With the existing hash function some nodes could end up with 3 times the
number of keys as others. The following changes improve that to roughly
less than 5% differences across nodes.
- switch from fnv-1a to murmur3_128 hash
- increase repetitions for ketama algorithm
- test to analyze distribution
Also updates spymemcached for recent bugfixes
this PR adds a JavaScriptWorkerSelectStrategy which allows defining
arbitrary logic for selecting workers to run task using a JavaScript
function.
This gives users full control to implement complex worker selection
strategies based on task attributes.
more tests and a complex javascript config
fix for java8 modify for nashorn compatibility
It is highly recommended to keep loading threads single threaded. There can be benefits to having multiple bootstrapping threads, though. Those bootstrapping threads shouldn't be kept alive for the whole life of the process, however.
Also, adjust some logging and do some code cleanup.
which could cause Guice duplicate binding errors.
Add unit test to replicate duplicate extension issue. Update documents. Add an package accessible getter for loadersMap for testing only.
Notice that extensions explicitly specified in druid.extensions.coordinates have a higher priority than ones included in the classpath.
Extension modules that don't have a canonical class name will be ignored.
- pass rowFlushboundary correctly instead of using default.
- fixes indexTask failing with
io.druid.segment.incremental.IndexSizeExceededException when
rowFlushboundary is set higher than
RealtimeTuningConfig.defaultMaxRowsInMemory
rename test method
* Moves many items to JobHelper
* Remove dependencies of these functions on HadoopDruidIndexerConfig in favor of more general items
* Changes functionalities of some of the path methods to always return a path with scheme
* Adds retry to uploads
* Change output loadSpec determining from using outputFS.getClass().getName() to using outputFS.getScheme()
* Fix resource leak in `io.druid.segment.IndexIO.DefaultIndexIOHandler#validateTwoSegments(java.io.File, java.io.File)`
* Un-deprecate `close()` in `QueryableIndex` and make it inherit `Closeable`
* Fix resource leaks in various unit tests
* Add `CloserRule` for closing out resources
Anyways, @guobingkun , if possible, can you please test proposed change and send another PR to fix#1016 . but that doesn't necessarily hold this PR, so I'm merging this.
Hao Xia
Charles Allen
Xavier Léauté
sahner
nishant
Xavier Léauté
Eric Tschetter
Fangjin Yang
Bingkun Guo
cheddar
Parag Jain
Himanshu