* Reduce # of max subTasks to 2
* fix typo and add more doc
* add more doc and link
* change default and add warning
* fix doc
* add test
* fix it test
#### `EventReceiverFirehoseFactory`
Fixed several concurrency bugs in `EventReceiverFirehoseFactory`:
- Race condition over putting an entry into `producerSequences` in `checkProducerSequence()`.
- `Stopwatch` used to measure time across threads, but it's a non-thread-safe class.
- Use `System.nanoTime()` instead of `System.currentTimeMillis()` because the latter are [not suitable](https://stackoverflow.com/a/351571/648955) for measuring time intervals.
- `close()` was not synchronized by could be called from multiple threads concurrently.
Removed unnecessary `readLock` (protecting `hasMore()` and `nextRow()` which are always called from a single thread). Removed unnecessary `volatile` modifiers.
Documented threading model and concurrent control flow of `EventReceiverFirehose` instances.
**Important:** please read the updated Javadoc for `EventReceiverFirehose.addAll()`. It allows events from different requests (batches) to be interleaved in the buffer. Is this OK?
#### `TimedShutoffFirehoseFactory`
- Fixed a race condition that was possible because `close()` that was not properly synchronized.
Documented threading model and concurrent control flow of `TimedShutoffFirehose` instances.
#### `Firehose`
Refined concurrency contract of `Firehose` based on `EventReceiverFirehose` implementation. Importantly, now it states that `close()` doesn't affect `hasMore()` and `nextRow()` and could be called concurrently with them. In other words, specified that `close()` is for "row supply" side rather than "row consume" side. However, I didn't check that other `Firehose` implementatations adhere to this contract.
<hr>
This issue is the result of reviewing `EventReceiverFirehose` and `TimedShutoffFirehose` using [this checklist](https://medium.com/@leventov/code-review-checklist-java-concurrency-49398c326154).
* Remove DataSegmentFinder, InsertSegmentToDb, and descriptor.json file
* delete descriptor.file when killing segments
* fix test
* Add doc for ha
* improve warning
* index_parallel: support !appendToExisting with no explicit intervals
This enables ParallelIndexSupervisorTask to dynamically request locks at runtime
if it is run without explicit intervals in the granularity spec and with
appendToExisting set to false. Previously, it behaved as if appendToExisting
was set to true, which was undocumented and inconsistent with IndexTask and
Hadoop indexing.
Also, when ParallelIndexSupervisorTask allocates segments in the explicit
interval case, fail if its locks on the interval have been revoked.
Also make a few other additions/clarifications to native ingestion docs.
Fixes#6989.
* Review feedback.
PR description on GitHub updated to match.
* Make native batch ingestion partitions start at 0
* Fix to previous commit
* Unit test. Verified to fail without the other commits on this branch.
* Another round of review
* Slightly scarier warning
* document middle manager api
* re-arrange
* correction
* document more missing overlord api calls, minor re-arrange of some code i was referencing
* fix it
* this will fix it
* fixup
* link to other docs
* Support kafka transactional topics
* update kafka to version 2.0.0
* Remove the skipOffsetGaps option since it's not used anymore
* Adjust kafka consumer to use transactional semantics
* Update tests
* Remove unused import from test
* Fix compilation
* Invoke transaction api to fix a unit test
* temporary modification of travis.yml for debugging
* another attempt to get travis tasklogs
* update kafka to 2.0.1 at all places
* Remove druid-kafka-eight dependency from integration-tests, remove the kafka firehose test and deprecate kafka-eight classes
* Add deprecated in docs for kafka-eight and kafka-simple extensions
* Remove skipOffsetGaps and code changes for transaction support
* Fix indentation
* remove skipOffsetGaps from kinesis
* Add transaction api to KafkaRecordSupplierTest
* Fix indent
* Fix test
* update kafka version to 2.1.0
* maintenance mode for Historical
forbidden api fix, config deserialization fix
logging fix, unit tests
* addressed comments
* addressed comments
* a style fix
* addressed comments
* a unit-test fix due to recent code-refactoring
* docs & refactoring
* addressed comments
* addressed a LoadRule drop flaw
* post merge cleaning up
* Add published segment cache in broker
* Change the DataSegment interner so it's not based on DataSEgment's equals only and size is preserved if set
* Added a trueEquals to DataSegment class
* Use separate interner for realtime and historical segments
* Remove trueEquals as it's not used anymore, change log message
* PR comments
* PR comments
* Fix tests
* PR comments
* Few more modification to
* change the coordinator api
* removeall segments at once from MetadataSegmentView in order to serve a more consistent view of published segments
* Change the poll behaviour to avoid multiple poll execution at same time
* minor changes
* PR comments
* PR comments
* Make the segment cache in broker off by default
* Added a config to PlannerConfig
* Moved MetadataSegmentView to sql module
* Add doc for new planner config
* Update documentation
* PR comments
* some more changes
* PR comments
* fix test
* remove unintentional change, whether to synchronize on lifecycleLock is still in discussion in PR
* minor changes
* some changes to initialization
* use pollPeriodInMS
* Add boolean cachePopulated to check if first poll succeeds
* Remove poll from start()
* take the log message out of condition in stop()
* blooming aggs
* partially address review
* fix docs
* minor test refactor after rebase
* use copied bloomkfilter
* add ByteBuffer methods to BloomKFilter to allow agg to use in place, simplify some things, more tests
* add methods to BloomKFilter to get number of set bits, use in comparator, fixes
* more docs
* fix
* fix style
* simplify bloomfilter bytebuffer merge, change methods to allow passing buffer offsets
* oof, more fixes
* more sane docs example
* fix it
* do the right thing in the right place
* formatting
* fix
* avoid conflict
* typo fixes, faster comparator, docs for comparator behavior
* unused imports
* use buffer comparator instead of deserializing
* striped readwrite lock for buffer agg, null handling comparator, other review changes
* style fixes
* style
* remove sync for now
* oops
* consistency
* inspect runtime shape of selector instead of selector plus, static comparator, add inner exception on serde exception
* CardinalityBufferAggregator inspect selectors instead of selectorPluses
* fix style
* refactor away from using ColumnSelectorPlus and ColumnSelectorStrategyFactory to instead use specialized aggregators for each supported column type, other review comments
* adjustment
* fix teamcity error?
* rename nil aggs to empty, change empty agg constructor signature, add comments
* use stringutils base64 stuff to be chill with master
* add aggregate combiner, comment
* add "bloom_filter_test" druid expression to support bloom filters in ExpressionVirtualColumn and ExpressionDimFilter and sql expressions
* more docs
* use java.util.Base64, doc fixes
* use SqlLifecyle to manage sql execution, add sqlId
* add sql request logger
* fix UT
* rename sqlId to sqlQueryId, sql/time to sqlQuery/time, etc
* add docs and more sql request logger impls
* add UT for http and jdbc
* fix forbidden use of com.google.common.base.Charsets
* fix UT in QuantileSqlAggregatorTest, supressed unused warning of getSqlQueryId
* do not use default method in QueryMetrics interface
* capitalize 'sql' everywhere in the non-property parts of the docs
* use RequestLogger interface to log sql query
* minor bugfixes and add switching request logger
* add filePattern configs for FileRequestLogger
* address review comments, adjust sql request log format
* fix inspection error
* try SuppressWarnings("RedundantThrows") to fix inspection error on ComposingRequestLoggerProvider
PR #6605 added support to the statsd emitter for DogStatsD tags. This commit
lets you specify "constant tags" in the config file which are included with
every event. This is helpful if you are running in an environment where you
cannot configure your datadog-agent with tags like "cluster name" --- eg, a
Kubernetes cluster with a datadog-agent on each node and different Druid
deployments in different namespaces but sharing the same datadog-agent
daemonset.
Also fix the name of an existing boolean getter to start with 'is'.
* created seekablestream classes
* created seekablestreamsupervisor class
* first attempt to integrate kafa indexing service to use SeekableStream
* seekablestream bug fixes
* kafkarecordsupplier
* integrated kafka indexing service with seekablestream
* implemented resume/suspend and refactored some package names
* moved kinesis indexing service into core druid extensions
* merged some changes from kafka supervisor race condition
* integrated kinesis-indexing-service with seekablestream
* unite tests for kinesis-indexing-service
* various bug fixes for kinesis-indexing-service
* refactored kinesisindexingtask
* finished up more kinesis unit tests
* more bug fixes for kinesis-indexing-service
* finsihed refactoring kinesis unit tests
* removed KinesisParititons and KafkaPartitions to use SeekableStreamPartitions
* kinesis-indexing-service code cleanup and docs
* merge #6291
merge #6337
merge #6383
* added more docs and reordered methods
* fixd kinesis tests after merging master and added docs in seekablestream
* fix various things from pr comment
* improve recordsupplier and add unit tests
* migrated to aws-java-sdk-kinesis
* merge changes from master
* fix pom files and forbiddenapi checks
* checkpoint JavaType bug fix
* fix pom and stuff
* disable checkpointing in kinesis
* fix kinesis sequence number null in closed shard
* merge changes from master
* fixes for kinesis tasks
* capitalized <partitionType, sequenceType>
* removed abstract class loggers
* conform to guava api restrictions
* add docker for travis other modules test
* address comments
* improve RecordSupplier to supply records in batch
* fix strict compile issue
* add test scope for localstack dependency
* kinesis indexing task refactoring
* comments
* github comments
* minor fix
* removed unneeded readme
* fix deserialization bug
* fix various bugs
* KinesisRecordSupplier unable to catch up to earliest position in stream bug fix
* minor changes to kinesis
* implement deaggregate for kinesis
* Merge remote-tracking branch 'upstream/master' into seekablestream
* fix kinesis offset discrepancy with kafka
* kinesis record supplier disable getPosition
* pr comments
* mock for kinesis tests and remove docker dependency for unit tests
* PR comments
* avg lag in kafkasupervisor #6587
* refacotred SequenceMetadata in taskRunners
* small fix
* more small fix
* recordsupplier resource leak
* revert .travis.yml formatting
* fix style
* kinesis docs
* doc part2
* more docs
* comments
* comments*2
* revert string replace changes
* comments
* teamcity
* comments part 1
* comments part 2
* comments part 3
* merge #6754
* fix injection binding
* comments
* KinesisRegion refactor
* comments part idk lol
* can't think of a commit msg anymore
* remove possiblyResetDataSourceMetadata() for IncrementalPublishingTaskRunner
* commmmmmmmmmments
* extra error handling in KinesisRecordSupplier getRecords
* comments
* quickfix
* typo
* oof
* Broker: Await initialization before finishing startup.
In particular, hold off on announcing the service and starting the
HTTP server until the server view and SQL metadata cache are finished
initializing. This closes a window of time where a Broker could return
partial results shortly after startup.
As part of this, some simplification of server-lifecycle service
announcements. This helps ensure that the two different kinds of
announcements we do (legacy and new-style) stay in sync.
* Remove unused imports.
* Fix NPE in ServerRunnable.
* FilteredRequestLogger: Fix start/stop, invalid delegate behavior.
Fixes two bugs:
1) FilteredRequestLogger did not start/stop the delegate.
2) FilteredRequestLogger would ignore an invalid delegate type, and
instead silently substitute the "noop" logger. This was due to a larger
problem with RequestLoggerProvider setup in general; the fix here is
to remove "defaultImpl" from the RequestLoggerProvider interface, and
instead have JsonConfigurator be responsible for creating the
default implementations. It is stricter about things than the old system
was, and is only willing to make a noop logger if it doesn't see any
request logger configs. Otherwise, it'll raise a provision error.
* Remove unneeded annotations.
The documentation for Bound filter's lowerStrict/upperStrict is incorrect. It is not consistent with the examples provided and actual behaviour of the bound filter. Correct this.
* autosize processing buffers based on direct memory sizing
* remove oops, more test
* max 1gb autosize buffers, test, start of docs
* fix oops
* revert accidental change
* print buffer size in exception
* change the things
* Replace StatsD client library
The [Datadog package][1] is a StatsD compatible drop-in replacement for the
client library, but it seems to be [better maintained][2] and has support for
Datadog DogStatsD specific features, which will be made use of in a subsequent
commit.
The `count`, `time`, and `gauge` methods are actually exactly compatible with
the previous library and the modifications shouldn't be required, but EasyMock
seems to have a hard time dealing with the variable arguments added by the
DogStatsD library and causes tests to fail if no arguments are provided for the
last String vararg. Passing an empty array fixes the test failures.
[1]: https://github.com/DataDog/java-dogstatsd-client
[2]: https://github.com/tim-group/java-statsd-client/issues/37#issuecomment-248698856
* Retain dimension key information for StatsD metrics
This doesn't change behavior, but allows separating dimensions from the metric
name in subsequent commits.
There is a possible order change for values from
`dimsBuilder.build().values()`, but from the tests it looks like it doesn't
affect actual behavior and the order of user dimensions is also retained.
* Support DogStatsD style tags in statsd-emitter
Datadog [doesn't support name-encoded dimensions and uses a concept of _tags_
instead.][1] This change allows Datadog users to send the metrics without
having to encode the various dimensions in the metric names. This enables
building graphs and monitors with and without aggregation across various
dimensions from the same data.
As tests in this commit verify, the behavior remains the same for users who
don't enable the `druid.emitter.statsd.dogstatsd` configuration flag.
[1]: https://www.datadoghq.com/blog/the-power-of-tagged-metrics/#tags-decouple-collection-and-reporting
* Disable convertRange behavior for DogStatsD users
DogStatsD, unlike regular StatsD, supports floating-point values, so this
behavior is unnecessary. It would be possible to still support `convertRange`,
even with `dogstatsd` enabled, but that would mean that people using the
default mapping would have some of the gauges unnecessarily converted.
`time` is in milliseconds and doesn't support floating-point values.
Also add a "fromIndex" argument to the strpos expression function. There
are some -1 and +1 adjustment terms due to the fact that the strpos
expression behaves like Java indexOf (0-indexed), but the POSITION SQL
function is 1-indexed.
* move parquet-extensions from contrib to core, adds new hadoop parquet parser that does not convert to avro first and supports flattenSpec and int96 columns, add support for flattenSpec for parquet-avro conversion parser, much test with a bunch of files lifted from spark-sql
* fix avro flattener to support nullable primitives for auto discovery and now only supports primitive arrays instead of all arrays
* remove leftover print
* convert micro timestamp to millis
* checkstyle
* add ignore for .parquet and .parq to rat exclude
* fix legit test failure from avro flattern behavior change
* fix rebase
* add exclusions to pom to cut down on redundant jars
* refactor tests, add support for unwrapping lists for parquet-avro, review comments
* more comment
* fix oops
* tweak parquet-avro list handling
* more docs
* fix style
* grr styles
* Period load/drop/broadcast rules should include the future by default
* address comments
* adjust coordinator console and tweak docs
* address comments
* fix travis-ci
* include mysql-metadata-storage extension in distribution, but without the GPL-licensed connector library
* Install mysql connector package
* use symlinks to avoid versioning issues
* add documentation for fetching the mysql connector
* add PrefixFilteredDimensionSpec for multi-value dimensions
* add docs for PrefixFilteredDimensionSpec
* remove unnecessary null handling
* add null check to the result of NullHandling
* o- Query Response format to be based on http 'accept' header & Query Payload contenty type to be based on 'content-type' header
* o- Query Response format to be based on http 'accept' header & Query Payload contenty type to be based on 'content-type' header
o- if Accept header is absent, it defaults to Content-Type header
* Feature: Query Response format to be based on http 'accept' header & Query Payload content type to be based on 'content-type' PR #4033
Minor change to a comment - restoring to previous wording
* Feature: Query Response format to be based on http 'accept' header & Query Payload content type to be based on 'content-type' PR #4033
o- minor change to check for empty string
* Added SystemSchema with following tables (#5989)
* SEGMENTS table provides details on served and published segments
* SERVERS table provides details on data servers
* SERVERSEGMETS table is the JOIN of SEGMENTS and SERVERS
* TASKS table provides details on tasks
* Add documentation for system schema
* Fix static-analysis warnings
* Address PR comments
*Add unit tests
* Fix a test
* Try to fix a test
* Fix a bug around replica count
* rename io.druid to org.apache.druid
* Major change is to make tasks and segment queries streaming
* Made tasks/segments stream to calcite instead of storing it in memory
* Add num_rows to segments table
* Refactor JsonParserIterator
* Replace with closeable iterator
* Fix docs, make num_rows column nullable, some unit test changes
* make num_rows column type long, allow it to be null
fix a compile error after merge, add TrafficCop param to InputStreamResponseHandler
* Filter null rows for segments table from Linq4j enumerable
* change num_replicas datatype to long in segments table
* Fix some tests and address comments
* Doc updates, other PR comments
* Update tests
* Address comments
* Add auth check
* Update docs
* Refactoring
* Fix teamcity warning, change the getQueryableServer in TimelineServerView
* Fix compilation after rebase
* Use the stream API from AuthorizationUtils
* Added LeaderClient interface and NoopDruidLeaderClient class
* Revert "Added LeaderClient interface and NoopDruidLeaderClient class"
This reverts commit 100fa46e39.
* Make the naming consistent to server_segments for the join table
* Add ForbiddenException on auth check failure
* Remove static block from SystemSchema
* Try to fix a test in CalciteQueryTest due to rename of server_segments
* Fix the json output format in the coordinator API
* Add auth check in the segments API
* Add null check to avoid NPE
* Use annonymous class object instead of mock for DruidLeaderClient in SqlBenchmark
* Fix test failures, type long/BIGINT can be nullable
* Revert long nullability to fix tests
* Fix style for tests
* PR comments
* Address PR comments
* Add the missing BytesAccumulatingResponseHandler class
* Use Sequences.withBaggage in DruidPlanner
* Fix docs, add comments
* Close the iterator if hasNext returns false
* fix opentsdb emitter always be running
* check if emitter started
* add more details about consumeDelay in doc
* fix possible thread unsafe
* fix fail sending tags whose value contain colon
* 'suspend' and 'resume' support for kafka indexing service
changes:
* introduces `SuspendableSupervisorSpec` interface to describe supervisors which support suspend/resume functionality controlled through the `SupervisorManager`, which will gracefully shutdown the supervisor and it's tasks, update it's `SupervisorSpec` with either a suspended or running state, and update with the toggled spec. Spec updates are provided by `SuspendableSupervisorSpec.createSuspendedSpec` and `SuspendableSupervisorSpec.createRunningSpec` respectively.
* `KafkaSupervisorSpec` extends `SuspendableSupervisorSpec` and now supports suspend/resume functionality. The difference in behavior between 'running' and 'suspended' state is whether the supervisor will attempt to ensure that indexing tasks are or are not running respectively. Behavior is identical otherwise.
* `SupervisorResource` now provides `/druid/indexer/v1/supervisor/{id}/suspend` and `/druid/indexer/v1/supervisor/{id}/resume` which are used to suspend/resume suspendable supervisors
* Deprecated `/druid/indexer/v1/supervisor/{id}/shutdown` and moved it's functionality to `/druid/indexer/v1/supervisor/{id}/terminate` since 'shutdown' is ambiguous verbage for something that effectively stops a supervisor forever
* Added ability to get all supervisor specs from `/druid/indexer/v1/supervisor` by supplying the 'full' query parameter `/druid/indexer/v1/supervisor?full` which will return a list of json objects of the form `{"id":<id>, "spec":<SupervisorSpec>}`
* Updated overlord console ui to enable suspend/resume, and changed 'shutdown' to 'terminate'
* move overlord console status to own column in supervisor table so does not look like garbage
* spacing
* padding
* other kind of spacing
* fix rebase fail
* fix more better
* all supervisors now suspendable, updated materialized view supervisor to support suspend, more tests
* fix log
* Broker backpressure.
Adds a new property "druid.broker.http.maxQueuedBytes" and a new context
parameter "maxQueuedBytes". Both represent a maximum number of bytes queued
per query before exerting backpressure on the channel to the data server.
Fixes#4933.
* Fix query context doc.
* resolves#5898 by adding maxTotalRows to incremental publishing kafka index task and appenderator based realtime indexing task, as available in IndexTask
* address review comments
* changes due to review
* merge fail
Jonathan Wei
jorbay-au
Clint Wylie
Jihoon Son
Xue Yu
Roman Leventov
Mingming Qiu
David Glasser
Surekha
scrawfor
awelsh93
Edward Gan
Egor Riashin
Justin Borromeo
lxqfy
Furkan KAMACI
Gian Merlino
Navin Kumar
Janek Lasocki-Biczysko
Dayue Gao
dongyifeng
Benjamin Hopp
thomask
Joshua Sun
David Lim
Vadim Ogievetsky
Vincent Newkirk
rcgarcia74
Evans Hauser
seoeun
Caroline1000
Deiwin Sarjas
Niketh Sabbineni
taiii
patelh
robertervin
vishnu rao
Atul Mohan
Shiv Toolsidass
adursun
Nishant Bangarwa
Benedict Jin
Alexander Saydakov
Eyal Yurman