[
https://issues.apache.org/jira/browse/HIVE-19053?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16502363#comment-16502363
]
Sahil Takiar commented on HIVE-19053:
-------------------------------------
Makes sense, just a few comments:
* Can you remove the call to {{LOG#warn}} - its redundant, the exception thrown
will already get logged
* For the handling of the {{ExecutionException}} can you add in the underlying
exception message to the {{ErrorMsg}} message - e.g. add in "... due to
exception {0}" and use {{Throwables.getRootCause(e).getMessage()}} to extract
the underlying message cause
> RemoteSparkJobStatus#getSparkJobInfo treats all exceptions as timeout errors
> ----------------------------------------------------------------------------
>
> Key: HIVE-19053
> URL: https://issues.apache.org/jira/browse/HIVE-19053
> Project: Hive
> Issue Type: Sub-task
> Components: Spark
> Reporter: Sahil Takiar
> Assignee: Aihua Xu
> Priority: Major
> Attachments: HIVE-19053.1.patch
>
>
> {code}
> Future<SparkJobInfo> getJobInfo = sparkClient.run(
> new GetJobInfoJob(jobHandle.getClientJobId(), sparkJobId));
> try {
> return getJobInfo.get(sparkClientTimeoutInSeconds, TimeUnit.SECONDS);
> } catch (Exception e) {
> LOG.warn("Failed to get job info.", e);
> throw new HiveException(e, ErrorMsg.SPARK_GET_JOB_INFO_TIMEOUT,
> Long.toString(sparkClientTimeoutInSeconds));
> }
> {code}
> It should only throw {{ErrorMsg.SPARK_GET_JOB_INFO_TIMEOUT}} if a
> {{TimeoutException}} is thrown. Other exceptions should be handled
> independently.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
