You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@nifi.apache.org by "Gábor Gyimesi (Jira)" <ji...@apache.org> on 2022/11/16 13:46:00 UTC

[jira] [Commented] (MINIFICPP-1987) Configuring a metrics for Processors for C2 Protocol

    [ https://issues.apache.org/jira/browse/MINIFICPP-1987?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17634848#comment-17634848 ] 

Gábor Gyimesi commented on MINIFICPP-1987:
------------------------------------------

We could implement a solution to use a regex filter in the C2 properties for defining which processor metrics we want to publish. The filter would something look like: processors/.* just to specify that this filter refers only to the processor metrics.

> Configuring a metrics for Processors for C2 Protocol
> ----------------------------------------------------
>
>                 Key: MINIFICPP-1987
>                 URL: https://issues.apache.org/jira/browse/MINIFICPP-1987
>             Project: Apache NiFi MiNiFi C++
>          Issue Type: Improvement
>    Affects Versions: 0.12.0
>            Reporter: Kondakov Artem
>            Priority: Major
>             Fix For: 0.13.0
>
>
> Hello.
> I tried to setup receiving metrics using the C2 protocol for my processors.
> The documentation describes that you must explicitly set the metric classes, for example like this
> nifi.c2.root.class.definitions.metrics.metrics.processorMetrics.classes=ListenHTTPMetrics,PostElasticsearchMetrics,InvokeHTTPMetrics,BinFilesMetrics,MergeContentMetrics...
> It turns out a complex configuration, if I want to receive data on all my processors, I need to update the list of these classes every time if I add a new or remove type of processor.
> If I try to configure classes for all possible processors, then I get 50+ errors in the logs, , because not all processor types are involved, for example
> "No metric defined for PublishKafkaMetrics"
> "No metric defined for ConsumeKafkaMetrics"
> etc...
> It seems to me that it is worth making one universal metric class for all processes, or lower the error logging level when searching for a metric class to warning



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