* Removing proxy execution context
Retrieving query info using new query plan retriever
* Implementing PR comments
* Returning bad request for cross partition non value aggregates
* Fixing failing aggregate tests
* Fixing failing aggregate tests
* Fixing spywirecontent test
* Adding a feedoption to allow empty pages in feedresponse
* Adding licence
* Minor refactoring
* Fixing test failures
* Fixing test failures
* Queries with partitionKeyRangeIdInternal or partitionKey are routed to appropriate partition ranges
* Fixing test failures
* Implementing PR comments
* draft fix for global strong retry
* draft fix for global strong retry
* improved test setup time by doing bulk insert in parallel instead of serial
* bumped timeout values
* debugging
* cleanup
* added e2e test profile, relaxed replica counting test
* read writes test
* tune test
* retry analyzer renamed
* fixed test
* fixed test
* fix tests
* travis fix
* Removed skip exception for TCP as the test is disabled for all protocols for now
* update
* try/catch for diagnostic string
* fixed compilation error
* fixed test
* fixed test
* WIP (Work In Progress) Merged internal direct implementation to open source branch
* Added localhost to test configurations
* Safe Deleting collection after test case is done
* Trying smaller set of collections
* Reverting it back to normal
* Merging latest develop branch from internal code base
* Merge RNTBD work-in-progress from internal repository
* Fixed some request timer issues
* Fixed some request timer issues
* Fixed some request timer issues
* FakeEndpoint is now RntbdRequestTimer-capable
* Added a blank line and confirmed that all unit tests pass
* Renamed RntbdTransportClient.Options.maxChannels as RntbdTransportClient.Options.maxChannelsPerEndpoint and reduced its default from from 65535 to 10. Impact: Improved reliability and performance.
* Renamed RntbdRequestManager.PendingRequest as RntbdRequestRecord and improved error handling in RntbdRequestManager.write to ensure correct retry behavior in the layer above RntbdTransportClient.
* RntbdRequestRecord is now a CompletableFuture<StoreResponse>. RntbdServiceEndpoint instances now write RntbdRequestRecord instances. The result: improved encapsulation and code readability.
* Code tweaks for correctness
* Refactored RntbdTransportClient.Options for improved usability.
* Refactored for improved diagnstics, testability, and usability.
* Improved request timeout error message
* Marked some classes as final
* Improved metrics and logger.debug messages. Also: confirmed direct and simple (fast), and unit tests pass locally
* Added and updated dependencies (benchmark and direct-impl) on metrics
* Corrected merge issues
* Improved RntbdMetrics and corrected some deficiencies in RntbdClientChannelPool
* Improved RntbdMetrics and corrected some deficiencies in RntbdClientChannelPool
* We now report illegal state instead of throwing an illegal state exception in two cases which may be at the root of a failure related to Read/WriteTimeoutExceptions
* Corrected an error message
* Corrected an error message
* Addressed some error recovery/reporting issues
* Added RntbdTransportClient specification
* Tweaked error handling and removed some dead code
* Tweaked error handling and removed some dead code
* Tweaked error handling and removed some dead code
* Tweaked error handling
* Tidied top-level pom.xml
* Enabled back-pressure tests for Direct TCP and removed some test TODOs. In the process: found a ByteBuf memory leak and found and fixed a NullPointerException in RntbdClientChannelPool. Also: Improved channel load balancing algorithm a little. See the changes to RntbdClientChannelPool.
* Minor cleanup (e.g., removed unused imports and unnecessary uses of 'throws' signatures)
* Tweaked logging
* Added call to SslHandler.closeOutbound to ensure that all ByteBuf and SslEngine resources are released when a channel closes exceptionally
* Upgraded jackson-databind version to 2.9.9 because Snyk-Dependency-Vulnerabilities-Detection-SQL-RxJava complained that 2.9.8 presents a vulnerability
* Revised error handling in RntbdRequestManager.channelRead
* Improvements to error/issue reporting
* Tweaked a message
* Removed some debug checks that are no longer required
* Revised a loop and added a logger message for use as a breakpoint
* tweaks
* tweaks
* tweaks
* tweaks
* tweaks
* tweaks
* tweaks
* tweaks
* Tweaked completion handlers
* We now return an RntbdRequestRecord, not a Completion stage, to RntbdTransportClient. This improves our ability to track requests.
* Tweaks
* Unit tests pass
* RntbdReporter now protects against a class initialization error that may occur when a Java Security Manager prohibites access to the class's protection domain
* Tweaks for code clarity and improved error messages
* Channel pool refinements
* Channel pool and request manager refinements. Confirmed: fast, long, and unit tests pass locally.
* Channel pool and request manager refinements. Confirmed: fast, long, and unit tests pass locally.
* Added fix for Windows to RntbdReporter.java
* Added fix for Windows to RntbdReporter.java
* Bumped timeout interval to avoid an intermittent CI issue on Emulator runs.
* Attempt to remove dependency security vulnerability described here: https://snyk.io/vuln/SNYK-JAVA-COMMONSBEANUTILS-30077
* Updated dependency on netty-tcnative from 2.0.20.Final to 2.0.25.Final
* Resolved issue #140
* Resolved issue #140
* Removed unused dependency
* Address issue #163
* Address backpressure test failures by updating the definition of RntbdRequestManager.isServiceable
* Address backpressure test failures by updating the definition of RntbdRequestManager.isServiceable
* WIP (Work In Progress) Merged internal direct implementation to open source branch
* Added localhost to test configurations
* Safe Deleting collection after test case is done
* Trying smaller set of collections
* Reverting it back to normal
* Merging latest develop branch from internal code base
* Merge RNTBD work-in-progress from internal repository
* Fixed some request timer issues
* Fixed some request timer issues
* Fixed some request timer issues
* FakeEndpoint is now RntbdRequestTimer-capable
* Added a blank line and confirmed that all unit tests pass
* Renamed RntbdTransportClient.Options.maxChannels as RntbdTransportClient.Options.maxChannelsPerEndpoint and reduced its default from from 65535 to 10. Impact: Improved reliability and performance.
* Renamed RntbdRequestManager.PendingRequest as RntbdRequestRecord and improved error handling in RntbdRequestManager.write to ensure correct retry behavior in the layer above RntbdTransportClient.
* RntbdRequestRecord is now a CompletableFuture<StoreResponse>. RntbdServiceEndpoint instances now write RntbdRequestRecord instances. The result: improved encapsulation and code readability.
* Code tweaks for correctness
* Refactored RntbdTransportClient.Options for improved usability.
* Refactored for improved diagnstics, testability, and usability.
* Improved request timeout error message
* Marked some classes as final
* Improved metrics and logger.debug messages. Also: confirmed direct and simple (fast), and unit tests pass locally
* Added and updated dependencies (benchmark and direct-impl) on metrics
* Corrected merge issues
* Improved RntbdMetrics and corrected some deficiencies in RntbdClientChannelPool
* Improved RntbdMetrics and corrected some deficiencies in RntbdClientChannelPool
* We now report illegal state instead of throwing an illegal state exception in two cases which may be at the root of a failure related to Read/WriteTimeoutExceptions
* Corrected an error message
* Corrected an error message
* Addressed some error recovery/reporting issues
* Added RntbdTransportClient specification
* Tweaked error handling and removed some dead code
* Tweaked error handling and removed some dead code
* Tweaked error handling and removed some dead code
* Tweaked error handling
* Tidied top-level pom.xml
David.Noble@microsoft.com
mbhaskar
Christopher Anderson
Mohammad Derakhshani
David Noble