* support unit suffix on byte-related properties
* add doc
* change default value of byte-related properites in example files
* fix coding style
* fix doc
* fix CI
* suppress spelling errors
* improve code according to comments
* rename Bytes to HumanReadableBytes
* add getBytesInInt to get value safely
* improve doc
* fix problem reported by CI
* fix problem reported by CI
* resolve code review comments
* improve error message
* improve code & doc according to comments
* fix CI problem
* improve doc
* suppress spelling check errors
* Add validation for authorizer name
* fix deps
* add javadocs
* Do not use resource filters
* Fix BasicAuthenticatorResource as well
* Add integration tests
* fix test
* fix
* Add availability and consistency docs.
Describes transactional ingestion and atomic replacement. Also, this patch
deletes some bad advice from the javadocs for SegmentTransactionalInsertAction.
* Fix missing word.
* QueryCountStatsMonitor can be injected in the Peon
This change fixes a dependency injection bug where there is a circular
dependency on getting the MonitorScheduler when a user configures the
QueryCountStatsMonitor to be used.
* fix tests
* Actually fix the tests this time
* Fix native batch range partition segment sizing
Fixes#10057.
Native batch range partitioning was only considering the partition
dimension value when grouping rows instead of using all of the row's
partition values. Thus, for schemas with multiple dimensions, the rollup
was overestimated, which would cause too many dimension values to be
packed into the same range partition. The resulting segments would then
be overly large (and not honor the target or max partition sizes).
Main changes:
- PartialDimensionDistributionTask: Consider all dimension values when
grouping row
- RangePartitionMultiPhaseParallelIndexingTest: Regression test by
having input with rows that should roll up and rows that should not
roll up
* Use hadoop & native hash ingestion row group key
* Fix missing temp dir for native single_dim
Native single dim indexing throws a file not found exception from
InputEntityIteratingReader.java:81. This MR creates the required
temporary directory when setting up the
PartialDimensionDistributionTask. The change was tested on a Druid
cluster. After installing the change native single_dim indexing
completes successfully.
* Fix indentation
* Use SinglePhaseSubTask as example for creating the temp dir
* Move temporary indexing dir creation in to TaskToolbox
* Remove unused dependency
Co-authored-by: Morri Feldman <morri@appsflyer.com>
* Fill in the core partition set size properly for batch ingestion with
dynamic partitioning
* incomplete javadoc
* Address comments
* fix tests
* fix json serde, add tests
* checkstyle
* Set core partition set size for hash-partitioned segments properly in
batch ingestion
* test for both parallel and single-threaded task
* unused variables
* fix test
* unused imports
* add hash/range buckets
* some test adjustment and missing json serde
* centralized partition id allocation in parallel and simple tasks
* remove string partition chunk
* revive string partition chunk
* fill numCorePartitions for hadoop
* clean up hash stuffs
* resolved todos
* javadocs
* Fix tests
* add more tests
* doc
* unused imports
* Allow append to existing datasources when dynamic partitioing is used
* fix test
* checkstyle
* checkstyle
* fix test
* fix test
* fix other tests..
* checkstyle
* hansle unknown core partitions size in overlord segment allocation
* fail to append when numCorePartitions is unknown
* log
* fix comment; rename to be more intuitive
* double append test
* cleanup complete(); add tests
* fix build
* add tests
* address comments
* checkstyle
* Fill in the core partition set size properly for batch ingestion with
dynamic partitioning
* incomplete javadoc
* Address comments
* fix tests
* fix json serde, add tests
* checkstyle
* Set core partition set size for hash-partitioned segments properly in
batch ingestion
* test for both parallel and single-threaded task
* unused variables
* fix test
* unused imports
* add hash/range buckets
* some test adjustment and missing json serde
* centralized partition id allocation in parallel and simple tasks
* remove string partition chunk
* revive string partition chunk
* fill numCorePartitions for hadoop
* clean up hash stuffs
* resolved todos
* javadocs
* Fix tests
* add more tests
* doc
* unused imports
* IntelliJ inspection and checkstyle rule for "Collection.EMPTY_* field accesses replaceable with Collections.empty*()"
* Reverted checkstyle rule
* Added tests to pass CI
* Codestyle
* move benchmark data generator into druid-processing, add a GeneratorInputSource to fill up a cluster with data
* newlines
* make test coverage not fail maybe
* remove useless test
* Update pom.xml
* Update GeneratorInputSourceTest.java
* less passive aggressive test names
* Empty partitionDimension has less rollup compared to the case when it is explicitly specified
* Adding a unit test for the empty partitionDimension scenario. Fixing another test which was failing
* Fixing CI Build Inspection Issue
* Addressing all review comments
* Updating the javadocs for the hash method in HashBasedNumberedShardSpec
This change removes ListenableFutures.transformAsync in favor of the
existing Guava Futures.transform implementation. Our own implementation
had a bug which did not fail the future if the applied function threw an
exception, resulting in the future never completing.
An attempt was made to fix this bug, however when running againts Guava's own
tests, our version failed another half dozen tests, so it was decided to not
continue down that path and scrap our own implementation.
Explanation for how was this bug manifested itself:
An exception thrown in BaseAppenderatorDriver.publishInBackground when
invoked via transformAsync in StreamAppenderatorDriver.publish will
cause the resulting future to never complete.
This explains why when encountering https://github.com/apache/druid/issues/9845
the task will never complete, forever waiting for the publishFuture to
register the handoff. As a result, the corresponding "Error while
publishing segments ..." message only gets logged once the index task
times out and is forcefully shutdown when the future is force-cancelled
by the executor.
* refactor SeekableStreamSupervisor usage of RecordSupplier to reduce contention between background threads and main thread, refactor KinesisRecordSupplier, refactor Kinesis lag metric collection and emitting
* fix style and test
* cleanup, refactor, javadocs, test
* fixes
* keep collecting current offsets and lag if unhealthy in background reporting thread
* review stuffs
* add comment
* add flag to flattenSpec to keep null columns
* remove changes to inputFormat interface
* add comment
* change comment message
* update web console e2e test
* move keepNullColmns to JSONParseSpec
* fix merge conflicts
* fix tests
* set keepNullColumns to false by default
* fix lgtm
* change Boolean to boolean, add keepNullColumns to hash, add tests for keepKeepNullColumns false + true with no nuulul columns
* Add equals verifier tests
* IntelliJ inspections cleanup
* Standard Charset object can be used
* Redundant Collection.addAll() call
* String literal concatenation missing whitespace
* Statement with empty body
* Redundant Collection operation
* StringBuilder can be replaced with String
* Type parameter hides visible type
* fix warnings in test code
* more test fixes
* remove string concatenation inspection error
* fix extra curly brace
* cleanup AzureTestUtils
* fix charsets for RangerAdminClient
* review comments
* WIP integration tests
* Add integration test for ingestion with transformSpec
* WIP almost working tests
* Add ignored tests
* checkstyle stuff
* remove newPage from index task ingestion spec
* more test cleanup
* still not quite working
* Actually disable the tests
* working tests
* fix codestyle
* dont use junit in integration tests
* actually fix the bug
* fix checkstyle
* bring index tests closer to reindex tests
* fix nullhandling exceptions related to test ordering
Tests might get executed in different order depending on the maven
version and the test environment. This may lead to "NullHandling module
not initialized" errors for some tests where we do not initialize
null-handling explicitly.
* use InitializedNullHandlingTest
* DruidSegmentReader should work if timestamp is specified as a dimension
* Add integration tests
Tests for compaction and re-indexing a datasource with the timestamp column
* Instructions to run integration tests against quickstart
* address pr
* add kinesis lag metric
* fixes
* heh
* do it right this time
* more test
* split out supervisor report lags into lagMillis, remove latest offsets from kinesis supervisor report since always null, review stuffs
* Use the iterator instead of higherKey(); use the iterator API instead of stream
* Fix tests; fix a concurrency bug in timeline
* fix test
* add tests for findNonOvershadowedObjectsInInterval
* fix test
* add missing tests; fix a bug in QueueEntry
* equals tests
* fix test
* Fix superbatch merge last partition boundaries
A bug in the computation for the last parallel merge partition could
cause an IndexOutOfBoundsException or precondition failure due to an
empty partition.
* Improve comments and tests
* Create splits of multiple files for parallel indexing
* fix wrong import and npe in test
* use the single file split in tests
* rename
* import order
* Remove specific local input source
* Update docs/ingestion/native-batch.md
Co-Authored-By: sthetland <steve.hetland@imply.io>
* Update docs/ingestion/native-batch.md
Co-Authored-By: sthetland <steve.hetland@imply.io>
* doc and error msg
* fix build
* fix a test and address comments
Co-authored-by: sthetland <steve.hetland@imply.io>
More functional tests to cover handling of input data that has a
partition dimension that contains:
1) Null values: Should be in first partition
2) Multi values: Should cause superbatch task to abort
* Codestyle - use java style array declaration
Replaced C-style array declarations with java style declarations and marked
the intelliJ inspection as an error
* cleanup test code
* Refactoring codes around ingestion:
- Parallel index task and simple task now use the same segment allocator implementation. This is reusable for the future implementation as well.
- Added PartitionAnalysis to store the analysis of the partitioning
- Move some util methods to SegmentLockHelper and rename it to TaskLockHelper
* fix build
* fix SingleDimensionShardSpecFactory
* optimize SingledimensionShardSpecFactory
* fix test
* shard spec builder
* import order
* shardSpecBuilder -> partialShardSpec
* build -> complete
* fix comment; add unit tests for partitionBoundaries
* add more tests and fix javadoc
* fix toString(); add serde tests for HashBasedNumberedPartialShardSpec and SegmentAllocateAction
* fix test
* add equality test for hash and range partial shard specs
* Forbid easily misused HashSet and HashMap constructors
* Add two LinkedHashMap constructors to forbidden-apis and create utility method as replacement for them
* Fix visibility of constant in CollectionUtils.java
* Make an exception for an instance of LinkedHashMap#<init>(int) because proper sizing is used
* revert changes to sql module tests that should be in separate PR
* Finish reverting changes to sql module tests that were flagged in checkstyle during CI
* Add netty dependency resulting from SupressForbidden
* intelliJ inspections cleanup
- remove redundant escapes
- performance warnings
- access static member via instance reference
- static method declared final
- inner class may be static
Most of these changes are aesthetic, however, they will allow inspections to
be enabled as part of CI checks going forward
The valuable changes in this delta are:
- using StringBuilder instead of string addition in a loop
indexing-hadoop/.../Utils.java
processing/.../ByteBufferMinMaxOffsetHeap.java
- Use class variables instead of static variables for parameterized test
processing/src/.../ScanQueryLimitRowIteratorTest.java
* Add intelliJ inspection warnings as errors to druid profile
* one more static inner class
* Reconcile terminology and method naming to 'used/unused segments'; Don't use terms 'enable/disable data source'; Rename MetadataSegmentManager to MetadataSegments; Make REST API methods which mark segments as used/unused to return server error instead of an empty response in case of error
* Fix brace
* Import order
* Rename withKillDataSourceWhitelist to withSpecificDataSourcesToKill
* Fix tests
* Fix tests by adding proper methods without interval parameters to IndexerMetadataStorageCoordinator instead of hacking with Intervals.ETERNITY
* More aligned names of DruidCoordinatorHelpers, rename several CoordinatorDynamicConfig parameters
* Rename ClientCompactTaskQuery to ClientCompactionTaskQuery for consistency with CompactionTask; ClientCompactQueryTuningConfig to ClientCompactionTaskQueryTuningConfig
* More variable and method renames
* Rename MetadataSegments to SegmentsMetadata
* Javadoc update
* Simplify SegmentsMetadata.getUnusedSegmentIntervals(), more javadocs
* Update Javadoc of VersionedIntervalTimeline.iterateAllObjects()
* Reorder imports
* Rename SegmentsMetadata.tryMark... methods to mark... and make them to return boolean and the numbers of segments changed and relay exceptions to callers
* Complete merge
* Add CollectionUtils.newTreeSet(); Refactor DruidCoordinatorRuntimeParams creation in tests
* Remove MetadataSegmentManager
* Rename millisLagSinceCoordinatorBecomesLeaderBeforeCanMarkAsUnusedOvershadowedSegments to leadingTimeMillisBeforeCanMarkAsUnusedOvershadowedSegments
* Fix tests, refactor DruidCluster creation in tests into DruidClusterBuilder
* Fix inspections
* Fix SQLMetadataSegmentManagerEmptyTest and rename it to SqlSegmentsMetadataEmptyTest
* Rename SegmentsAndMetadata to SegmentsAndCommitMetadata to reduce the similarity with SegmentsMetadata; Rename some methods
* Rename DruidCoordinatorHelper to CoordinatorDuty, refactor DruidCoordinator
* Unused import
* Optimize imports
* Rename IndexerSQLMetadataStorageCoordinator.getDataSourceMetadata() to retrieveDataSourceMetadata()
* Unused import
* Update terminology in datasource-view.tsx
* Fix label in datasource-view.spec.tsx.snap
* Fix lint errors in datasource-view.tsx
* Doc improvements
* Another attempt to please TSLint
* Another attempt to please TSLint
* Style fixes
* Fix IndexerSQLMetadataStorageCoordinator.createUsedSegmentsSqlQueryForIntervals() (wrong merge)
* Try to fix docs build issue
* Javadoc and spelling fixes
* Rename SegmentsMetadata to SegmentsMetadataManager, address other comments
* Address more comments
* Add JoinableFactory interface and use it in the query stack.
Also includes InlineJoinableFactory, which enables joining against
inline datasources. This is the first patch where a basic join query
actually works. It includes integration tests.
* Fix test issues.
* Adjustments from code review.
Builds on #9235, using the datasource analysis functionality to replace various ad-hoc
approaches. The most interesting changes are in ClientQuerySegmentWalker (brokers),
ServerManager (historicals), and SinkQuerySegmentWalker (indexing tasks).
Other changes related to improving how we analyze queries:
1) Changes TimelineServerView to return an Optional timeline, which I thought made
the analysis changes cleaner to implement.
2) Added QueryToolChest#canPerformSubquery, which is now used by query entry points to
determine whether it is safe to pass a subquery dataSource to the query toolchest.
Fixes an issue introduced in #5471 where subqueries under non-groupBy-typed queries
were silently ignored, since neither the query entry point nor the toolchest did
anything special with them.
3) Removes the QueryPlus.withQuerySegmentSpec method, which was mostly being used in
error-prone ways (ignoring any potential subqueries, and not verifying that the
underlying data source is actually a table). Replaces with a new function,
Queries.withSpecificSegments, that includes sanity checks.
* Add join-related DataSource types, and analysis functionality.
Builds on #9111 and implements the datasource analysis mentioned in #8728. Still can't
handle join datasources, but we're a step closer.
Join-related DataSource types:
1) Add "join", "lookup", and "inline" datasources.
2) Add "getChildren" and "withChildren" methods to DataSource, which will be used
in the future for query rewriting (e.g. inlining of subqueries).
DataSource analysis functionality:
1) Add DataSourceAnalysis class, which breaks down datasources into three components:
outer queries, a base datasource (left-most of the highest level left-leaning join
tree), and other joined-in leaf datasources (the right-hand branches of the
left-leaning join tree).
2) Add "isConcrete", "isGlobal", and "isCacheable" methods to DataSource in order to
support analysis.
Other notes:
1) Renamed DataSource#getNames to DataSource#getTableNames, which I think is clearer.
Also, made it a Set, so implementations don't need to worry about duplicates.
2) The addition of "isCacheable" should work around #8713, since UnionDataSource now
returns false for cacheability.
* Remove javadoc comment.
* Updates reflecting code review.
* Add comments.
* Add more comments.
Add more unit tests for range partition native batch parallel indexing.
Also, fix a bug where ParallelIndexPhaseRunner incorrectly thinks that
identical collected DimensionDistributionReports are not equal due to
not overriding equals() in DimensionDistributionReport.
* Doc update for new input source and input format.
- The input source and input format are promoted in all docs under docs/ingestion
- All input sources including core extension ones are located in docs/ingestion/native-batch.md
- All input formats and parsers including core extension ones are localted in docs/ingestion/data-formats.md
- New behavior of the parallel task with different partitionsSpecs are documented in docs/ingestion/native-batch.md
* parquet
* add warning for range partitioning with sequential mode
* hdfs + s3, gs
* add fs impl for gs
* address comments
* address comments
* gcs
* Optimize CachingLocalSegmentAllocator#getSequenceName
Replace StringUtils#format with string addition to generate the sequence
name for an interval and partition. This is faster because format uses a
Matcher under the covers to replace the string format with the variables.
* fix imports and add test
* Add comment about optimization
* Use renamed function for TaskToolbox
* Move tests after refactor
* Rename tests
* Fail superbatch range partition multi dim values
Change the behavior of parallel indexing range partitioning to fail
ingestion if any row had multiple values for the partition dimension.
After this change, the behavior matches that of hadoop indexing.
(Previously, rows with multiple dimension values would be skipped.)
* Improve err msg, rename method, rename test class
* HRTR: make pending task execution handling to go through all tasks on
not finding worker slots
* make HRTR methods package private that are meant to be used only in HttpRemoteTaskRunnerResource
* mark HttpRemoteTaskRunnerWorkItem.State global variables final
* hrtr: move immutableWorker NULL check outside of try-catch or finally block could have NPE
* add some explanatory comments
* add comment on explaining mechanics around hand off of pending tasks from submission to it getting picked up by a task execution thread
* fix spelling
* Parallel indexing single dim partitions
Implements single dimension range partitioning for native parallel batch
indexing as described in #8769. This initial version requires the
druid-datasketches extension to be loaded.
The algorithm has 5 phases that are orchestrated by the supervisor in
`ParallelIndexSupervisorTask#runRangePartitionMultiPhaseParallel()`.
These phases and the main classes involved are described below:
1) In parallel, determine the distribution of dimension values for each
input source split.
`PartialDimensionDistributionTask` uses `StringSketch` to generate
the approximate distribution of dimension values for each input
source split. If the rows are ungrouped,
`PartialDimensionDistributionTask.UngroupedRowDimensionValueFilter`
uses a Bloom filter to skip rows that would be grouped. The final
distribution is sent back to the supervisor via
`DimensionDistributionReport`.
2) The range partitions are determined.
In `ParallelIndexSupervisorTask#determineAllRangePartitions()`, the
supervisor uses `StringSketchMerger` to merge the individual
`StringSketch`es created in the preceding phase. The merged sketch is
then used to create the range partitions.
3) In parallel, generate partial range-partitioned segments.
`PartialRangeSegmentGenerateTask` uses the range partitions
determined in the preceding phase and
`RangePartitionCachingLocalSegmentAllocator` to generate
`SingleDimensionShardSpec`s. The partition information is sent back
to the supervisor via `GeneratedGenericPartitionsReport`.
4) The partial range segments are grouped.
In `ParallelIndexSupervisorTask#groupGenericPartitionLocationsPerPartition()`,
the supervisor creates the `PartialGenericSegmentMergeIOConfig`s
necessary for the next phase.
5) In parallel, merge partial range-partitioned segments.
`PartialGenericSegmentMergeTask` uses `GenericPartitionLocation` to
retrieve the partial range-partitioned segments generated earlier and
then merges and publishes them.
* Fix dependencies & forbidden apis
* Fixes for integration test
* Address review comments
* Fix docs, strict compile, sketch check, rollup check
* Fix first shard spec, partition serde, single subtask
* Fix first partition check in test
* Misc rewording/refactoring to address code review
* Fix doc link
* Split batch index integration test
* Do not run parallel-batch-index twice
* Adjust last partition
* Split ITParallelIndexTest to reduce runtime
* Rename test class
* Allow null values in range partitions
* Indicate which phase failed
* Improve asserts in tests
* Address security vulnerabilities CVSS >= 7
Update dependencies to address security vulnerabilities with CVSS scores
of 7 or higher. A new Travis CI job is added to prevent new
high/critical security vulnerabilities from being added.
Updated dependencies:
- api-util 1.0.0 -> 1.0.3
- jackson 2.9.10 -> 2.10.1
- kafka 2.1.0 -> 2.1.1
- libthrift 0.10.0 -> 0.13.0
- protobuf 3.2.0 -> 3.11.0
The following high/critical security vulnerabilities are currently
suppressed (so that the new Travis CI job can be added now) and are left
as future work to fix:
- hibernate-validator:5.2.5
- jackson-mapper-asl:1.9.13
- libthrift:0.6.1
- netty:3.10.6
- nimbus-jose-jwt:4.41.1
* Rename EDL1 license file
* Fix inspection errors
* Support orc format for native batch ingestion
* fix pom and remove wrong comment
* fix unnecessary condition check
* use flatMap back to handle exception properly
* move exceptionThrowingIterator to intermediateRowParsingReader
* runtime
* Fix the potential race SplittableInputSource.getNumSplits() and SplittableInputSource.createSplits() in TaskMonitor
* Fix docs and javadoc
* Add unit tests for large or small estimated num splits
* add override
* Add FileUtils.createTempDir() and enforce its usage.
The purpose of this is to improve error messages. Previously, the error
message on a nonexistent or unwritable temp directory would be
"Failed to create directory within 10,000 attempts".
* Further updates.
* Another update.
* Remove commons-io from benchmark.
* Fix tests.
* add TsvInputFormat
* refactor code
* fix grammar
* use enum replace string literal
* code refactor
* code refactor
* mark abstract for base class meant not to be instantiated
* remove constructor for test
* Refactor parallel indexing perfect rollup partitioning
Refactoring to make it easier to later add range partitioning for
perfect rollup parallel indexing. This is accomplished by adding several
new base classes (e.g., PerfectRollupWorkerTask) and new classes for
encapsulating logic that needs to be changed for different partitioning
strategies (e.g., IndexTaskInputRowIteratorBuilder).
The code is functionally equivalent to before except for the following
small behavior changes:
1) PartialSegmentMergeTask: Previously, this task had a priority of
DEFAULT_TASK_PRIORITY. It now has a priority of
DEFAULT_BATCH_INDEX_TASK_PRIORITY (via the new PerfectRollupWorkerTask
base class), since it is a batch index task.
2) ParallelIndexPhaseRunner: A decorator was added to
subTaskSpecIterator to ensure the subtasks are generated with unique
ids. Previously, only tests (i.e., MultiPhaseParallelIndexingTest)
would have this decorator, but this behavior is desired for non-test
code as well.
* Fix forbidden apis and pmd warnings
* Fix analyze dependencies warnings
* Fix IndexTask json and add IT diags
* Fix parallel index supervisor<->worker serde
* Fix TeamCity inspection errors/warnings
* Fix TeamCity inspection errors/warnings again
* Integrate changes with those from #8823
* Address review comments
* Address more review comments
* Fix forbidden apis
* Address more review comments
* Tidy up lifecycle, query, and ingestion logging.
The goal of this patch is to improve the clarity and usefulness of
Druid's logging for cluster operators. For more information, see
https://twitter.com/cowtowncoder/status/1195469299814555648.
Concretely, this patch does the following:
- Changes a lot of INFO logs to DEBUG, and DEBUG to TRACE, with the
goal of reducing redundancy and improving clarity by avoiding
showing rarely-useful log messages. This includes most "starting"
and "stopping" messages, and most messages related to individual
columns.
- Adds new log4j2 templates that show operators how to enabled DEBUG
logging for certain important packages.
- Eliminate stack traces for query errors, unless log level is DEBUG
or more. This is useful because query errors often indicate user
error rather than system error, but dumping stack trace often gave
operators the impression that there was a system failure.
- Adds task id to Appenderator, AppenderatorDriver thread names. In
the default log4j2 configuration, this will put them in log lines
as well. It's very useful if a user is using the Indexer, where
multiple tasks run in the same JVM.
- More consistent terminology when it comes to "sequences" (sets of
segments that are handed-off together by Kafka ingestion) and
"offsets" (cursors in partitions). These terms had been confused in
some log messages due to the fact that Kinesis calls offsets
"sequence numbers".
- Replaces some ugly toString calls with either the JSONification or
something more operator-accessible (like a URL or segment identifier,
instead of JSON object representing the same).
* Adjustments.
* Adjust integration test.
* Use earliest offset on kafka newly discovered partitions
* resolve conflicts
* remove redundant check cases
* simplified unit tests
* change test case
* rewrite comments
* add regression test
* add junit ignore annotation
* minor modifications
* indent
* override testableKafkaSupervisor and KafkaRecordSupplier to make the test runable
* modified test constructor of kafkaRecordSupplier
* simplify
* delegated constructor
* transformSpec + array expressions
changes:
* added array expression support to transformSpec
* removed ParseSpec.verify since its only use afaict was preventing transform expr that did not replace their input from functioning
* hijacked index task test to test changes
* remove docs about being unsupported
* re-arrange test assert
* unused imports
* imports
* fix tests
* preserve types
* suppress warning, fixes, add test
* formatting
* cleanup
* better list to array type conversion and tests
* fix oops
frank chen
Jihoon Son
Suneet Saldanha
Gian Merlino
Parag Jain
Clint Wylie
Chi Cao Minh
morrifeldman
Aleksey Plekhanov
Jonathan Wei
Yuanli Han
Mainak Ghosh
Xavier Léauté
Maytas Monsereenusorn
mcbrewster
Suneet Saldanha
Lijia Liu
Lucas Capistrant
Roman Leventov
Atul Mohan
Himanshu
Fangyuan Deng
jon-wei
Rye
Surekha
chencb