samza-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Shadi A. Noghabi" <abdol...@illinois.edu>
Subject Re: Review Request 36006: SAMZA-724: Writing a tool to read from the coordinator stream and react to config changes accordingly.
Date Fri, 14 Aug 2015 01:25:13 GMT

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/36006/
-----------------------------------------------------------

(Updated Aug. 14, 2015, 1:25 a.m.)


Review request for samza, Yi Pan (Data Infrastructure), Navina Ramesh, and Naveen Somasundaram.


Changes
-------

Fixed some java doc nits.


Repository: samza


Description
-------

This RBis for SAMZA-724.

After a job is submitted, it might need some configuration change, specifically it might need
more containers. In SAMZA-704 a tool is being added to write to the coordinator stream (CoordinatorStreamWriter).
 This tool can be used to write new configurations to the coordinator stream. However, another
tool (ConfigManager) is needed to read the config changes and react to them, which is the
goal of this task. This tool should be brought up after the job is submitted and read any
config changes added to the coordinator stream, and react to each accordingly. 

This tool, called the Config Manager, is focusing on handling container changs by reacting
to set-config massages with key "yarn.container.count". 

The config manager is a separate standa alone module, that should be brought up separately
after the submission of a job. Therefore, you have to add two configurations to the input
config file:
1. yarn.rm.address= <ip of resource manager in yarn. ex: localhost >
2. yarn.rm.port= <the port of the resource manager http server. ex: 8088 >

The config manger will periodically poll the coordinator stream to see if there are any new
messages. This period is set to 100 ms by deafualt. However, it can be configured by adding
configManager.polling.interval=<polling interval> to the input config file. Thus, overal
the command to run the config manager along with the job would be:


<path to samza deployment>/bin/run-config-manager.sh --config-factory=<config factory>
--config-path=<path to config file of a job>


Diffs (updated)
-----

  build.gradle a935088eccb3aee4fbde21275fa8e701c215a69e 
  checkstyle/import-control.xml 24ed680785175f3cdf955602b7a813684edd813e 
  gradle/dependency-versions.gradle fb06e8ed393d1a38abfa1a48fe5244fc7f6c7339 
  samza-autoscaling/src/main/java/org/apache/samza/autoscaling/deployer/ConfigManager.java
PRE-CREATION 
  samza-autoscaling/src/main/java/org/apache/samza/autoscaling/utils/YarnUtil.java PRE-CREATION

  samza-core/src/main/java/org/apache/samza/coordinator/stream/CoordinatorStreamSystemConsumer.java
2277a732b9ab763edf19a0fbec288ff72b27583b 
  samza-core/src/main/scala/org/apache/samza/job/JobRunner.scala d7c928c7401e539a370d4e82276e7dabbce1b638

  samza-shell/src/main/bash/run-config-manager.sh PRE-CREATION 
  samza-test/src/test/scala/org/apache/samza/test/integration/StreamTaskTestUtil.scala c6e994ff707af802ded57c3bc1762971892014da

  samza-yarn/src/main/scala/org/apache/samza/job/yarn/SamzaAppMasterService.scala ce88698c12c4bf6f4cf128f92d60b0b9496997d7

  settings.gradle a8d2c885254ca3994327fda18e09c49bc9c5e830 

Diff: https://reviews.apache.org/r/36006/diff/


Testing
-------

Tested with hello samza and works properly.


Thanks,

Shadi A. Noghabi


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message