HDFS-11114. Support for running async disk checks in DataNode.#153
HDFS-11114. Support for running async disk checks in DataNode.#153arp7 wants to merge 2 commits intoapache:trunkfrom arp7:trunk
Conversation
Change-Id: Ib21cd21fe9b67ca35b38f8462c138e90b55f33df
| * | ||
| * @throws InterruptedException | ||
| */ | ||
| void join(long timeout, TimeUnit timeUnit) throws InterruptedException; |
There was a problem hiding this comment.
Just trying to understand this a little better, From the signature and implementation this function looks more a awaitTermination in executorService. That is this function will wait for a while and cancel and task if the timeout occurs, in that case would you consider calling this await or awaitTermination. Java "join" seems to imply a wait without timeouts. Just making sure that the intended was indeed a shutdown/await pattern.
There was a problem hiding this comment.
Thanks for taking a look @anuengineer. The method covers both shutdown and awaitTermination semantics. I could call it shutdownAndAwaitTermination() to make it clearer.
Change-Id: I3b6dd71f2421b6703fc09b0f3491bd3e18bbf898
|
Renamed it to shutdownAndWait. |
anuengineer
left a comment
There was a problem hiding this comment.
+1, looks good to me. Thanks for providing this patch.
…kController Addresses the following: * Makes LeaderElectionListener to be explicitly registered by the caller * Removes debouncetimer dependency from ZkLeaderElector implementation * [Bug] onBecomeLeader was scheduling a task in timer under "OnBecomeLeader", when it should actually be the same as "OnProcessorChange". Otherwise, it will not cancel when there is a new OnProcessorChange event. * [Transient Test Failure] `TestScheduleAfterDebounceTime` tests were relying on timing controlled by sleep. Fixed it by using latch Author: Navina Ramesh <navina@apache.org> Reviewers: Xinyu Liu <xiliu@linkedin.com>, Boris Shkolnik <boryas@apache.org> Closes apache#153 from navina/SAMZA-1251
…bounceTimer This PR depends on PR apache#153 * Treats all errors in jobcoordinator as FATAL and shuts-down the streamprocessor * [Bug] Fixed bug reported in SAMZA-1241 * Introduced a callback to be associated with the timer (same callback for every Runnable failure) **TBD**: some more unit tests Author: Navina Ramesh <navina@apache.org> Reviewers: Xinyu Liu <xiliu@apache.org>, Prateek Maheshwari <pmaheshw@linkedin.com> Closes apache#166 from navina/SAMZA-1150
Interface for running async checks on a resource.
The implementation ThrottledAsyncChecker supports throttling and result-caching. DataNode changes to use it will be done in another Jira.