[ 
https://issues.apache.org/jira/browse/CAMEL-21056?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Claus Ibsen updated CAMEL-21056:
--------------------------------
    Priority: Minor  (was: Major)

> Delay when using readLockMinAge
> -------------------------------
>
>                 Key: CAMEL-21056
>                 URL: https://issues.apache.org/jira/browse/CAMEL-21056
>             Project: Camel
>          Issue Type: Bug
>          Components: camel-file
>    Affects Versions: 3.20.6
>            Reporter: Stephan Balmer
>            Priority: Minor
>
> When using the file component with the readLockMinAge setting, consuming each 
> file incurs a delay the length of readLockMinAge . So when 
> `readLockMinAge=5s` and there are 1000 files, it takes 5000 seconds to 
> consume all files.
> Example route URL: 
> file:/tmp/test?readLock=changed&readLockMinAge=5s&recursive=true&delete=true
> This has two disadvantages:
>  * The maximal rate at which files are processed is 1/readLockMinAge. This 
> severly limits throughput on a route.
>  * When a route is started it will initially wait for readLockMinAge before 
> processing a file. Even if the file was written long ago.
> Camel up to version 3.20.5 processes files instantly as soon as the min age 
> is reached. Since version 3.20.6 files are not processed instantly even if 
> they are older than min age. Instead, the file component will wait for 
> readLockMinAge for each file individually.
> This regression was introduced in commit 
> [https://github.com/apache/camel/commit/5459f1da2673205a6c24041a8631ac99088b5eae
>  
> |https://github.com/apache/camel/commit/5459f1da2673205a6c24041a8631ac99088b5eae]I
>  can sympathize with the goal of that change, as I've experienced that 
> problem as well because file-managers preserve last-changed dates when 
> copying files to new places. However I do judge the importance of throughput 
> in the usual case higher than avoiding premature processing of files in the 
> edge case. I think the change should be reverted. A separate strategy or 
> option for that use-case could be introduced.



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

Reply via email to