You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Bowen Li (JIRA)" <ji...@apache.org> on 2017/08/31 21:39:00 UTC
[jira] [Created] (FLINK-7566) if there's only one checkpointing
metadata file in , `flink run -s ` should resume from that
metadata file
Bowen Li created FLINK-7566:
-------------------------------
Summary: if there's only one checkpointing metadata file in <dir>, `flink run -s <dir>` should resume from that metadata file
Key: FLINK-7566
URL: https://issues.apache.org/jira/browse/FLINK-7566
Project: Flink
Issue Type: Improvement
Components: Startup Shell Scripts
Affects Versions: 1.3.2
Environment: Currently, if we want to start a Flink job from a checkpointing file, we have to run `flink run -s <dir>/checkpoint_metadata-xxxxx` by explicitly specifying the checkpoint metadata file name 'checkpoint_metadata-xxxxx'. Since metadata file name always changes, it's not easy to programmatically restart a failed Flink job.
What I want is like this: users should be able to start a Flink job by running `flink run -s <dir>` if there's only one checkpointing metadata file in <dir>. If there's none or more than 1 metadata file, the command can fail like it is right now. This way, we can programmatically restart a failed Flink job by hardcoding <dir>
Reporter: Bowen Li
Fix For: 1.4.0
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)