[Hot Fix][Streaming] Fix WriteAheadLogBackedBlockRDDSuite

[scwf]

WriteAheadLogBackedBlockRDDSuite failed due to the same random file name(https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/24456/consoleFull)

/tmp/1418645532799-0/??????????
[info] - Read data available only in write ahead log, and test storing in block manager *** FAILED *** (2 milliseconds)

Using another way to produce random file name(http://alvinalexander.com/scala/creating-random-strings-in-scala).

[scwf]

link to #3687

[SparkQA]

Test build #24458 has started for PR 3701 at commit b7f8955.

• This patch merges cleanly.

[srowen]

This is for SPARK-4826 right? can you explain the fix? is it that reserved characters like / or non-ASCII chars were part of the file name?

[scwf]

Yes, it is. I have not noticed that PR. Based on my test, this failure is caused by Random.nextString(10), it produce the same string each time when test(a debug link on jenkins: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/24456/consoleFull) for unkonw reason. I am not sure this is caused by bugs of Random.nextString or race condition

[SparkQA]

Test build #24458 has finished for PR 3701 at commit b7f8955.

• 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/24458/
Test PASSed.

[JoshRosen]

I've opened a new PR with what I think is is a more direct approach to fixing this issue: #3704. Please take a look and let me know what you think.