[SPARK-12396][Core]Once driver connect to a master successfully, stop it connect to master again.

[echoTomei]

No description provided.

[andrewor14]

ok to test

[andrewor14]

Good catch, this change looks fine.

[SparkQA]

Test build #48125 has finished for PR 10407 at commit 7959c1f.

• This patch fails Scala style tests.
• This patch merges cleanly.
• This patch adds no public classes.

[andrewor14]

@echoTomei it's failing style tests because there's a whitespace at the end of the line

[echoTomei]

@andrewor14 Sorry for my careless.I'll create another PR.

[SparkQA]

Test build #48151 has finished for PR 10407 at commit 3a38c14.

• This patch fails to build.
• This patch merges cleanly.
• This patch adds no public classes.

[echoTomei]

@SparkQA The build error shows that the function cancel() is not a member of [java.util.concurrent.ScheduledFuture[_]] . I check the interface ScheduledFuture API ,confirming that this interface has cancel() function. So I don't understand why occur this error. Can you give me some suggestion?Thanks.

[andrewor14]

retest this please

[SparkQA]

Test build #48208 has finished for PR 10407 at commit 3a38c14.

• This patch fails to build.
• This patch merges cleanly.
• This patch adds no public classes.

[andrewor14]

That is really confusing... @srowen do you know if we're missing something obvious?

[srowen]

Actually the error says that the var is an AtomicReference and this does not have a cancel method. Indeed, it does in master, after e1bcf6a

Is this PR against a quite old version of master? while it doesn't technically need a rebase, it really does.

[zsxwing]

I think a better approach is changing registrationRetryThread.scheduleAtFixedRate to registrationRetryThread.schedule.

[andrewor14]

I think a better approach is changing registrationRetryThread.scheduleAtFixedRate to registrationRetryThread.schedule.

+1. Right now we call it both recursively and iteratively, which doesn't make sense...

[echoTomei]

It seemed that change to registrationRetryThread.schedule is a better way.

[zsxwing]

Could you rebase on the latest master?

[SparkQA]

Test build #48231 has finished for PR 10407 at commit 93c005d.

• This patch fails R style tests.
• This patch does not merge cleanly.
• This patch adds no public classes.

[andrewor14]

Can you close this PR now that you've opened another one #10447