Github user vanzin commented on the pull request:
https://github.com/apache/spark/pull/969#issuecomment-46051206
I rebased my code in pr #560, and tries this command:
YARN_CONF_DIR=/home/vanzin/tmp/yarn-conf ./bin/spark-submit \
/work/spark_stream_test/target/scala-2.10/streamingtest_2.10-1.0.jar \
--properties-file /home/vanzin/tmp/spark.conf --master yarn
--deploy-mode cluster \
--class com.cloudera.ss.SillyDepSleeper --jars
local:/tmp/kite-data-core-0.11.0.jar \
--files /home/vanzin/tmp/log4j.properties --archives
/home/vanzin/tmp/wordcount.tgz
`spark.conf` just sets spark.yarn.jar to `local:/something`. Note I'm using
both "--files" and "--archives", which get translated into
`spark.yarn.dist.files` and `spark.yarn.dist.archives`. I'm also adding a
custom jar to the classpath using `--jars`.
The job runs and succeeds (it depends on a class in the kite jar I
referenced in the command line), and when I look at the local directory for the
Yarn container I see this there:
/yarn/nm/usercache/vanzin/appcache/application_1402520894306_0014/container_1402520894306_0014_01_000003/wordcount.tgz
/yarn/nm/usercache/vanzin/appcache/application_1402520894306_0014/container_1402520894306_0014_01_000003/log4j.properties
(Among others.) The log file I set up in log4j.properties is also used by
the process, so it's correctly uploading my local file.
So @witgo, can you explain more clearly what is it that is not working, and
which case is not handled by my patch?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
