[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16850490#comment-16850490 ] Congxian Qiu(klion26) commented on FLINK-12296: --- Thanks for the reminder [~sunjincheng121]

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16850421#comment-16850421 ] sunjincheng commented on FLINK-12296: - Thanks for double check this JIRA, and close it. :) > Data

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16846382#comment-16846382 ] Congxian Qiu(klion26) commented on FLINK-12296: --- merged * master        

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16827806#comment-16827806 ] Congxian Qiu(klion26) commented on FLINK-12296: --- After discussed offline with [~srichter],

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16826787#comment-16826787 ] Stefan Richter commented on FLINK-12296: [~klion26] I was just taking a look into the code and

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16825650#comment-16825650 ] Ning Shi commented on FLINK-12296: -- [~srichter] The assumption totally made sense in that context. I

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16825199#comment-16825199 ] Congxian Qiu(klion26) commented on FLINK-12296: --- [~srichter] I first implemented as you

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16825138#comment-16825138 ] Stefan Richter commented on FLINK-12296: [~klion26] In general I think it is a good idea to

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16825087#comment-16825087 ] Ning Shi commented on FLINK-12296: -- [~srichter] You are right, we use reinterpretAsKeyedStream between

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16825034#comment-16825034 ] Congxian Qiu(klion26) commented on FLINK-12296: --- [~srichter]  from what [~ningshi] have

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16824945#comment-16824945 ] Stefan Richter commented on FLINK-12296: I guess this is one remaining relic from how backends

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16824491#comment-16824491 ] seed Z commented on FLINK-12296: Unchaining the operator makes the new job be able to save the state via

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16824187#comment-16824187 ] Ning Shi commented on FLINK-12296: -- Unchaining the two stateful operators seems to have worked around

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16823998#comment-16823998 ] Congxian Qiu(klion26) commented on FLINK-12296: --- I think this is a long existing issue

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16823995#comment-16823995 ] Ning Shi commented on FLINK-12296: -- We observed the issue on Flink 1.6.1. We haven’t tested on later

[jira] [Commented] (FLINK-12296) Data loss silently in RocksDBStateBackend when more than one operator(has states) chained in a single task

[ https://issues.apache.org/jira/browse/FLINK-12296?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16823988#comment-16823988 ] Yu Li commented on FLINK-12296: --- Thanks for filing the JIRA [~klion26]. Is this some regression issue or