[
https://issues.apache.org/jira/browse/EAGLE-750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15650720#comment-15650720
]
ASF GitHub Bot commented on EAGLE-750:
--------------------------------------
GitHub user garrettlish opened a pull request:
https://github.com/apache/incubator-eagle/pull/638
[EAGLE-750] Improve coordinator schedule strategy to reuse alert work slot
1. Add switch coordinator.reuseBoltInStreams to enable/disable reuse bolt
in multiple streams
2. Add config coordinator.streamsPerBolt to set maximum streams in one bolt
3. Enable dedicated bolt for specific stream
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/garrettlish/incubator-eagle master
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/incubator-eagle/pull/638.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #638
----
commit 1be5974a460fdb61b9bf836145d3292be690d6cb
Author: Xiancheng Li <xiancheng...@ebay.com>
Date: 2016-11-09T07:17:15Z
[EAGLE-750] Improve coordinator schedule strategy to reuse alert work slot
1. Add switch coordinator.reuseBoltInStreams to enable/disable reuse bolt
in multiple streams
2. Add config coordinator.streamsPerBolt to set maximum streams in one bolt
3. Enable dedicated bolt for specific stream
----
> Improve coordinator schedule strategy to reuse alert work slot
> --------------------------------------------------------------
>
> Key: EAGLE-750
> URL: https://issues.apache.org/jira/browse/EAGLE-750
> Project: Eagle
> Issue Type: Bug
> Affects Versions: v0.5.0
> Reporter: Hao Chen
> Assignee: Garrett Li
> Fix For: v0.5.0
>
>
> We catch with some critical problem about alert engine policy schedule
> strategy:
> For example for an alert topology with 20 alert bolts , and after boarding 4
> policies and each assigning 5 slots, then the current alert coordinator will
> treat the work slots as used up and be unable to schedule any more new
> policies.
> But in fact some typical eagle deployment will have at least 20+ streams with
> lots of different partition requirements, which means we could just define
> very few policies under such scheduling strategy and may waste slot resources
> of eagle’s deployment infrastructure.
> In original design, we should have designed to reused slot queue among
> in-conflict monitored stream (stream-partition-sort), then eagle could try to
> reuse alert slots if possible to improve resource utilization, and it should
> be time to implement it now.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
