[
https://issues.apache.org/jira/browse/GOBBLIN-245?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Zhixiong Chen updated GOBBLIN-245:
----------------------------------
Issue Type: Task (was: Bug)
> Create topic specific extract for a WorkUnit in KafkaSource
> -----------------------------------------------------------
>
> Key: GOBBLIN-245
> URL: https://issues.apache.org/jira/browse/GOBBLIN-245
> Project: Apache Gobblin
> Issue Type: Task
> Reporter: Zhixiong Chen
> Assignee: Zhixiong Chen
> Labels: Bug:Generic, Core:JobManagement, Core:Publisher
>
> runOnce feature of Gobblin does not work correctly. A job without a schedule
> will be re-run upon Gobblin restart or modifications on the file.
> Additionally, `*.done` files are not being written.
> Root cause:
> In JobScheduler, Gobblin checks whether the job has a schedule in the method
> `scheduleJob(Properties, JobListener, Map, Class)` and sets the key
> `ConfigurationKeys.JOB_RUN_ONCE_KEY` accordingly. On the other hand, the
> method `scheduleGeneralConfiguredJobs()` checks the key
> `ConfigurationKeys.JOB_RUN_ONCE_KEY` and if runonce, creates the
> `RunOnceJobListener` that creates the `*.done` file. However,
> `scheduleGeneralConfiguredJobs()` is called before `scheduleJob(Properties,
> JobListener, Map, Class)`, so the property has not been set yet, and the
> `*.done` file is never written.
> On Gobblin restart, Gobblin checks for presence of done files, and skips jobs
> that have already been executed. However, the done file is not present, so
> the job gets repeated.
>
> *Github Url* : https://github.com/linkedin/gobblin/issues/1195
> *Github Reporter* : [~ibuenros]
> *Github Created At* : 2016-08-11T21:03:24Z
> *Github Updated At* : 2017-01-12T04:59:43Z
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
