You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ambari.apache.org by "Sumit Mohanty (JIRA)" <ji...@apache.org> on 2016/08/30 01:57:20 UTC

[jira] [Updated] (AMBARI-16828) Support round-robin scheduling with failover for Sinks with distributed collector

     [ https://issues.apache.org/jira/browse/AMBARI-16828?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sumit Mohanty updated AMBARI-16828:
-----------------------------------
    Fix Version/s:     (was: 2.4.1)
                   trunk

> Support round-robin scheduling with failover for Sinks with distributed collector
> ---------------------------------------------------------------------------------
>
>                 Key: AMBARI-16828
>                 URL: https://issues.apache.org/jira/browse/AMBARI-16828
>             Project: Ambari
>          Issue Type: Task
>          Components: amvari-me
>    Affects Versions: 2.4.1
>            Reporter: Siddharth Wagle
>            Assignee: Siddharth Wagle
>             Fix For: trunk
>
>         Attachments: AMBARI-16828.patch
>
>
> - Initial set of collectors is configured in the configuration files
> - Find available collectors by connecting to zookeeper thereafter 
> - Remember available collectors, refresh this information only when one collector cannot be reached with a very low frequency of checks, example: random interval between 10-12 minutes, check if a new collector is available. Set a low client side zk timeout.
> - Round robin the write between the collector choosing the first one at random
> - If a write timed out, choose the next available collector, remember the attempts with the first one
> - Set a configurable attempt count for failed connector (default = 3), after which the failed connector is no longer in the available collectors list. 
> - The next retry will be triggered after refresh with zookeeper is successful
> - If no failed collectors available, zk refresh interval should be chosen randomly between 1-2 minutes.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)