eb8c420edb
### What changes were proposed in this pull request? Support FETCH_PRIOR fetching in Thriftserver, and report correct fetch start offset it TFetchResultsResp.results.startRowOffset The semantics of FETCH_PRIOR are as follow: Assuming the previous fetch returned a block of rows from offsets [10, 20) * calling FETCH_PRIOR(maxRows=5) will scroll back and return rows [5, 10) * calling FETCH_PRIOR(maxRows=10) again, will scroll back, but can't go earlier than 0. It will nevertheless return 10 rows, returning rows [0, 10) (overlapping with the previous fetch) * calling FETCH_PRIOR(maxRows=4) again will again return rows starting from offset 0 - [0, 4) * calling FETCH_NEXT(maxRows=6) after that will move the cursor forward and return rows [4, 10) ##### Client/server backwards/forwards compatibility: Old driver with new server: * Drivers that don't support FETCH_PRIOR will not attempt to use it * Field TFetchResultsResp.results.startRowOffset was not set, old drivers don't depend on it. New driver with old server * Using an older thriftserver with FETCH_PRIOR will make the thriftserver return unsupported operation error. The driver can then recognize that it's an old server. * Older thriftserver will return TFetchResultsResp.results.startRowOffset=0. If the client driver receives 0, it can know that it can not rely on it as correct offset. If the client driver intentionally wants to fetch from 0, it can use FETCH_FIRST. ### Why are the changes needed? It's intended to be used to recover after connection errors. If a client lost connection during fetching (e.g. of rows [10, 20)), and wants to reconnect and continue, it could not know whether the request got lost before reaching the server, or on the response back. When it issued another FETCH_NEXT(10) request after reconnecting, because TFetchResultsResp.results.startRowOffset was not set, it could not know if the server will return rows [10,20) (because the previous request didn't reach it) or rows [20, 30) (because it returned data from the previous request but the connection got broken on the way back). Now, with TFetchResultsResp.results.startRowOffset the client can know after reconnecting which rows it is getting, and use FETCH_PRIOR to scroll back if a fetch block was lost in transmission. Driver should always use FETCH_PRIOR after a broken connection. * If the Thriftserver returns unsuported operation error, the driver knows that it's an old server that doesn't support it. The driver then must error the query, as it will also not support returning the correct startRowOffset, so the driver cannot reliably guarantee if it hadn't lost any rows on the fetch cursor. * If the driver gets a response to FETCH_PRIOR, it should also have a correctly set startRowOffset, which the driver can use to position itself back where it left off before the connection broke. * If FETCH_NEXT was used after a broken connection on the first fetch, and returned with an startRowOffset=0, then the client driver can't know if it's 0 because it's the older server version, or if it's genuinely 0. Better to call FETCH_PRIOR, as scrolling back may anyway be possibly required after a broken connection. This way it is implemented in a backwards/forwards compatible way, and doesn't require bumping the protocol version. FETCH_ABSOLUTE might have been better, but that would require a bigger protocol change, as there is currently no field to specify the requested absolute offset. ### Does this PR introduce any user-facing change? ODBC/JDBC drivers connecting to Thriftserver may now implement using the FETCH_PRIOR fetch order to scroll back in query results, and check TFetchResultsResp.results.startRowOffset if their cursor position is consistent after connection errors. ### How was this patch tested? Added tests to HiveThriftServer2Suites Closes #26014 from juliuszsompolski/SPARK-29349. Authored-by: Juliusz Sompolski <julek@databricks.com> Signed-off-by: Yuming Wang <wgyumg@gmail.com> |
||
---|---|---|
.github | ||
assembly | ||
bin | ||
build | ||
common | ||
conf | ||
core | ||
data | ||
dev | ||
docs | ||
examples | ||
external | ||
graph | ||
graphx | ||
hadoop-cloud | ||
launcher | ||
licenses | ||
licenses-binary | ||
mllib | ||
mllib-local | ||
project | ||
python | ||
R | ||
repl | ||
resource-managers | ||
sbin | ||
sql | ||
streaming | ||
tools | ||
.gitattributes | ||
.gitignore | ||
appveyor.yml | ||
CONTRIBUTING.md | ||
LICENSE | ||
LICENSE-binary | ||
NOTICE | ||
NOTICE-binary | ||
pom.xml | ||
README.md | ||
scalastyle-config.xml |
Apache Spark
Spark is a unified analytics engine for large-scale data processing. It provides high-level APIs in Scala, Java, Python, and R, and an optimized engine that supports general computation graphs for data analysis. It also supports a rich set of higher-level tools including Spark SQL for SQL and DataFrames, MLlib for machine learning, GraphX for graph processing, and Structured Streaming for stream processing.
Online Documentation
You can find the latest Spark documentation, including a programming guide, on the project web page. This README file only contains basic setup instructions.
Building Spark
Spark is built using Apache Maven. To build Spark and its example programs, run:
./build/mvn -DskipTests clean package
(You do not need to do this if you downloaded a pre-built package.)
You can build Spark using more than one thread by using the -T option with Maven, see "Parallel builds in Maven 3". More detailed documentation is available from the project site, at "Building Spark".
For general development tips, including info on developing Spark using an IDE, see "Useful Developer Tools".
Interactive Scala Shell
The easiest way to start using Spark is through the Scala shell:
./bin/spark-shell
Try the following command, which should return 1,000,000,000:
scala> spark.range(1000 * 1000 * 1000).count()
Interactive Python Shell
Alternatively, if you prefer Python, you can use the Python shell:
./bin/pyspark
And run the following command, which should also return 1,000,000,000:
>>> spark.range(1000 * 1000 * 1000).count()
Example Programs
Spark also comes with several sample programs in the examples
directory.
To run one of them, use ./bin/run-example <class> [params]
. For example:
./bin/run-example SparkPi
will run the Pi example locally.
You can set the MASTER environment variable when running examples to submit
examples to a cluster. This can be a mesos:// or spark:// URL,
"yarn" to run on YARN, and "local" to run
locally with one thread, or "local[N]" to run locally with N threads. You
can also use an abbreviated class name if the class is in the examples
package. For instance:
MASTER=spark://host:7077 ./bin/run-example SparkPi
Many of the example programs print usage help if no params are given.
Running Tests
Testing first requires building Spark. Once Spark is built, tests can be run using:
./dev/run-tests
Please see the guidance on how to run tests for a module, or individual tests.
There is also a Kubernetes integration test, see resource-managers/kubernetes/integration-tests/README.md
A Note About Hadoop Versions
Spark uses the Hadoop core library to talk to HDFS and other Hadoop-supported storage systems. Because the protocols have changed in different versions of Hadoop, you must build Spark against the same version that your cluster runs.
Please refer to the build documentation at "Specifying the Hadoop Version and Enabling YARN" for detailed guidance on building for a particular distribution of Hadoop, including building for particular Hive and Hive Thriftserver distributions.
Configuration
Please refer to the Configuration Guide in the online documentation for an overview on how to configure Spark.
Contributing
Please review the Contribution to Spark guide for information on how to get started contributing to the project.