[ 
https://issues.apache.org/jira/browse/FLINK-28274?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17562419#comment-17562419
 ] 

zl commented on FLINK-28274:
----------------------------

-- Does this problem only occur for \{ContinuousFileMonitoringFunction} or also 
for the new \{FileSource}

As [~wanglijie95] said, with FLIP-27, FileSource does not need 
{{ContinuousFileMonitoringFunction}}  to plays the role of SplitEnumerator, so 
this problem won't occur for new FileSource.

 

-- does it also occur with other operators whose parallelism must be 1 ?

>From what I understand, if the max parallelism is not set to 1, it's possible 
>that this problem may occur for these operators whose parallelism must be 1.

 

-- why fix it if it's only occurring in a legacy component ?

The legacy source are still widely used, for example, When reading files with 
datastream api, the first thought is to use readFile/readTextFile() methods, 
especially for starters. Since the fix work won't take much effort, may be it's 
worthy.

> ContinuousFileMonitoringFunction doesn't work with reactive mode
> ----------------------------------------------------------------
>
>                 Key: FLINK-28274
>                 URL: https://issues.apache.org/jira/browse/FLINK-28274
>             Project: Flink
>          Issue Type: Bug
>          Components: API / DataStream, Runtime / Coordination
>    Affects Versions: 1.16.0
>            Reporter: Robert Metzger
>            Assignee: zl
>            Priority: Major
>              Labels: pull-request-available
>
> This issue was first reported in the Flink Slack: 
> https://apache-flink.slack.com/archives/C03G7LJTS2G/p1656257678477659
> It seems that reactive mode is changing the parallelism of the 
> `ContinuousFileMonitoringFunction`, which is supposed to always run with a 
> parallelism of 1.
> This is the error
> {code}
> INITIALIZING to FAILED with failure cause: 
> java.lang.IllegalArgumentException: ContinuousFileMonitoringFunction 
> retrieved invalid state.
>       at 
> org.apache.flink.util.Preconditions.checkArgument(Preconditions.java:138)
> {code}
> You can see from the logs that the parallelism is changing on a rescale event:
> {code}
> 2022-06-27 13:38:54,979 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (1/1) (cbaad20beee908b95c9fe5c34ba76bfa) switched from 
> RUNNING to CANCELING.
> 2022-06-27 13:38:55,254 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (1/1) (cbaad20beee908b95c9fe5c34ba76bfa) switched from 
> CANCELING to CANCELED.
> 2022-06-27 13:38:55,657 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (2/2) (6ceaacbe8d9aa507b0a56c850082da8c) switched from 
> DEPLOYING to INITIALIZING.
> 2022-06-27 13:38:55,722 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (2/2) (6ceaacbe8d9aa507b0a56c850082da8c) switched from 
> INITIALIZING to RUNNING.
> 2022-06-27 13:44:54,058 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (1/2) (665b12194741744d6bba4408a252fa45) switched from 
> RUNNING to CANCELING.
> 2022-06-27 13:45:00,825 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (1/2) (3cc408fd0eb9ddfa97b22f4dfc09d8dc) switched from 
> DEPLOYING to INITIALIZING.
> 2022-06-27 13:45:00,826 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (1/2) (3cc408fd0eb9ddfa97b22f4dfc09d8dc) switched from 
> INITIALIZING to RUNNING.
> 2022-06-27 13:45:01,434 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (2/2) (79338042d84b6458c34760bc85145512) switched from 
> DEPLOYING to INITIALIZING.
> 2022-06-27 13:45:02,427 | INFO  | .executiongraph.ExecutionGraph  | Source: 
> Custom File Source (2/2) (79338042d84b6458c34760bc85145512) switched from 
> INITIALIZING to RUNNING.
> {code}



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to