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 `SegmentLockReleaseAction` as it is not used anywhere.
It is not even registered as a known sub-type of `TaskAction`.
- Minor refactor in `TaskLockbox`. No functional change.
- Remove `ExpectedException` from `TaskLockboxTest`
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.
Remove software.amazon.ion:ion-java from the licenses as it is no longer a transient dependency of aws-java-sdk-core
Verified that after version 1.12.638 of aws-java-sdk-core doesnt have the ion-java as a dependency
Fixes a few minor issues with scripts.
- Add additional information around since it was confusing, and not clear that the number was the ID from github and not just the major version number.
- Fix an issue where the milestone displayed in an output message was the milestone supplied as an argument, instead of the number of the milestone the PR is already tagged against in Github, from the sent request.
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.
* specify node type so that the log filename can get resolved
* Update distribution/docker/druid.sh
Co-authored-by: Benedict Jin <asdf2014@apache.org>
---------
Co-authored-by: Benedict Jin <asdf2014@apache.org>
This parameter has been removed for awhile now as of Druid 0.23.0
https://github.com/apache/druid/pull/12187.
The code was only used in tests to verify that serialization works.
Now remove all references to avoid any confusion.
* Add native filter conversion for SCALAR_IN_ARRAY.
Main changes:
1) Add an implementation of "toDruidFilter" in ScalarInArrayOperatorConversion.
2) Split up Expressions.literalToDruidExpression into two functions, so the first
half (literalToExprEval) can be used by ScalarInArrayOperatorConversion to more
efficiently create the list of match values.
* Fix type in time arithmetic conversion.
* Test updates.
* Update test cases to use null instead of '' in default-value mode.
* Switch test from msqIncompatible to compatible with a different result.
* Update one more test.
* Fix test.
* Update tests.
* Use ExprEvalWrapper to differentiate between empty string and null.
* Fix tests some more.
* Fix test.
* Additional comment.
* Style adjustment.
* Fix tests.
* trueValue -> actualValue.
* Use different approach, DruidLiteral instead of ExprEvalWrapper.
* Revert changes in ArrayOfDoublesSketchSqlAggregatorTest.
* QueryableIndex: Close columns after failed vector cursor setup.
If anything fails while setting up a vector cursor, the prior code in
QueryableIndex would not close its ColumnCache and would therefore leak
columns. Columns often contain references to buffers that must be closed.
* Fix style.
* * 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
* add rate and stats
* better tabs
* detail
* add recent errors
* update tests
* don't let people hide the actions column because why
* don't sort on actions
* better way to agg
* add timeouts
* show error only once
* fix tests and Explain showing up
* only consider active tasks
* refresh
* fix tests
* better formatting
Description:
All the streaming ingestion tasks for a given datasource share the same lock for a given interval.
Changing lock types in the supervisor can lead to segment allocation errors due to lock conflicts
for the new tasks while the older tasks are still running.
Fix:
Allow locks of different types (EXCLUSIVE, SHARED, APPEND, REPLACE) to co-exist if they have
the same interval and the same task group.
Changes:
- Rename `DataSegmentChangeRequestAndStatus` to `DataSegmentChangeResponse`
- Rename `SegmentLoadDropHandler.Status` to `SegmentChangeStatus`
- Remove method `CoordinatorRunStats.getSnapshotAndReset()` as it was used only in
load queue peon implementations. Using an atomic reference is much simpler.
- Remove `ServerTestHelper.MAPPER`. Use existing `TestHelper.makeJsonMapper()` instead.
* test scoped jdbc driver for druidtest:/// backed DruidAvaticaTestDriver
** DecoupledTestConfig is used inside the URI - this will make it possible to attach to existing things more easily
* DruidQuidemTestBase can be used to create module level set of quidem tests
* added quidem commands: !convertedPlan, !logicalPlan, !druidPlan, !nativePlan
** for these I've used some values of the Hook which was there in calcite
* there are some shortcuts with proxies(they are only used during testing) - we can probably remove those later
* 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.
Follow up to #15705
Changes:
- Remove references to ZK-based segment loading in the docs
- Fix doc for existing config `druid.coordinator.loadqueuepeon.http.repeatDelay`
* Move druid-sql completions to dsql mode
* Use font-size 12
* Convert ace-modes to typescript
* Move aceCompleters to class member
* Use namespace imports
This PR fixes the first and last vector aggregators and improves their readability. Following changes are introduced
The folding is broken in the vectorized versions. We consider time before checking the folded object.
If the numerical aggregator gets passed any other object type for some other reason (like String), then the aggregator considers it to be folded, even though it shouldn’t be. We should convert these objects to the desired type, and aggregate them properly.
The aggregators must properly use generics. This would minimize the ClassCastException issues that can happen with mixed segment types. We are unifying the string first/last aggregators with numeric versions as well.
The aggregators must aggregate null values (https://github.com/apache/druid/blob/master/processing/src/main/java/org/apache/druid/query/aggregation/first/StringFirstLastUtils.java#L55-L56 ). The aggregator should only ignore pairs with time == null, and not value == null
Time nullity is ignored when trying to vectorize the data.
String versions initialized with DateTimes.MIN that is equal to Long.MIN / 2. This can cause incorrect results in case the user enters a custom time column. NOTE: This is still present because it would require a larger refactor in all of the versions.
There is a difference in what users might expect from the results because the code flow is changed (for example, the direction of the for loops, etc), however, this will only change the results, and not the contract set by first/last aggregators, which is that if multiple values have the same timestamp, then any of them can get picked.
If the column is non-existent, the users might expect a change in the timestamp from DateTime.MAX to Long.MAX, because the code incorrectly used DateTime.MAX to initialize the aggregator, however, in case of a custom timestamp column, this might not be the case. The SQL query might be prohibited from using any Long since it requires a cast to the timestamp function that can fail, but AFAICT native queries don't have such limitations.
#16068 modified DimensionHandlerUtils to accept complex types to be dimensions. This had an unintended side effect of allowing complex types to be joined upon (which wasn't guarded explicitly, it doesn't work).
This PR modifies the IndexedTable to reject building the index on the complex types to prevent joining on complex types. The PR adds back the check in the same place, explicitly.
* Delta Lake support for filters.
* Updates
* cleanup comments
* Docs
* Remmove Enclosed runner
* Rename
* Cleanup test
* Serde test for the Delta input source and fix jackson annotation.
* Updates and docs.
* Update error messages to be clearer
* Fixes
* Handle NumberFormatException to provide a nicer error message.
* Apply suggestions from code review
Co-authored-by: 317brian <53799971+317brian@users.noreply.github.com>
* Doc fixes based on feedback
* Yes -> yes in docs; reword slightly.
* Update docs/ingestion/input-sources.md
Co-authored-by: Laksh Singla <lakshsingla@gmail.com>
* Update docs/ingestion/input-sources.md
Co-authored-by: Laksh Singla <lakshsingla@gmail.com>
* Documentation, javadoc and more updates.
* Not with an or expression end-to-end test.
* Break up =, >, >=, <, <= into its own types instead of sub-classing.
---------
Co-authored-by: 317brian <53799971+317brian@users.noreply.github.com>
Co-authored-by: Laksh Singla <lakshsingla@gmail.com>
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:
- Add `LookupLoadingSpec` to support 3 modes of lookup loading: ALL, NONE, ONLY_REQUIRED
- Add method `Task.getLookupLoadingSpec()`
- Do not load any lookups for `KillUnusedSegmentsTask`
Akshat Jain
Laksh Singla
Sensor
aho135
Alberic Liu
Benedict Jin
Kashif Faraz
Igor Berman
Charles Smith
Adarsh Sanjeev
Zoltan Haindrich
Rishabh Singh
Rushikesh Bankar
Misha
dependabot[bot]
Pranav
Abhishek Radhakrishnan
Gian Merlino
zachjsh
Vadim Ogievetsky
AmatyaAvadhanula
Jan Werner
John Gozde
Adithya Chakilam