[SPARK-4989][CORE] avoid wrong eventlog conf cause cluster down in standalone mode

[liyezhang556520]

when enabling eventlog in standalone mode, if give the wrong configuration, the standalone cluster will down (cause master restart, lose connection with workers).
How to reproduce: just give an invalid value to "spark.eventLog.dir", for example: spark.eventLog.dir=hdfs://tmp/logdir1, hdfs://tmp/logdir2. This will throw illegalArgumentException, which will cause the Master restart. And the whole cluster is not available.

[SparkQA]

Test build #24858 has started for PR 3824 at commit 12eee85.

• This patch merges cleanly.

[srowen]

The word Relay was correct here

[liyezhang556520]

Yes, you are right, relay is correct, replay is not correct, thanks.

[SparkQA]

Test build #24861 has started for PR 3824 at commit a49c52f.

• This patch merges cleanly.

[SparkQA]

Test build #24858 has finished for PR 3824 at commit 12eee85.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/24858/
Test PASSed.

[SparkQA]

Test build #24861 has finished for PR 3824 at commit a49c52f.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/24861/
Test PASSed.

[JoshRosen]

It looks like eventLogFile is only read from inside the try block on the following line, so why not move it inside and make it into a val instead?

[liyezhang556520]

Hi @JoshRosen , eventLogFile is also used in catch block in this file.

[JoshRosen]

This change seems okay to me overall, aside from one minor nit. Most of the change is just broadening the scope of the try block to handle some cases that didn't seem like they could fail.

[SparkQA]

Test build #25025 has started for PR 3824 at commit 3c1ac2e.

• This patch merges cleanly.

[SparkQA]

Test build #25025 has finished for PR 3824 at commit 3c1ac2e.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/25025/
Test PASSed.

[andrewor14]

as commented elsewhere, I don't think logging the stack trace here is particularly useful

[SparkQA]

Test build #25199 has started for PR 3824 at commit 3c24d98.

• This patch merges cleanly.

[SparkQA]

Test build #25199 has finished for PR 3824 at commit 3c24d98.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/25199/
Test PASSed.

[andrewor14]

great. This used to say something like

No event logs found for application X in /history/not-found

which doesn't make sense.

[andrewor14]

Ok LGTM I'm merging this into master thanks

[liyezhang556520]

@andrewor14 , I received an email of your comment about creating other PRs to fix this issue for other older branches, but not found on this page. I think you might have removed that comment, so do I still need to make new PRs or just ignore that message?

[andrewor14]

Yes that would be great if you have the time. It seems that not all of the changes in this PR are applicable there, however.

[liyezhang556520]

ok, I'll make new PRs for those old branches 1.0, 1.1, and 1.2.