[GitHub] [druid] jasonk000 opened a new pull request #12102: HttpPostEmitter back off send() busy-loop

[GitBox <gi...@apache.org>]

jasonk000 opened a new pull request #12102:
URL: https://github.com/apache/druid/pull/12102


   ### Description
   
   Introduce a back-off to the `HttpPostEmitter` `send()` thread loop to minimise the amount of log spamming and CPU.
   
   Fixes #11279.
   
   #### Introduce HttpPostEmitterLoggerStressTest
   
   Introduce a test to count how many logged events are triggered during a loop scenario. It confirms that no more than some maximum send timeout log events are written in duration of the test. Test runs typically for 6 seconds, and expects limit of 1000 send attempts in that 6 second period.
   
   Currently, the HttpPostEmitter thread submits 150K events in the test window.
   ```
   [ERROR] Failures: 
   [ERROR] org.apache.druid.java.util.emitter.core.HttpPostEmitterLoggerStressTest.testBurstFollowedByQuietPeriod(org.apache.druid.java.util.emitter.core.HttpPostEmitterLoggerStressTest)
   [ERROR]   Run 1: HttpPostEmitterLoggerStressTest.testBurstFollowedByQuietPeriod:98 too many timeouts (159620), expect less than (1000)
   [ERROR]   Run 2: HttpPostEmitterLoggerStressTest.testBurstFollowedByQuietPeriod:98 too many timeouts (166201), expect less than (1000)
   [ERROR]   Run 3: HttpPostEmitterLoggerStressTest.testBurstFollowedByQuietPeriod:98 too many timeouts (173177), expect less than (1000)
   [ERROR]   Run 4: HttpPostEmitterLoggerStressTest.testBurstFollowedByQuietPeriod:98 too many timeouts (175135), expect less than (1000)
   ```
   
   #### Introduce a backoff to the send() loop when the specific timeout is experienced.
   
   A simple backoff is chosen at 1/5th the minimum time.
   
   An alternative solution which might be better involves a larger rewrite of this class, as the class currently splits responsibilities for managing the batching and sending across multiple classes and threads. A simpler codebase might run a simpler submission loop with a more explicit throttle and batching mechanism instead of the current batching handoffs.
   
   This PR has:
   - [x] been self-reviewed.
      - [ ] N/A using the [concurrency checklist](https://github.com/apache/druid/blob/master/dev/code-review/concurrency.md) (Remove this item if the PR doesn't have any relation to concurrency.)
   - [x] added documentation for new or modified features or behaviors.
   - [x] added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
   - [ ] N/A added or updated version, license, or notice information in [licenses.yaml](https://github.com/apache/druid/blob/master/dev/license.md)
   - [x] added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
   - [x] added unit tests or modified existing tests to cover new code paths, ensuring the threshold for [code coverage](https://github.com/apache/druid/blob/master/dev/code-review/code-coverage.md) is met.
   - [ ] N/A added integration tests.
   - [ ] been tested in a test Druid cluster.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@druid.apache.org
For additional commands, e-mail: commits-help@druid.apache.org