[
https://issues.apache.org/jira/browse/NIFI-1329?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15192428#comment-15192428
]
ASF subversion and git services commented on NIFI-1329:
-------------------------------------------------------
Commit 0cb4047adde5f82a1554fa13431909c5e5eb4f25 in nifi's branch
refs/heads/master from [~boardm26]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=0cb4047 ]
NIFI-627 removed flowfile penalization which could skew behavior when
processor's Time Duration was less than Penalty Duration, improved over
throttle penalization NIFI-990 corrected failure path NIFI-1329 refactored
using FlowFileFilter to avoid repeatedly returning flowfiles to the input
queue, producing misleading stats and excessive Tasks/Time used
Signed-off-by: joewitt <joew...@apache.org>
> ControlRate processor should not be penalizing FlowFiles
> --------------------------------------------------------
>
> Key: NIFI-1329
> URL: https://issues.apache.org/jira/browse/NIFI-1329
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Affects Versions: 0.4.1
> Reporter: Matthew Clarke
> Assignee: Michael Moser
> Priority: Minor
> Fix For: 0.6.0
>
>
> The controlRate processor currently penalizes all FlowFiles still on input
> queue if flow rate has been reached. Those penalized FlowFiles will not be
> processed by the ControlRate processor again until that penalization has
> expired causing an impact on the desired controlled throughput of the
> processor.
> Example ControlRate processor setup:
> Penalty duration left at default 30 sec
> Strategy set to FlowFile count
> Max rate set at 100
> time duration set to 1 sec
> Let say the input queue to this processor has 2000 files dumped in it. The
> ControlRate processor will read all 2000 files, allow 100 through and
> penalize the other 1900 fro 30 seconds. Assuming no other files enter queue
> at this point, after 30 seconds another 100 will pass and 1800 will be
> penalized again. This pattern continues....
> You can work around this defect by changing the penalization default from 30
> to say 1 sec.
> Why is the "In" stat reflecting reading all files on input queue when only
> 100 files are actually passing through processor? <-- this leads to
> confusion for the users.
> Why is penalty being applied here?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
