* * 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
* SQL tests: avoid mixing skip and cannot vectorize.
skipVectorize switches off vectorization tests completely, and
cannotVectorize turns vectorization tests into negative tests. It doesn't
make sense to use them together, so this patch makes it an error to do so,
and cleans up cases where both are mentioned.
This patch also has the effect of changing various tests from skipVectorize
to cannotVectorize, because in the past when both were mentioned,
skipVectorize would take priority.
* Fix bug with StringAnyAggregatorFactory attempting to vectorize when it cannt.
* Fix tests.
Compaction in the native engine by default records the state of compaction for each segment in the lastCompactionState segment field. This PR adds support for doing the same in the MSQ engine, targeted for future cases such as REPLACE and compaction done via MSQ.
Note that this PR doesn't implicitly store the compaction state for MSQ replace tasks; it is stored with flag "storeCompactionState": true in the query context.
Current Runtime Exceptions generated while writing frames only include the exception itself without including the name of the column they were encountered in. This patch introduces the further information in the error and makes it non-retryable.
This PR logs the segment type and reason chosen. It also adds it to the query report, to be displayed in the UI.
This PR adds a new section to the reports, segmentReport. This contains the segment type created, if the query is an ingestion, and null otherwise.
* Reduce upload buffer size in GoogleTaskLogs.
Use a 1MB upload buffer, rather than the default of 15 MB in the API client. This is
mainly because MMs may upload logs in parallel, and typically have small heaps. The
default-sized 15 MB buffers add up quickly and can cause a MM to run out of memory.
* Make bufferSize a nullable Integer. Add tests.
Support for exporting msq results to gcs bucket. This is essentially copying the logic of s3 export for gs, originally done by @adarshsanjeev in this PR - #15689
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
This PR aims to introduce Window functions on MSQ by doing the following:
Introduce a Window querykit for handling window queries along with its factory and a processor for window queries
If a window operator is present with a partition by clause, pushes the partition as a shuffle spec of the previous stage
In presence of empty OVER() clause lets all operators loose on a single rac
In presence of no empty OVER() clause, breaks down each window into individual stages
Associated machinery to handle window functions in MSQ
Introduced a separate hidden engine feature WINDOW_LEAF_OPERATOR which is set only for MSQ engine. In presence of this feature, the planner plans without the leaf operators by creating a window query over an inner scan query. In case of native this is set to false and the planner generates the leafOperators
Guardrails around materialization
Comprehensive UTs
WorkerAssignmentStrategy.AUTO was missing a check for maxWorkerCount
in the case where the inputs to a stage are not dynamically sliceable.
A common case here is when the inputs to a stage are other stages.
* MSQ: Controller checker should check for "closed" only.
Currently, the worker's controller checker will exit the worker if
the controller location is "closed" (no longer running) or if its location
is empty (i.e. location unknown).
This patch changes to only exit on "closed". We shouldn't exit on empty
location, because that may happen if the Overlord is slow to acknowledge the
location of a task.
* Fix test.
* SortMerge join support for IS NOT DISTINCT FROM.
The patch adds a "requiredNonNullKeyParts" field to the sortMerge
processor, which has the list of key parts that must be nonnull for
an equijoin condition to match. Conditions with SQL "=" are present in
the list; conditions with SQL "IS NOT DISTINCT FROM" are absent from
the list.
* Fix test.
* Update javadoc.
* Update Calcite*Test to use junit5
* change the way temp dirs are handled
* add openrewrite workflow to safeguard upgrade
* replace junitparamrunner with standard junit5 parametered tests
* update a few rules to junit5 api
* lots of boring changes
* cleanup QueryLogHook
* cleanup
* fix compile error: ARRAYS_DATASOURCE
* fix test
* remove enclosed
* empty
+TEST:TDigestSketchSqlAggregatorTest,HllSketchSqlAggregatorTest,DoublesSketchSqlAggregatorTest,ThetaSketchSqlAggregatorTest,ArrayOfDoublesSketchSqlAggregatorTest,BloomFilterSqlAggregatorTest,BloomDimFilterSqlTest,CatalogIngestionTest,CatalogQueryTest,FixedBucketsHistogramQuantileSqlAggregatorTest,QuantileSqlAggregatorTest,MSQArraysTest,MSQDataSketchesTest,MSQExportTest,MSQFaultsTest,MSQInsertTest,MSQLoadedSegmentTests,MSQParseExceptionsTest,MSQReplaceTest,MSQSelectTest,InsertLockPreemptedFaultTest,MSQWarningsTest,SqlMSQStatementResourcePostTest,SqlStatementResourceTest,CalciteSelectJoinQueryMSQTest,CalciteSelectQueryMSQTest,CalciteUnionQueryMSQTest,MSQTestBase,VarianceSqlAggregatorTest,SleepSqlTest,SqlRowTransformerTest,DruidAvaticaHandlerTest,DruidStatementTest,BaseCalciteQueryTest,CalciteArraysQueryTest,CalciteCorrelatedQueryTest,CalciteExplainQueryTest,CalciteExportTest,CalciteIngestionDmlTest,CalciteInsertDmlTest,CalciteJoinQueryTest,CalciteLookupFunctionQueryTest,CalciteMultiValueStringQueryTest,CalciteNestedDataQueryTest,CalciteParameterQueryTest,CalciteQueryTest,CalciteReplaceDmlTest,CalciteScanSignatureTest,CalciteSelectQueryTest,CalciteSimpleQueryTest,CalciteSubqueryTest,CalciteSysQueryTest,CalciteTableAppendTest,CalciteTimeBoundaryQueryTest,CalciteUnionQueryTest,CalciteWindowQueryTest,DecoupledPlanningCalciteJoinQueryTest,DecoupledPlanningCalciteQueryTest,DecoupledPlanningCalciteUnionQueryTest,DrillWindowQueryTest,DruidPlannerResourceAnalyzeTest,IngestTableFunctionTest,QueryTestRunner,SqlTestFrameworkConfig,SqlAggregationModuleTest,ExpressionsTest,GreatestExpressionTest,IPv4AddressMatchExpressionTest,IPv4AddressParseExpressionTest,IPv4AddressStringifyExpressionTest,LeastExpressionTest,TimeFormatOperatorConversionTest,CombineAndSimplifyBoundsTest,FiltrationTest,SqlQueryTest,CalcitePlannerModuleTest,CalcitesTest,DruidCalciteSchemaModuleTest,DruidSchemaNoDataInitTest,InformationSchemaTest,NamedDruidSchemaTest,NamedLookupSchemaTest,NamedSystemSchemaTest,RootSchemaProviderTest,SystemSchemaTest,CalciteTestBase,SqlResourceTest
* use @Nested
* add rule to remove enclosed; upgrade surefire
* remove enclosed
* cleanup
* add comment about surefire exclude
TaskTracker's status is null when TaskTrackers are first set up, and
stay null until the first status call comes back. This patch handles
that case and sets the status code to null in the WorkerStats object
in live reports.
Prior to this patch, when canceled, workers would keep trying to contact
the controller: they would attempt to report an error, and if they were
in the midst of some other call (like a counters push) they would keep
trying it.
This can cause cancellation to be delayed, because the controller shuts
down its HTTP server before it cancels workers. Workers are then stuck
retrying calls to the controller that will never succeed. The retry loops
are broken when the controller gives up on them (one minute later) and
exits for real. Then, the controller failure detection logic on the worker
detects that the controller has failed, and the worker finally shuts down.
This patch speeds up worker cancellation by bypassing communication
with the controller. There is no real need for it. If the controller
canceled the workers, it isn't interested in further communications from
them. If the workers were canceled out-of-band, the controller can
detect this through worker monitoring and report it as a WorkerFailed
error.
Changes:
- Remove deprecated `DruidException` (old one) and `EntryExistsException`
- Use newly added comprehensive `DruidException` instead
- Update error message in `SqlMetadataStorageActionHandler` when max packet limit is violated.
- Factor out common code from several faults into `BaseFault`.
- Slightly update javadoc in `DruidException` to render it correctly
- Remove unused classes `SegmentToMove`, `SegmentToDrop`
- Move `ServletResourceUtils` from module `druid-processing` to `druid-server`
- Add utility method to build error Response from `DruidException`.
* MSQ: Validate that strings and string arrays are not mixed.
When multi-value strings and string arrays coexist in the same column,
it causes problems with "classic MVD" style queries such as:
select * from wikipedia -- fails at runtime
select count(*) from wikipedia where flags = 'B' -- fails at planning time
select flags, count(*) from wikipedia group by 1 -- fails at runtime
To avoid these problems, this patch adds type verification for INSERT
and REPLACE. It is targeted: the only type changes that are blocked are
string-to-array and array-to-string. There is also a way to exclude
certain columns from the type checks, if the user really knows what
they're doing.
* Fixes.
* Tests and docs and error messages.
* More docs.
* Adjustments.
* Adjust message.
* Fix tests.
* Fix test in DV mode.
* MSQ: Plan without implicit sorting.
This patch adds an EngineFeature "GROUPBY_IMPLICITLY_SORTS" and sets
it true for native, false for MSQ. It's useful for two reasons:
1) In the future we'll likely want MSQ to hash-partition for GROUP BY
instead of using a global sort, which would mean MSQ would not
implicitly ORDER BY when there is a GROUP BY.
2) When doing REPLACE with MSQ, CLUSTERED BY is transformed to ORDER BY.
We should retain that ORDER BY, as it may be a subset of the GROUP BY,
and it is important to remember which fields the user wanted to include in
range shard specs.
* Fix tests.
* Fix tests for real.
* Fix test.
* Pull up literals in InputAccessor
* pull up literals in `InputAccessor`
* remove the need to pass `constants` of `Window` operator
Fixes#15353
* update test
* enable relax_nulls
The previously used GCS API client library returned last update time for objects directly in milliseconds. The new library returns it in OffsetDateTime format which was being converted to seconds and stored against the object. This fix converts the time back to ms before storing it.
Fixes # size blowup regression introduced in https://github.com/apache/druid/pull/15443
This PR removes the transitive dependency of ranger-plugins-audit to reduce the size of the compiled artifacts
* add aws-logs-sdk to ensure that all the transitive dependencies are satisfied
* replace aws-bundle-sdk with aws-logs-sdk
* add additional guidance on ranger update, add dependency ignore to satisfy dependency analyzer
* add aws-sdk-logs to list of ignored dependencies to satisfy the maven plugin
* align aws-sdk versions
BaseNodeRoleWatcher counts down cacheInitialized after a timeout, but also sets some flag that it was a timed-out initialization. and call nodeViewInitializationTimedOut (new method on listeners) instead of nodeViewInitialized. Then listeners can do what is most appropriate with this information.
* Add support for AzureDNSZone enabled storage accounts used for deep storage
Added a new config to AzureAccountConfig
`storageAccountEndpointSuffix`
which allows the user to specify a storage account endpoint suffix where the underlying
storage account is enabled for AzureDNSZone. The previous config `endpointSuffix`, did not allow
support for such accounts. The previous config has been deprecated in favor of this new config. Also
fixed an issue where `managedIdentityClientId` was not being set properly
* * address review comments
* * add back azure government link and docs
* Move retries into DataSegmentPusher implementations.
The individual implementations know better when they should and should
not retry. They can also generate better error messages.
The inspiration for this patch was a situation where EntityTooLarge was
generated by the S3DataSegmentPusher, and retried uselessly by the
retry harness in PartialSegmentMergeTask.
* Fix missing var.
* Adjust imports.
* Tests, comments, style.
* Remove unused import.
* MSQ: Nicer error when sortMerge join falls back to broadcast.
In certain cases, joins run as broadcast even when the user hinted
that they wanted sortMerge. This happens when the sortMerge algorithm
is unable to process the join, because it isn't a direct comparison
between two fields on the LHS and RHS.
When this happens, the error message from BroadcastTablesTooLargeFault
is quite confusing, since it mentions that you should try sortMerge
to fix it. But the user may have already configured sortMerge.
This patch fixes it by having two error messages, based on whether
broadcast join was used as a primary selection or as a fallback selection.
* Style.
* Better message.
Currently, while reading results from realtime tasks, requests are sent on a segment level. This is slightly wasteful, as when contacting a data servers, it is possible to transfer results for all segments which it is hosting, instead of only one segment at a time.
One change this PR makes is to group the segments on the basis of servers. This reduces the number of queries to data servers made. Since we don't have access to the number of rows for realtime segments, the grouping is done with a fixed estimated number of rows for each realtime segment.
The code in the groupBy engine and the topN engine assume that the dimensions are comparable and can call dimA.compareTo(dimB) to sort the dimensions and group them together.
This works well for the primitive dimensions, because they are Comparable, however falls apart when the dimensions can be arrays (or in future scenarios complex columns). In cases when the dimensions are not comparable, Druid resorts to having a wrapper type ComparableStringArray and ComparableList, which is a Comparable, based on the list comparator.
Changes:
- Add visibility into number of records processed by each streaming task per partition
- Add field `recordsProcessed` to `IngestionStatsAndErrorsTaskReportData`
- Populate number of records processed per partition in `SeekableStreamIndexTaskRunner`
This PR contains a portion of the changes from the inactive draft PR for integrating the catalog with the Calcite planner https://github.com/apache/druid/pull/13686 from @paul-rogers, Refactoring the IngestHandler and subclasses to produce a validated SqlInsert instance node instead of the previous Insert source node. The SqlInsert node is then validated in the calcite validator. The validation that is implemented as part of this pr, is only that for the source node, and some of the validation that was previously done in the ingest handlers. As part of this change, the partitionedBy clause can be supplied by the table catalog metadata if it exists, and can be omitted from the ingest time query in this case.
Apache Druid brings the dependency json-path which is affected by CVE-2023-51074.
Its latest version 2.9.0 fixes the above CVE.
Append function has been added to json-path and so the unit test to check for the append function not present has been updated.
---------
Co-authored-by: Xavier Léauté <xvrl@apache.org>
* Rework ExprMacro base classes to simplify implementations.
This patch removes BaseScalarUnivariateMacroFunctionExpr, adds
BaseMacroFunctionExpr at the top of the hierarchy (a suitable base class
for ExprMacros that take either arrays or scalars), and adds an
implementation for "visit" to BaseMacroFunctionExpr.
The effect on implementations is generally cleaner code:
- Exprs no longer need to implement "visit".
- Exprs no longer need to implement "stringify", even if they don't
use all of their args at runtime, because BaseMacroFunctionExpr has
access to even unused args.
- Exprs that accept arrays can extend BaseMacroFunctionExpr and
inherit a bunch of useful methods. The only one they need to
implement themselves that scalar exprs don't is "supplyAnalyzeInputs".
* Make StringDecodeBase64UTFExpression a static class.
* Remove unused import.
* Formatting, annotation changes.
* Fix HllSketchHolderObjectStrategy#isSafeToConvertToNullSketch.
The prior code from #15162 was reading only the low-order byte of an int
representing the size of a coupon set. As a result, it would erroneously
believe that a coupon set with a multiple of 256 elements was empty.
During ingestion, incremental segments are created in memory for the different time chunks and persisted to disk when certain thresholds are reached (max number of rows, max memory, incremental persist period etc). In the case where there are a lot of dimension and metrics (1000+) it was observed that the creation/serialization of incremental segment file format for persistence and persisting the file took a while and it was blocking ingestion of new data. This affected the real-time ingestion. This serialization and persistence can be parallelized across the different time chunks. This update aims to do that.
The patch adds a simple configuration parameter to the ingestion tuning configuration to specify number of persistence threads. The default value is 1 if it not specified which makes it the same as it is today.
PassthroughAggregatorFactory overrides a deprecated method in the AggregatorFactory, on which it relies on for serializing one of its fields complexTypeName. This was accidentally removed, leading to a bug in the factory, where the type name doesn't get serialized properly, and places null in the type name. This PR revives that method with a different name and adds tests for the same.
- After upgrading the pac4j version in: https://github.com/apache/druid/pull/15522. We were not able to access the druid ui.
- Upgraded the Nimbus libraries version to a compatible version to pac4j.
- In the older pac4j version, when we return RedirectAction there we also update the webcontext Response status code and add the authentication URL to the header. But in the newer pac4j version, we just simply return the RedirectAction. So that's why it was not getting redirected to the generated authentication URL.
- To fix the above, I have updated the NOOP_HTTP_ACTION_ADAPTER to JEE_HTTP_ACTION_ADAPTER and it updates the HTTP Response in context as per the HTTP Action.
As part of becoming FIPS compliance, we are seeing this error: salt must be at least 128 bits when we run the Druid code against FIPS Compliant cryptographic security providers.
This PR fixes the salt size used in Pac4jSessionStore.java
### Description
Our Kinesis consumer works by using the [GetRecords API](https://docs.aws.amazon.com/kinesis/latest/APIReference/API_GetRecords.html) in some number of `fetchThreads`, each fetching some number of records (`recordsPerFetch`) and each inserting into a shared buffer that can hold a `recordBufferSize` number of records. The logic is described in our documentation at: https://druid.apache.org/docs/27.0.0/development/extensions-core/kinesis-ingestion/#determine-fetch-settings
There is a problem with the logic that this pr fixes: the memory limits rely on a hard-coded “estimated record size” that is `10 KB` if `deaggregate: false` and `1 MB` if `deaggregate: true`. There have been cases where a supervisor had `deaggregate: true` set even though it wasn’t needed, leading to under-utilization of memory and poor ingestion performance.
Users don’t always know if their records are aggregated or not. Also, even if they could figure it out, it’s better to not have to. So we’d like to eliminate the `deaggregate` parameter, which means we need to do memory management more adaptively based on the actual record sizes.
We take advantage of the fact that GetRecords doesn’t return more than 10MB (https://docs.aws.amazon.com/streams/latest/dev/service-sizes-and-limits.html ):
This pr:
eliminates `recordsPerFetch`, always use the max limit of 10000 records (the default limit if not set)
eliminate `deaggregate`, always have it true
cap `fetchThreads` to ensure that if each fetch returns the max (`10MB`) then we don't exceed our budget (`100MB` or `5% of heap`). In practice this means `fetchThreads` will never be more than `10`. Tasks usually don't have that many processors available to them anyway, so in practice I don't think this will change the number of threads for too many deployments
add `recordBufferSizeBytes` as a bytes-based limit rather than records-based limit for the shared queue. We do know the byte size of kinesis records by at this point. Default should be `100MB` or `10% of heap`, whichever is smaller.
add `maxBytesPerPoll` as a bytes-based limit for how much data we poll from shared buffer at a time. Default is `1000000` bytes.
deprecate `recordBufferSize`, use `recordBufferSizeBytes` instead. Warning is logged if `recordBufferSize` is specified
deprecate `maxRecordsPerPoll`, use `maxBytesPerPoll` instead. Warning is logged if maxRecordsPerPoll` is specified
Fixed issue that when the record buffer is full, the fetchRecords logic throws away the rest of the GetRecords result after `recordBufferOfferTimeout` and starts a new shard iterator. This seems excessively churny. Instead, wait an unbounded amount of time for queue to stop being full. If the queue remains full, we’ll end up right back waiting for it after the restarted fetch.
There was also a call to `newQ::offer` without check in `filterBufferAndResetBackgroundFetch`, which seemed like it could cause data loss. Now checking return value here, and failing if false.
### Release Note
Kinesis ingestion memory tuning config has been greatly simplified, and a more adaptive approach is now taken for the configuration. Here is a summary of the changes made:
eliminates `recordsPerFetch`, always use the max limit of 10000 records (the default limit if not set)
eliminate `deaggregate`, always have it true
cap `fetchThreads` to ensure that if each fetch returns the max (`10MB`) then we don't exceed our budget (`100MB` or `5% of heap`). In practice this means `fetchThreads` will never be more than `10`. Tasks usually don't have that many processors available to them anyway, so in practice I don't think this will change the number of threads for too many deployments
add `recordBufferSizeBytes` as a bytes-based limit rather than records-based limit for the shared queue. We do know the byte size of kinesis records by at this point. Default should be `100MB` or `10% of heap`, whichever is smaller.
add `maxBytesPerPoll` as a bytes-based limit for how much data we poll from shared buffer at a time. Default is `1000000` bytes.
deprecate `recordBufferSize`, use `recordBufferSizeBytes` instead. Warning is logged if `recordBufferSize` is specified
deprecate `maxRecordsPerPoll`, use `maxBytesPerPoll` instead. Warning is logged if maxRecordsPerPoll` is specified
* Clear "lineSplittable" for JSON when using KafkaInputFormat.
JsonInputFormat has a "withLineSplittable" method that can be used to
control whether JSON is read line-by-line, or as a whole. The intent
is that in streaming ingestion, "lineSplittable" is false (although it
can be overridden by "assumeNewlineDelimited"), and in batch ingestion,
lineSplittable is true.
When a "json" format is wrapped by a "kafka" format, this isn't set
properly. This patch updates KafkaInputFormat to set this on an
underlying "json" format.
The tests for KafkaInputFormat were overriding the "lineSplittable"
parameter explicitly, which wasn't really fair, because that made them
unrealistic to what happens in production. Now they omit the parameter
and get the production behavior.
* Add test.
* Fix test coverage.
* Faster parsing: reduce String usage, list-based input rows.
Three changes:
1) Reworked FastLineIterator to optionally avoid generating Strings
entirely, and reduce copying somewhat. Benefits the line-oriented
JSON, CSV, delimited (TSV), and regex formats.
2) In the delimited (TSV) format, when the delimiter is a single byte,
split on UTF-8 bytes directly.
3) In CSV and delimited (TSV) formats, use list-based input rows when
the column list is provided upfront by the user.
* Fix style.
* Fix inspections.
* Restore validation.
* Remove fastutil-extra.
* Exception type.
* Fixes for error messages.
* Fixes for null handling.
MSQ now allows empty ingest queries by default. For such queries that don't generate any output rows, the query counters in the async status result object/task report don't contain numTotalRows and totalSizeInBytes. These properties when not set/undefined can be confusing to API clients. For example, the web-console treats it as unknown values.
This patch fixes the counters by explicitly reporting them as 0 instead of null for empty ingest queries.
* support groups windowing mode; which is a close relative of ranges (but not in the standard)
* all windows with range expressions will be executed wit it groups
* it will be 100% correct in case for both bounds its true that: isCurrentRow() || isUnBounded()
* this covers OVER ( ORDER BY COL )
* for other cases it will have some chances of getting correct results...
Changes:
- Add new task context flag useConcurrentLocks.
- This can be set for an individual task or at a cluster level using `druid.indexer.task.default.context`.
- When set to true, any appending task would use an APPEND lock and any other
ingestion task would use a REPLACE lock when using time chunk locking.
- If false (default), we fall back on the context flag taskLockType and then useSharedLock.
* Add ImmutableLookupMap for static lookups.
This patch adds a new ImmutableLookupMap, which comes with an
ImmutableLookupExtractor. It uses a fastutil open hashmap plus two
lists to store its data in such a way that forward and reverse
lookups can both be done quickly. I also observed footprint to be
somewhat smaller than Java HashMap + MapLookupExtractor for a 1 million
row lookup.
The main advantage, though, is that reverse lookups can be done much
more quickly than MapLookupExtractor (which iterates the entire map
for each call to unapplyAll). This speeds up the recently added
ReverseLookupRule (#15626) during SQL planning with very large lookups.
* Use in one more test.
* Fix benchmark.
* Object2ObjectOpenHashMap
* Fixes, and LookupExtractor interface update to have asMap.
* Remove commented-out code.
* Fix style.
* Fix import order.
* Add fastutil.
* Avoid storing Map entries.
* Faster k-way merging using tournament trees, 8-byte key strides.
Two speedups for FrameChannelMerger (which does k-way merging in MSQ):
1) Replace the priority queue with a tournament tree, which does fewer
comparisons.
2) Compare keys using 8-byte strides, rather than 1 byte at a time.
* Adjust comments.
* Fix style.
* Adjust benchmark and test.
* Add eight-list test (power of two).
Add class PasswordHashGenerator. Move hashing logic from BasicAuthUtils to this new class.
Add cache in the hash generator to contain the computed hash of passwords and boost validator performance
Cache has max size 1000 and expiry 1 hour
Key of the cache is an SHA-256 hash of the (password + random salt generated on service startup)
Currently, If 2 tasks are consuming from the same partitions, try to publish the segment and update the metadata, the second task can fail because the end offset stored in the metadata store doesn't match with the start offset of the second task. We can fix this by retrying instead of failing.
AFAIK apart from the above issue, the metadata mismatch can happen in 2 scenarios:
- when we update the input topic name for the data source
- when we run 2 replicas of ingestion tasks(1 replica will publish and 1 will fail as the first replica has already updated the metadata).
Implemented the comparable function to compare the last committed end offset and new Sequence start offset. And return a specific error msg for this.
Add retry logic on indexers to retry for this specific error msg.
Updated the existing test case.
Added support for Azure Government storage in Druid Azure-Extensions. This enhancement allows the Azure-Extensions to be compatible with different Azure storage types by updating the endpoint suffix from a hardcoded value to a configurable one.
* overhaul DruidPredicateFactory to better handle 3VL
fixes some bugs caused by some limitations of the original design of how DruidPredicateFactory interacts with 3-value logic. The primary impacted area was with how filters on values transformed with expressions or extractionFn which turn non-null values into nulls, which were not possible to be modelled with the 'isNullInputUnknown' method
changes:
* adds DruidObjectPredicate to specialize string, array, and object based predicates instead of using guava Predicate
* DruidPredicateFactory now uses DruidObjectPredicate
* introduces DruidPredicateMatch enum, which all predicates returned from DruidPredicateFactory now use instead of booleans to indicate match. This means DruidLongPredicate, DruidFloatPredicate, DruidDoublePredicate, and the newly added DruidObjectPredicate apply methods all now return DruidPredicateMatch. This allows matchers and indexes
* isNullInputUnknown has been removed from DruidPredicateFactory
* rename, fix test
* adjust
* style
* npe
* more test
* fix default value mode to not match new test
FILTER_INTO_JOIN is mainly run along with the other rules with the Volcano planner; however if the query starts highly underdefined (join conditions in the where clauses) that generic query could give a lot of room for the other rules to play around with only enabled it for when the join uses subqueries for its inputs.
PROJECT_FILTER rule is not that useful. and could increase planning times by providing new plans. This problem worsened after we started supporting inner joins with arbitrary join conditions in https://github.com/apache/druid/pull/15302
* Reverse lookup fixes and enhancements.
1) Add a "mayIncludeUnknown" parameter to DimFilter#optimize. This is important
because otherwise the reverse-lookup optimization is done improperly when
the "in" filter appears under a "not", and the lookup extractionFn may return
null for some possible values of the filtered column. The "includeUnknown" test
cases in InDimFilterTest illustrate the difference in behavior.
2) Enhance InDimFilter#optimizeLookup to handle "mayIncludeUnknown", and to be able
to do a reverse lookup in a wider variety of cases.
3) Make "unapply" protected in LookupExtractor, and move callers to "unapplyAll".
The main reason is that MapLookupExtractor, a common implementation, lacks a
reverse mapping and therefore does a scan of the map for each call to "unapply".
For performance sake these calls need to be batched.
* Remove optimize call from BloomDimFilter.
* Follow the law.
* Fix tests.
* Fix imports.
* Switch function.
* Fix tests.
* More tests.
* Allow empty inserts and replace.
- Introduce a new query context failOnEmptyInsert which defaults to false.
- When this context is false (default), MSQE will now allow empty inserts and replaces.
- When this context is true, MSQE will throw the existing InsertCannotBeEmpty MSQ fault.
- For REPLACE ALL over an ALL grain segment, the query will generate a tombstone spanning eternity
which will be removed eventually be the coordinator.
- Add unit tests in MSQInsertTest, MSQReplaceTest to test the new default behavior (i.e., when failOnEmptyInsert = false)
- Update unit tests in MSQFaultsTest to test the non-default behavior (i.e., when failOnEmptyInsert = true)
* Ignore test to see if it's the culprit for OOM
* Add heap dump config
* Bump up -Xmx from 1500 MB to 2048 MB
* Add steps to tarball and collect hprof dump to GHA action
* put back mx to 1500MB to trigger the failure
* add the step to reusable unit test workflow as well
* Revert the temp heap dump & @Ignore changes since max heap size is increased
* Minor updates
* Review comments
1. Doc suggestions
2. Add tests for empty insert and replace queries with ALL grain and limit in the
default failOnEmptyInsert mode (=false). Add similar tests to MSQFaultsTest with
failOnEmptyInsert = true, so the query does fail with an InsertCannotBeEmpty fault.
3. Nullable annotation and javadocs
* Add comment
replace_limit.patch
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.
The PR addresses 2 things:
Add MSQ durable storage connector for GCS
Change GCS client library from the old Google API Client Library to the recommended Google Cloud Client Library. Ref: https://cloud.google.com/apis/docs/client-libraries-explained
* Upgrade org.pac4j:pac4j-oidc to 4.5.5 to address CVE-2021-44878
* add CVE suppression and notes, since vulnerability scan still shows this CVE
* Add tests to improve coverage
Update of direct dependencies:
* kubernetes java-client to 19.0.0
* docker-java-bom to 3.3.4
In order to update transitive dependencies:
* okio to 3.6.0
* bcjava to 1.76
To address CVES:
- CVE-2023-3635 in okio
- CVE-2023-33201 in bcjava
---------
Co-authored-by: Xavier Léauté <xvrl@apache.org>
Fixes a potential NPE which could occur while folding the HllSketchAggregator. If the sketch is null, druid could return a null HllSketchHolder object. Adding a null check here could help here
Resolves a null pointer exception in HllSketchAggregatorFactory
This change completes the change introduced in #15461
and unifies the version of gson dependency used between all the modules.
gson is used by kubernetes-extension, avro-extensions, ranger-security,
and as a test dependency in several core modules.
---------
Co-authored-by: Xavier Léauté <xl+github@xvrl.net>
* Excluding jackson-jaxrs dependency from ranger-plugin-common to address CVE regression introduced by ranger-upgrade: CVE-2019-10202, CVE-2019-10172
* remove the reference to outdated ranger 2.0 from the docs
---------
Co-authored-by: Xavier Léauté <xl+github@xvrl.net>
Recent upgrade of ranger introduced CVE regressions due to outdated elasticsearch components.
Druid-ranger-plugin does not elasticsearch components , and they have been explicitly removed.
Update woodstox-core to 6.4.0 to address GHSA-3f7h-mf4q-vrm4
This PR revives #14978 with a few more bells and whistles. Instead of an unconditional cross-join, we will now split the join condition such that some conditions are now evaluated post-join. To decide what sub-condition goes where, I have refactored DruidJoinRule class to extract unsupported sub-conditions. We build a postJoinFilter out of these unsupported sub-conditions and push to the join.
* update confluent's dependencies to common, supported version
Update io.confluent.* dependencies to common, updated version 6.2.12
currently used versions are EOL
* move version definition to the top level pom
Changes:
- Fix log `Got end of partition marker for partition [%s] from task [%s] in discoverTasks`
by fixing order of args
- Simplify in-line classes by using lambda
- Update kill task message from `Task [%s] failed to respond to [set end offsets]
in a timely manner, killing task` to `Failed to set end offsets, killing task`
- Clean up tests
There is a problem with Quantiles sketches and KLL Quantiles sketches.
Queries using the histogram post-aggregator fail if:
- the sketch contains at least one value, and
- the values in the sketch are all equal, and
- the splitPoints argument is not passed to the post-aggregator, and
- the numBins argument is greater than 2 (or not specified, which
leads to the default of 10 being used)
In that case, the query fails and returns this error:
{
"error": "Unknown exception",
"errorClass": "org.apache.datasketches.common.SketchesArgumentException",
"host": null,
"errorCode": "legacyQueryException",
"persona": "OPERATOR",
"category": "RUNTIME_FAILURE",
"errorMessage": "Values must be unique, monotonically increasing and not NaN.",
"context": {
"host": null,
"errorClass": "org.apache.datasketches.common.SketchesArgumentException",
"legacyErrorCode": "Unknown exception"
}
}
This behaviour is undesirable, since the caller doesn't necessarily
know in advance whether the sketch has values that are diverse
enough. With this change, the post-aggregators return [N, 0, 0...]
instead of crashing, where N is the number of values in the sketch,
and the length of the list is equal to numBins. That is what they
already returned for numBins = 2.
Here is an example of a query that would fail:
{"queryType":"timeseries",
"dataSource": {
"type": "inline",
"columnNames": ["foo", "bar"],
"rows": [
["abc", 42.0],
["def", 42.0]
]
},
"intervals":["0000/3000"],
"granularity":"all",
"aggregations":[
{"name":"the_sketch", "fieldName":"bar", "type":"quantilesDoublesSketch"}],
"postAggregations":[
{"name":"the_histogram",
"type":"quantilesDoublesSketchToHistogram",
"field":{"type":"fieldAccess","fieldName":"the_sketch"},
"numBins": 3}]}
I believe this also fixes issue #10585.
* Fix capacity response in mm-less ingestion (#14888)
Changes:
- Fix capacity response in mm-less ingestion.
- Add field usedClusterCapacity to the GET /totalWorkerCapacity response.
This API should be used to get the total ingestion capacity on the overlord.
- Remove method `isK8sTaskRunner` from interface `TaskRunner`
* Using Map to perform comparison
* Minor Change
---------
Co-authored-by: George Shiqi Wu <george.wu@imply.io>
Saw bug where MSQ controller task would continue to hold the task slot even after cancel was issued.
This was due to a deadlock created on work launch. The main thread was waiting for tasks to spawn and the cancel thread was waiting for tasks to finish.
The fix was to instruct the MSQWorkerTaskLauncher thread to stop creating new tasks which would enable the main thread to unblock and release the slot.
Also short circuited the taskRetriable condition. Now the check is run in the MSQWorkerTaskLauncher thread as opposed to the main event thread loop. This will result in faster task failure in case the task is deemed to be non retriable.
* MSQ generates tombstones honoring the query's granularity.
This change tweaks to only account for the infinite-interval tombstones.
For finite-interval tombstones, the MSQ query granualrity will be used
which is consistent with how MSQ works.
* more tests and some cleanup.
* checkstyle
* comment edits
* Throw TooManyBuckets fault based on review; add more tests.
* Add javadocs for both methods on reconciling the methods.
* review: Move testReplaceTombstonesWithTooManyBucketsThrowsException to MsqFaultsTest
* remove unused imports.
* Move TooManyBucketsException to indexing package for shared exception handling.
* lower max bucket for tests and fixup count
* Advance and count the iterator.
* checkstyle
* + Fix for Flaky Test
* + Replacing TreeMap with LinkedHashMap
* + Changing data structure from LinkedHashMap to HashMap
* Fixed flaky test in S3DataSegmentPusherConfigTest.testSerializationValidatingMaxListingLength
* Minor Changes
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.
Alberic Liu
zachjsh
Jan Werner
Gian Merlino
Adarsh Sanjeev
Adithya Chakilam
Arun Ramani
AmatyaAvadhanula
Jakub Matyszewski
Zoltan Haindrich
Laksh Singla
Rishabh Singh
Sree Charan Manamala
Parth Agrawal
Vishesh Garg
Akshat Jain
YongGang
Kashif Faraz
Parag Jain
Soumyava
Aru Raghuwanshi
AlbericByte
Karan Kumar
George Shiqi Wu
Clint Wylie
Lasse Mammen
Pramod Immaneni
PANKAJ KUMAR
Abhishek Radhakrishnan
Maytas Monsereenusorn
Ankit Kothari
Misha
Jonathan Wei
Keerthana Srikanth
Abhishek Agarwal
Vivek Dhiman
Magnus Henoch
Krishna Anandan
Pranav