johnyangk commented on a change in pull request #219: [NEMO-351] Empowering
Nemo with fast I/O using Apache Crail
URL: https://github.com/apache/incubator-nemo/pull/219#discussion_r292246340
##########
File path: client/src/main/java/org/apache/nemo/client/JobLauncher.java
##########
@@ -397,6 +397,8 @@ public static Configuration getJobConf(final String[]
args) throws IOException,
cl.registerShortNameOfClass(JobConf.MaxTaskAttempt.class);
cl.registerShortNameOfClass(JobConf.FileDirectory.class);
cl.registerShortNameOfClass(JobConf.GlusterVolumeDirectory.class);
+ cl.registerShortNameOfClass(JobConf.CrailVolumeDirectory.class);
Review comment:
Can you file a new JIRA issue to convert JobConf parameters for directory
paths (FileDirectory, GlusterVolumeDirectory, CrailVolumeDirectory) to
DataStoreProperty parameters? To do this, the type of the possible values of
the DataStoreProperty should also change from `enum` to `class`.
I suggest this change, because I think we should aim to integrate into the
IR DAG as many configuration parameters as possible. Another benefit is that it
saves the trouble of having to configure both JobConf and IR DAG to use a
specific Crail/Gluster/File directory. Finally, this enables us to configure
different directories for different IREdges.
A related issue is integrating the executor_json parameter into the IR DAG:
https://issues.apache.org/jira/browse/NEMO-382
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
