper` for a single implementation of serialization/deserialization of
audit payload and other utility methods.
* Add initial draft of MarkDanglingTombstonesAsUnused duty.
* Use overshadowed segments instead of all used segments.
* Add unit test for MarkDanglingSegmentsAsUnused duty.
* Add mock call
* Simplify code.
* Docs
* shorter lines formatting
* metric doc
* More tests, refactor and fix up some logic.
* update javadocs; other review comments.
* Make numCorePartitions as 0 in the TombstoneShardSpec.
* fix up test
* Add tombstone core partition tests
* Update docs/design/coordinator.md
Co-authored-by: 317brian <53799971+317brian@users.noreply.github.com>
* review comment
* Minor cleanup
* Only consider tombstones with 0 core partitions
* Need to register the test shard type to make jackson happy
* test comments
* checkstyle
* fixup misc typos in comments
* Update logic to use overshadowed segments
* minor cleanup
* Rename duty to eternity tombstone instead of dangling. Add test for full eternity tombstone.
* Address review feedback.
---------
Co-authored-by: 317brian <53799971+317brian@users.noreply.github.com>
In the current design, brokers query both data nodes and tasks to fetch the schema of the segments they serve. The table schema is then constructed by combining the schemas of all segments within a datasource. However, this approach leads to a high number of segment metadata queries during broker startup, resulting in slow startup times and various issues outlined in the design proposal.
To address these challenges, we propose centralizing the table schema management process within the coordinator. This change is the first step in that direction. In the new arrangement, the coordinator will take on the responsibility of querying both data nodes and tasks to fetch segment schema and subsequently building the table schema. Brokers will now simply query the Coordinator to fetch table schema. Importantly, brokers will still retain the capability to build table schemas if the need arises, ensuring both flexibility and resilience.
* Vectorizing earliest for numeric
* Vectorizing earliest string aggregator
* checkstyle fix
* Removing unnecessary exceptions
* Ignoring tests in MSQ as earliest is not supported for numeric there
* Fixing benchmarks
* Updating tests as MSQ does not support earliest for some cases
* Addressing review comments by adding the following:
1. Checking capabilities first before creating selectors
2. Removing mockito in tests for numeric first aggs
3. Removing unnecessary tests
* Addressing issues for dictionary encoded single string columns where we can use the dictionary ids instead of the entire string
* Adding a flag for multi value dimension selector
* Addressing comments
* 1 more change
* Handling review comments part 1
* Handling review comments and correctness fix for latest_by when the time expression need not be in sorted order
* Updating numeric first vector agg
* Revert "Updating numeric first vector agg"
This reverts commit 4291709901.
* Updating code for correctness issues
* fixing an issue with latest agg
* Adding more comments and removing an unnecessary check
* Addressing null checks for tie selector and only vectorize false for quantile sketches
* Merge core CoordinatorClient with MSQ CoordinatorServiceClient.
Continuing the work from #12696, this patch merges the MSQ
CoordinatorServiceClient into the core CoordinatorClient, yielding a single
interface that serves both needs and is based on the ServiceClient RPC
system rather than DruidLeaderClient.
Also removes the backwards-compatibility code for the handoff API in
CoordinatorBasedSegmentHandoffNotifier, because the new API was added
in 0.14.0. That's long enough ago that we don't need backwards
compatibility for rolling updates.
* Fixups.
* Trigger GHA.
* Remove unnecessary retrying in DruidInputSource. Add "about an hour"
retry policy and h
* EasyMock
* Add "stringEncoding" parameter to DataSketches HLL.
Builds on the concept from #11172 and adds a way to feed HLL sketches
with UTF-8 bytes.
This must be an option rather than always-on, because prior to this
patch, HLL sketches used UTF-16LE encoding when hashing strings. To
remain compatible with sketch images created prior to this patch -- which
matters during rolling updates and when reading sketches that have been
written to segments -- we must keep UTF-16LE as the default.
Not currently documented, because I'm not yet sure how best to expose
this functionality to users. I think the first place would be in the SQL
layer: we could have it automatically select UTF-8 or UTF-16LE when
building sketches at query time. We need to be careful about this, though,
because UTF-8 isn't always faster. Sometimes, like for the results of
expressions, UTF-16LE is faster. I expect we will sort this out in
future patches.
* Fix benchmark.
* Fix style issues, improve test coverage.
* Put round back, to make IT updates easier.
* Fix test.
* Fix issue with filtered aggregators and add test.
* Use DS native update(ByteBuffer) method. Improve test coverage.
* Add another suppression.
* Fix ITAutoCompactionTest.
* Update benchmarks.
* Updates.
* Fix conflict.
* Adjustments.
This commit does a complete revamp of the coordinator to address problem areas:
- Stability: Fix several bugs, add capabilities to prioritize and cancel load queue items
- Visibility: Add new metrics, improve logs, revamp `CoordinatorRunStats`
- Configuration: Add dynamic config `smartSegmentLoading` to automatically set
optimal values for all segment loading configs such as `maxSegmentsToMove`,
`replicationThrottleLimit` and `maxSegmentsInNodeLoadingQueue`.
Changed classes:
- Add `StrategicSegmentAssigner` to make assignment decisions for load, replicate and move
- Add `SegmentAction` to distinguish between load, replicate, drop and move operations
- Add `SegmentReplicationStatus` to capture current state of replication of all used segments
- Add `SegmentLoadingConfig` to contain recomputed dynamic config values
- Simplify classes `LoadRule`, `BroadcastRule`
- Simplify the `BalancerStrategy` and `CostBalancerStrategy`
- Add several new methods to `ServerHolder` to track loaded and queued segments
- Refactor `DruidCoordinator`
Impact:
- Enable `smartSegmentLoading` by default. With this enabled, none of the following
dynamic configs need to be set: `maxSegmentsToMove`, `replicationThrottleLimit`,
`maxSegmentsInNodeLoadingQueue`, `useRoundRobinSegmentAssignment`,
`emitBalancingStats` and `replicantLifetime`.
- Coordinator reports richer metrics and produces cleaner and more informative logs
- Coordinator uses an unlimited load queue for all serves, and makes better assignment decisions
Introduce DruidException, an exception whose goal in life is to be delivered to a user.
DruidException itself has javadoc on it to describe how it should be used. This commit both introduces the Exception and adjusts some of the places that are generating exceptions to generate DruidException objects instead, as a way to show how the Exception should be used.
This work was a 3rd iteration on top of work that was started by Paul Rogers. I don't know if his name will survive the squash-and-merge, so I'm calling it out here and thanking him for starting on this.
Description:
Druid allows a configuration of load rules that may cause a used segment to not be loaded
on any historical. This status is not tracked in the sys.segments table on the broker, which
makes it difficult to determine if the unavailability of a segment is expected and if we should
not wait for it to be loaded on a server after ingestion has finished.
Changes:
- Track replication factor in `SegmentReplicantLookup` during evaluation of load rules
- Update API `/druid/coordinator/v1metadata/segments` to return replication factor
- Add column `replication_factor` to the sys.segments virtual table and populate it in
`MetadataSegmentView`
- If this column is 0, the segment is not assigned to any historical and will not be loaded.
* Limit select results in MSQ
* reduce number of files in test
* add truncated flag
* avoid materializing select results to list, use iterable instead
* javadocs
Kashif Faraz
Vishesh Garg
Rishabh Singh
AmatyaAvadhanula
Gian Merlino
Adithya Chakilam
Clint Wylie
George Shiqi Wu
zachjsh
Laksh Singla
YongGang
Adarsh Sanjeev
Sensor
Abhishek Radhakrishnan
Ankit Kothari
Soumyava
imply-cheddar