carbondata-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From watermen <>
Subject [GitHub] incubator-carbondata pull request #863: [CARBONDATA-998] Don't request execu...
Date Thu, 27 Apr 2017 07:57:55 GMT
GitHub user watermen opened a pull request:

    [CARBONDATA-998] Don't request executors when we use carbon distribution

    In the current implementation, carbon request executors if it need more executors. This
brings the following questions:
    1. Carbon interferes with spark dispatch to make request/remove executors confusedly.
    2. Carbon will involk DistributionUtil.getDistinctNodesList and sleep some times to make
sure the executors are assigned, it wastes a lot of time.
    We'd better make the data scan and schedule independently.
    ### In CarbonData layout
    Make one task scan multiple splits to make the input size of task larger and the number
of tasks fewer.
    ### In Spark layout(dynamic allocation is enable)
    Request executors if some resources are released by other apps and we won't reach the
max number of executors.
    Now, I only invoke DistributionUtil.getNodeList insteads of DistributionUtil.ensureExecutorsAndGetNodeList.
We can talked about this idea here and then I can remove the useless code about DistributionUtil.ensureExecutorsAndGetNodeList.
    cc @jackylk @QiangCai @kumarvishal09 

You can merge this pull request into a Git repository by running:

    $ git pull CARBONDATA-998

Alternatively you can review and apply these changes as the patch at:

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

    This closes #863
commit 824d0f7790f5559d48be6b1878786bc384af51a9
Author: Yadong Qi <>
Date:   2017-04-27T07:54:49Z

    Don't request executors when we use carbon distribution


If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at or file a JIRA ticket
with INFRA.

View raw message