* Enable most IntelliJ 'Probable bugs' inspections
* Fix in RemoteTestNG
* Fix IndexSpec's equals() and hashCode() to include longEncoding
* Fix inspection errors
* Extract global isntance of natural().nullsFirst(); address comments
* Fix
* Use noinspection comments instead of SuppressWarnings on method for IntelliJ-specific inspections
* Prohibit Ordering.natural().nullsFirst() using Checkstyle
* Make using implicit system charset an error
* Use StringUtils.toUtf8() and fromUtf8() instead of String.getBytes() and new String()
* Use English locale in StringUtils.safeFormat()
* Restore comment
* Adding s3a schema and s3a implem to hdfs storage module.
* use 2.7.3
* use segment pusher to make loadspec
* move getStorageDir and makeLoad spec under DataSegmentPusher
* fix uts
* fix comment part1
* move to hadoop 2.8
* inject deep storage properties
* set version to 2.7.3
* fix build issue about static class
* fix comments
* fix default hadoop default coordinate
* fix create filesytem
* downgrade aws sdk
* bump the version
* fix TestKafkaExtractionCluster fail due to port already used
* explicitly unmap hydrant files when abandonSegment to recyle mmap memory
* address the comments
* apply to AppenderatorImpl
* Refactoring Appenderator
1) Added publishExecutor and handoffExecutor for background publishing and handing segments off
2) Change add() to not move segments out in it
* Address comments
1) Remove publishTimeout for KafkaIndexTask
2) Simplifying registerHandoff()
3) Add increamental handoff test
* Remove unused variable
* Add persist() to Appenderator and more tests for AppenderatorDriver
* Remove unused imports
* Fix strict build
* Address comments
* move ProtoBufInputRowParser from processing module to protobuf extensions
* Ported PR #3509
* add DynamicMessage
* fix local test stuff that slipped in
* add license header
* removed redundant type name
* removed commented code
* fix code style
* rename ProtoBuf -> Protobuf
* pom.xml: shade protobuf classes, handle .desc resource file as binary file
* clean up error messages
* pick first message type from descriptor if not specified
* fix protoMessageType null check. add test case
* move protobuf-extension from contrib to core
* document: add new configuration keys, and descriptions
* update document. add examples
* move protobuf-extension from contrib to core (2nd try)
* touch
* include protobuf extensions in the distribution
* fix whitespace
* include protobuf example in the distribution
* example: create new pb obj everytime
* document: use properly quoted json
* fix whitespace
* bump parent version to 0.10.1-SNAPSHOT
* ignore Override check
* touch
* Do not re-create prioritized servers on each call in server selector and extend TierSelectorStrategy interface with a method to pick multiple elements at once
* Fix compilation
* Replaces use of CountingMap with Object2LongMap from fastutil.
* Remove CountingMap classes and minor fixes
* Added additional test cases for DatasourceInputFormat.
* Added additional test cases for CoordinatorStats.
* Not materializing segment list.
* Put in this fix because it is failing the test on its expected behavior.
* Added missing header.
* Improve concurrency of SegmentManager
* Fix SegmentManager and add more tests
* Add more tests
* Add null check to TimelineEntry
* Remove empty data source and check null in getTimeline()
* Add a comment for returning null in compute()
* Make SegmentManager LazySingleton
* Timeout and maxScatterGatherBytes handling for queries run by Druid SQL
* Address PR comments
* Fix contexts in CalciteQueryTest
* Fix contexts in QuantileSqlAggregatorTest
* Add a ServerType for peons
* Add toString() method, toString() test, unsupported type check
* Use ServerType enum in DruidServer and DruidServerMetadata
* Don't pass QueryMetrics down in concurrent and async QueryRunners
* Rename QueryPlus.threadSafe() to withoutThreadUnsafeState(); Update QueryPlus.withQueryMetrics() Javadocs; Fix generics in MetricsEmittingQueryRunner and CpuTimeMetricQueryRunner; Make DefaultQueryMetrics to fail fast on modifications from concurrent threads
This is useful for putting them behind load balancers or proxies, as it lets
the load balancer know which server is currently active through an http health
check.
Also makes the method naming a little more consistent between coordinator and
overlord code.
* Optional long-polling based segment announcement via HTTP instead of Zookeeper
* address review comments
* make endpoint /druid-internal/v1 instead of /druid/internal so that jetty qos filters can be configured easily when needed
* update segment callback initialization to be called only after first segment list fetch has been succeeded from all servers
* address review comments
* remove size check not required anymore as only segment servers announce themselves and not all peon processes
* annouce segment server on historical only after cached segments are loaded
* fix checkstyle errors
* Fix lz4 library incompatibility in kafka-indexing-service extension #3266
* Bumped Kafka version to 0.10.2.0 for : Fix lz4 library incompatibility in kafka-indexing-service extension #3266
* Replaced Lists.newArrayList() with Collections.singletonList() For Fix lz4 library incompatibility in kafka-indexing-service extension #4115
* Fixed: Average Server Percent Used: NaN% Error when server startup is in progress #4214
* Make Errorprone the default compiler
* Address comments
* Make Error Prone's ClassCanBeStatic rule a error
* Preconditions allow only %s pattern
* Fix DruidCoordinatorBalancerTester
* Try to give the compiler more memory
* Remove distribution module activation on jdk 1.8 because only jdk 1.8 is used now
* Don't show compiler warnings
* Try different travis script
* Fix travis.yml
* Make Error Prone optional again
* For error-prone compiler
* Increase compiler's maxmem
* Don't run Error Prone for benchmarks because of OOM
* Skip install step in Travis
* Remove MetricHolder.writeToChannel()
* In travis.yml, check compilation before tests, because it may fail faster
* Add QueryPlus. Add QueryRunner.run(QueryPlus, Map) method with default implementation, to replace QueryRunner.run(Query, Map).
* Fix GroupByMergingQueryRunnerV2
* Fix QueryResourceTest
* Expand the comment to Query.run(walker, context)
* Remove legacy version of BySegmentSkippingQueryRunner.doRun()
* Add LegacyApiQueryRunnerTest and be more specific about legacy API removal plans in Druid 0.11 in Javadocs
* Ignore misnamed segment cache info files.
Fixes a bug where historical nodes could announce the same segment twice,
ultimately leading to historicals and their watchers (like coordinators
and brokers) being out of sync about which segments are served.
This could be caused if Druid is switched from local time to UTC, because
that causes the same segment descriptors to lead to different identifiers
(an identifier with local time interval before the switch, and UTC interval
after the switch). In turn this causes that segment descriptor to be written
to multiple segment cache info files and potentially get announced twice.
Later, if the historical receives a drop request, it drops the segment and
unannounces it once, but the other announcement would stick around in an
ephemeral znode forever, confusing coordinators and brokers.
* Only alert once.
* coordinator lookups mgmt improvements
* revert replaces removal, deprecate it instead
* convert and use older specs stored in db
* more tests and updates
* review comments
* add behavior for 0.10.0 to 0.9.2 downgrade
* incorporating more review comments
* remove explicit lock and use LifecycleLock in LookupReferencesManager. use LifecycleLock in LookupCoordinatorManager as well
* wip on LookupCoordinatorManager
* lifecycle lock
* refactor thread creation into utility method
* more review comments addressed
* support smooth roll back of lookup snapshots from 0.10.0 to 0.9.2
* correctly use LifecycleLock in LookupCoordinatorManager and remove synchronization from start/stop
* run lookup mgmt on leader coordinator only
* wip: changes to do multiple start() and stop() on LookupCoordinatorManager
* lifecycleLock fix usage in LookupReferencesManagerTest
* add LifecycleLock back
* fix license hdr
* some fixes
* make LookupReferencesManager.getAllLookupsState() consistent while still being lockless
* address review comments
* addressing leventov's comments
* address charle's comments
* add IOE.java
* for safety in LookupReferencesManager mainThread check for lifecycle started state on each loop in addition to interrupt
* move thread creation utility method to Execs
* fix names
* add tests for LookupCoordinatorManager.lookupManagementLoop()
* add further tests for figuring out toBeLoaded and toBeDropped on LookupCoordinatorManager
* address leventov comments
* remove LookupsStateWithMap and parameterize LookupsState
* address review comments
* address more review comments
* misc fixes
* DirectDruidClient: Fix division by zero.
Introduced in #3954 when some floating math was changed to
integer math. This patch restores the old math.
* Added comment.
* Make timeout behavior consistent to document
* Refactoring BlockingPool and add more methods to QueryContexts
* remove unused imports
* Addressed comments
* Address comments
* remove unused method
* Make default query timeout configurable
* Fix test failure
* Change timeout from period to millis
* RealtimeIndexTask to support alertTimeout in context and raise alert if task process exists after the timeout
* move alertTimeout config to tuningConfig and document
It wasn't doing anything useful (the sequences were being concatted, and
cursor.getTime() wasn't being called) and it defaulted to Granularities.NONE.
Changing it to Granularities.ALL gave me a 700x+ performance boost on a
small dataset I was reindexing (2m27s to 365ms). Most of that was from avoiding
making a lot of unnecessary column selectors.
* Allow compilation as Java8 source and target for everything except API
* Remove conditions in tests which assume that we may run with Java 7
* Update easymock to 3.4
* Make Animal Sniffer to check Java 1.8 usage; remove redundant druid-caffeine-cache configuration
* Use try-with-resources in LargeColumnSupportedComplexColumnSerializerTest.testSanity()
* Remove java7 special for druid-api
* In IndexMerger and IndexMergerV9, create temporary files under the output directory/tmpPeonFiles, instead of java.io.tmpdir
* Use FileUtils.forceMkdir() across the codebase and remove some unused code
* Fix test
* Fix PullDependencies.run()
* Unused import
* No more singleton. Reduce iterations
* Granularities
* Fix the delay in the test
* Add license header
* Remove unused imports
* Lot more unused imports from all the rearranging
* CR feedback
* Move javadoc to constructor
* Refactor Segment Granularity
* Beginning of one granularity
* Copy the fix for custom periods in segment-grunalrity over here.
* Remove the custom serialization for now.
* Compilation cleanup
* Reformat code
* Fixing unit tests
* Unify to use a single iterable
* Backward compatibility for rolling upgrade
* Minor check style. Cosmetic changes.
* Rename length and millis to duration
* CR feedback
* Minor changes.
GroupBy v2 doesn't cache on the broker, so it isn't actually testing
what the test was supposed to be testing. Also, the test failed due
to mismatched expectations.
* fix cache populate incorrect content when numBackgroundThreads>1
* simplify code by using Futures.allAsList and use CountDownLatch in UT
* fix test code style and assert countDownLatch.await()
* Require Java 8 and include some Java 8 dependencies.
- Upgrade Jetty to 9.3.16.v20170120.
- Upgrade DataSketches to 0.8.4.
- Bundle caffeine-cache by default.
- Still target Java 7 when compiling base Druid classes.
* Update cluster, quickstart docs.
* Remove oraclejdk7 from travis.yml.
* Less use of File.deleteOnExit()
* removed deleteOnExit from most of the tests/benchmarks/iopeon
* Made IOpeon closable
* Formatting.
* Revert DeterminePartitionsJobTest, remove cleanup method from IOPeon
* modified "end" column to `end`. "end" is interpretted as a string rather than dereferencing the column value
* SQLMetadataConnector.getQuoteString defines the string that should be used to quote string fields
* positional arguments for String.format
* for Connectors that use " need to include the \ escape as well
* Add PostAggregators to generator cache keys for top-n queries
* Add tests for strings
* Remove debug comments
* Add type keys and list sizes to cache key
* Make post aggregators used for sort are considered for cache key generation
* Use assertArrayEquals()
* Improve findPostAggregatorsForSort()
* Address comments
* fix test failure
* address comments
* Specialize LoadBalancingPool as MemcacheClientPool, reduce locking and don't override Object.finalize()
* Remove locking and don't override Object.finalize() in ReferenceCountingResourceHolder
* Add leak counts in ReferenceCountingResourceHolder and MemcacheClientPool. Add tests for ReferenceCountingResourceHolder and MemcacheClientPool
* Fix a race condition in ReferenceCountingResourceHolder.increment()
* Introduce SegmentizerFactory
- that knows how to deserialize specific type of segment
- Default implementation is MMappedQueryableSegmentizerFactory which creates QueryableIndexSegment
- Unit test for the default behavior
* review comments
* auto reset option for Kafka Indexing service in case message at the offset being fetched is not present anymore at kafka brokers
* review comments
* review comments
* reverted last change
* review comments
* review comments
* fix typo
* Removing unused code from io.druid.java.util.common.guava package; fix#3563 (more consistent and paranoiac resource handing in Sequences subsystem); Add Sequences.wrap() for DRY in MetricsEmittingQueryRunner, CPUTimeMetricQueryRunner and SpecificSegmentQueryRunner; Catch MissingSegmentsException in SpecificSegmentQueryRunner's yielder.next() method (follow up on #3617)
* Make Sequences.withEffect() execute the effect if the wrapped sequence throws exception from close()
* Fix strange code in MetricsEmittingQueryRunner
* Add comment on why YieldingSequenceBase is used in Sequences.withEffect()
* Use Closer in OrderedMergeSequence and MergeSequence to close multiple yielders
* SQL: Ditch CalciteConnection layer and add DruidMeta, extension aggregators.
Switched from CalciteConnection to Planner, bringing benefits:
- CalciteConnection's JDBC interface no longer sits between the SQL server
(HTTP/Avatica) and Druid's query layer. Instead, the SQL servers can use
Druid Sequence objects directly, reducing overhead in the query return path.
- Implemented our own Planner-based Avatica Meta, letting us control
connection timeouts and connection / statement limits. The previous
CalciteConnection-based implementation didn't have any limits or timeouts.
- The Planner interface lets us override the operator table, opening up
SQL language extensions. This patch includes two: APPROX_COUNT_DISTINCT
in core, and a QUANTILE aggregator in the druid-histogram extension.
Also:
- Added INFORMATION_SCHEMA metadata schema.
- Added tests for Unicode literals and escapes.
* Verify statement is actually open before closing it.
* More detailed INFORMATION_SCHEMA docs.
* streaming version of select query
* use columns instead of dimensions and metrics;prepare for valueVector;remove granularity
* respect query limit within historical
* use constant
* fix thread name corrupted bug when using jetty qtp thread rather than processing thread while working with SpecificSegmentQueryRunner
* add some test for scan query
* add scan query document
* fix merge conflicts
* add compactedList resultFormat, this format is better for json ser/der
* respect query timeout
* respect query limit on broker
* use static consts and remove unused code
* allow JsonConfigTesterBase to treat the fields of collections
* [Feature] Exhibitor Support (#3664)
This patch provides the integration of Druid & Netflix Exhibitor. Druid
currently use Apache Curator as ZooKeeper client. Curator can be
integrated with Exhibitor to achieve a live/updating list of the
ZooKeeper ensemble. This patch enables Druid to use this features.
* Fix#3795 (Java 7 compatibility).
Also introduce Animal Sniffer checks during build, which would
have caught the original problems.
* Add Animal Sniffer on caffeine-cache for JDK8.
* Add metrics for Query Count statistics
This PR adds a new metrics monitor “QueryCountStatsMonitor” which emits
three new metrics -
1) query/success/count - number of successful queries
2) query/failed/count - number of failed queries
3) query/interrupted/count - number of interrupted/timedout queries
fix bindings
* make fields final
* fix imports
* AsyncQueryForwardingServlet implement QueryStatsProvider
* remove unused import
* Enable parallel test
* Remove unnecessary NotThreadSafe annocation
* Randomize the start port when finding available ports
* Fix test failure
* Change to handle all negatives
* report message gap, source gap and sink count in RealtimePlumber
* report message gap, sink count in Appenderator
* add ingest/events/sourceGap in metrics.md
* remove source gap
* Fix PathChildrenCache's executorService leak in Announcer, CuratorInventoryManager and RemoteTaskRunner
* Use a single ExecutorService for all workerStatusPathChildrenCaches in RemoteTaskRunner
* make sure CliCoordinator initializes and starts DerbyMetadataStorage first if configured
* Revert "make sure CliCoordinator initializes and starts DerbyMetadataStorage first if configured"
This reverts commit 54f5644054626d4a9e2448bb4bd5e6ce9a9fca1d.
* make sure CliCoordinator initializes and starts DerbyMetadataStorage first if configured
* Normalized Cost Balancer
* Adding documentation and renaming to use diskNormalizedCostBalancer
* Remove balancer from the strings
* Update docs and include random cost balancer
* Fix checkstyle issues
This retries when the start condition is met but SELECT -> INSERT/UPDATE
fails, which indicates a race. If the start condition isn't met, there
won't be any retrying.
* Update emitter dependency to 0.4.0 and emit "version" dimension for all druid metrics, not only query metrics
* Remove unused imports
* Use empty string instead of "testing-version" as a version placeholder
Excludes tests from AvoidStaticImport, since those are used often there and
I didn't want to make this changeset too large. Production code use was minimal
and I switched those to non-static imports.
This is useful for the insert-segment-to-db tool, which would otherwise
potentially insert a lot of overshadowed segments as "used", causing
load and drop churn in the cluster.
* print exception details from QueryInterruptedException
* in QueryResource.java, set thread name to include dataSource names and not whole query string e.g. from QueryDataSource
* Add support for timezone in segment granularity
* CR feedback. Handle null timezone during equals check.
* Include timezone in docs.
Add timezone for ArbitraryGranularitySpec.
* Show candidate hosts for the given query
* Added test cases & minor changes to address comments
* Changed path-param to query-pram for intervals/numCandidates
* support renaming of outputName for cached select and search queries
* rebase and resolve conflicts
* rollback CacheStrategy interface change
* updated based on review comments
* introducing lists of existing columns in the fields of select queries' output
* rebase master
* address the comment. add test code for select query caching
* change the cache code in SelectQueryQueryToolChest to 0x16
Follow-up to #1773, which meant to add more useful query errors but
did not actually do so. Since that patch, any error other than
interrupt/cancel/timeout was reported as `{"error":"Unknown exception"}`.
With this patch, the error fields are:
- error, one of the specific strings "Query interrupted", "Query timeout",
"Query cancelled", or "Unknown exception" (same behavior as before).
- errorMessage, the message of the topmost non-QueryInterruptedException
in the causality chain.
- errorClass, the class of the topmost non-QueryInterruptedException
in the causality chain.
- host, the host that failed the query.
* ability to not rollup at index time, make pre aggregation an option
* rename getRowIndexForRollup to getPriorIndex
* fix doc misspelling
* test query using no-rollup indexes
* fix benchmark fail due to jmh bug
* Add numeric StringComparator
* Only use direct long comparison for numeric ordering in BoundFilter, add time filtering benchmark query
* Address PR comments, add multithreaded BoundDimFilter test
* Add comment on strlen tie handling
* Add timeseries interval filter benchmark
* Adjust docs
* Use jackson for StringComparator, address PR comments
* Add new TopNMetricSpec and SearchSortSpec with tests (WIP)
* More TopNMetricSpec and SearchSortSpec tests
* Fix NewSearchSortSpec serde
* Update docs for new DimensionTopNMetricSpec
* Delete NumericDimensionTopNMetricSpec
* Delete old SearchSortSpec
* Rename NewSearchSortSpec to SearchSortSpec
* Add TopN numeric comparator benchmark, address PR comments
* Refactor OrderByColumnSpec
* Add null checks to NumericComparator and String->BigDecimal conversion function
* Add more OrderByColumnSpec serde tests
This reverts commit a931debf79.
Fixes#3283
The core issue here is that realtime nodes announce their size as 0, so a coordinator which interns the realtime version of the data segment will not be able to see the new sized announcement when handoff occurs.
This is caused by the `eauals` method on a `DataSegment` only evaluating the identifier. the `eauals` method *should* be correct for object equivalence, and things which need to check equivalence of some sub-portion of the object should do so explicitly.
Fixes inconsistent metric handling between the two implementations. Formerly,
RealtimePlumber only emitted query/segmentAndCache/time and query/wait and
Appenderator only emitted query/partial/time and query/wait (all per sink).
Now they both do the same thing:
- query/segmentAndCache/time, query/segment/time are the time spent per sink.
- query/cpu/time is the CPU time spent per query.
- query/wait/time is the executor waiting time per sink.
These generally match historical metrics, except segmentAndCache & segment
mean the same thing here, because one Sink may be partially cached and
partially uncached and we aren't splitting that out.
All query metrics now start with toolChest.makeMetricBuilder, and all of
*those* now start with DruidMetrics.makePartialQueryTimeMetric. Also, "id"
moved to common code, since all query metrics added it anyway.
In particular this will add query-type specific dimensions like "threshold"
and "numDimensions" to servlet-originated metrics like query/time.
* Migrate IndexerSQLMetadataStorageCoordinator.getUnusedSegmentsForInterval to streaming
* Missed query from #2859
* Make inReadOnlyTransaction part of SQLMetadataConnector
This patch introduces a GroupByStrategy concept and two strategies: "v1"
is the current groupBy strategy and "v2" is a new one. It also introduces
a merge buffers concept in DruidProcessingModule, to try to better
manage memory used for merging.
Both of these are described in more detail in #2987.
There are two goals of this patch:
1. Make it possible for historical/realtime nodes to return larger groupBy
result sets, faster, with better memory management.
2. Make it possible for brokers to merge streams when there are no order-by
columns, avoiding materialization.
This patch does not do anything to help with memory management on the broker
when there are order-by columns or when there are nested queries. That could
potentially be done in a future patch.
* add get dimension rangeset to filters
* add get domain to ShardSpec and added chunk filter in caching clustered client
* add null check and modified not filter, started with unit test
* add filter test with caching
* refactor and some comments
* extract filtershard to helper function
* fixup
* minor changes
* update javadoc
* fix caching for search results
properly read count when reading from cache.
* fix NPE during merging search count and add test
* Update cache key to invalidate prev results
* validate X-Druid-Task-Id header in request and add header to response
* modify KafkaIndexTaskClient to take a TaskLocationProvider as the TaskLocation may not remain constant
* Allow dynamically setting of shutoffTime for EventReceiverFirehose
Allow dynamically setting shutoffTime for EventReceiverFirehose
review comments and tests
* shut down exec on close
* Datasource as lookup tier
* Adds an option to let indexing service tasks pull their lookup tier from the datasource they are working for.
* Fix bad docs for lookups lookupTier
* Add Datasource name holder
* Move task and datasource to be pulled from Task file
* Make LookupModule pull from bound dataSource
* Fix test
* Fix code style on imports
* Fix formatting
* Make naming better
* Address code comments about naming
* new interval based cost function
Addresses issues with balancing of segments in the existing cost function
- `gapPenalty` led to clusters of segments ~30 days apart
- `recencyPenalty` caused imbalance among recent segments
- size-based cost could be skewed by compression
New cost function is purely based on segment intervals:
- assumes each time-slice of a partition is a constant cost
- cost is additive, i.e. cost(A, B union C) = cost(A, B) + cost(A, C)
- cost decays exponentially based on distance between time-slices
* comments and formatting
* add more comments to explain the calculation
* Allow user to set cost balancer threads more than the number of cores.
Allow user to set cost balancer threads more than the number of cores.
* modify test
* Optimize CostBalancerStrategy
Ignore benchmark test in normal run
fix test
review comments
fix compilation
fix test
* review comments
* review comment
Roman Leventov
kaijianding
Charles Allen
Slim
Jihoon Son
Kenji Noguchi
dgolitsyn
Goh Wei Xiang
Jonathan Wei
Gian Merlino
Himanshu
Benedict Jin
satishbhor
David Lim
Parag Jain
Roman Leventov
praveev
Niketh Sabbineni
Jakub Kukul
Akash Dwivedi
michaelschiff
DaimonPl
Yuusaku Taniguchi
Nishant
John Zhang
Navis Ryu
Xiaoyao
Keuntae Park
Gleb Smirnov
jaehong choi
rajk-tetration
Hyukjin Kwon
Xavier Léauté
Dave Li
John Wang