问题

I have a typesafe configuration application.conf in the src/main/resourcesfolder which is loaded by default.

A single value can be overridden by specifying:

--conf spark.driver.extraJavaOptions=-DsomeValue="foo"

However, specifying a complete new, i.e. overriding application.conf file like:

spark-submit \
    --class my.Class \
    --master "local[2]" \
    --files foo.conf \
    --conf spark.driver.extraClassPath="-Dconfig.file=file:foo.conf" \
    --conf spark.driver.extraJavaOptions=-Dvalue="abcd" \
    job.jar

will fail to load foo.conf. Instead, the original file from the resources folder will be loaded. Trying the tricks from: Using typesafe config with Spark on Yarn did not help as well.

edit

Overriding multiple config values in Typesafe config when using an uberjar to deploy seems to be the answer for plain (without spark) programs. The question remains how to bring this to spark.

Also passing:

--conf spark.driver.extraClassPath="-Dconfig.resource=file:foo.conf"
--conf spark.driver.extraClassPath="-Dconfig.resource=foo.conf"

fails to load my configuration from the command line .

Though, according to the docs:

https://github.com/lightbend/config For applications using application.{conf,json,properties}, system properties can be used to force a different config source (e.g. from command line -Dconfig.file=path/to/config-file):

config.resource specifies a resource name - not a basename, i.e. application.conf not application

config.file specifies a filesystem path, again it should include the extension, not be a basename

config.url specifies a URL

These system properties specify a replacement for application.{conf,json,properties}, not an addition. They only affect apps using the default ConfigFactory.load() configuration. In the replacement config file, you can use include "application" to include the original default config file; after the include statement you could go on to override certain settings.

it should be possible with these parameters.

回答1:

spark-submit \
    --class my.Class \
    --master "local[2]" \
    --files foo.conf \
    --conf spark.driver.extraJavaOptions="-Dvalue='abcd' -Dconfig.file=foo.conf" \
    target/scala-2.11/jar-0.1-SNAPSHOT.jar

changing from spark.driver.extraClassPathto spark.driver.extraJavaOptions is doing the trick

来源：https://stackoverflow.com/questions/47737103/cant-override-typesafe-configuration-on-commmandline-in-spark

标签

apache-spark