Update to Curator 4.2.0, ZooKeeper 3.4.14.

[gianm]

Other than generally wanting to use the latest Curator and ZK, this change is motivated by an outage I encountered last night. I was debugging a cluster last night that was acting bizarrely, and in the end it turned out that it had two overlords that both thought they were leader. Shortly before they both gained leadership, the ZK quorum was unavailable for about 20 seconds. It doesn't look like Druid itself was doing anything particularly wrong: logs indicated the overlords weren't ignoring stopBeingLeader calls or anything like that.

For these reasons, I believe the cause of the outage was https://issues.apache.org/jira/browse/CURATOR-498. This comment indicates the bug could cause two LeaderLatch users to become leaders at once: https://issues.apache.org/jira/browse/CURATOR-498?focusedCommentId=16732419&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-16732419.

The bug was fixed in Curator 4.2.0.

[gianm]

The test failures look suspicious. There is no such method as ByteBuffer position(int) in java.nio.ByteBuffer in Java 8, which is what the error message is referring to. I wonder if the curator-framework jar was built with JDK 9+ (unintentionally? since I don't think they intended to drop Java 8 support).

[ERROR] testSanity(org.apache.druid.curator.announcement.AnnouncerTest)  Time elapsed: 1.158 s  <<< ERROR!
java.lang.NoSuchMethodError: java.nio.ByteBuffer.position(I)Ljava/nio/ByteBuffer;
	at org.apache.curator.framework.imps.GzipCompressionProvider.skip(GzipCompressionProvider.java:319)
	at org.apache.curator.framework.imps.GzipCompressionProvider.doReadHeader(GzipCompressionProvider.java:290)
	at org.apache.curator.framework.imps.GzipCompressionProvider.readGzipHeader(GzipCompressionProvider.java:273)
	at org.apache.curator.framework.imps.GzipCompressionProvider.decompress(GzipCompressionProvider.java:195)

[asdf2014]

@gianm Yep, this is a known bug for curator, more details:
https://issues.apache.org/jira/browse/CURATOR-523

[himanshug]

+1 when the build works

[fjy]

@gianm can you fix test failures?

[gianm]

@gianm can you fix test failures?

It's broken because of the upstream Curator issue that @asdf2014 linked.

[stale]

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If you think that's incorrect or this pull request should instead be reviewed, please simply write any comment. Even if closed, you can still revive the PR at any time or discuss it on the dev@druid.apache.org list. Thank you for your contributions.

[gianm]

Still waiting for a new release of Curator.

[stale]

This issue is no longer marked as stale.

[stale]

This pull request has been marked as stale due to 60 days of inactivity. It will be closed in 4 weeks if no further activity occurs. If you think that's incorrect or this pull request should instead be reviewed, please simply write any comment. Even if closed, you can still revive the PR at any time or discuss it on the dev@druid.apache.org list. Thank you for your contributions.

[gianm]

Fine, let's close this until a new Apache Curator is released.

[ccaominh]

Looks like curator 4.2.0 and 4.3.0 are now available.

[maytasm]

curator 4.3.0 should have the fix. I can make a PR for this