[ 
https://issues.apache.org/jira/browse/CAMEL-25447?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Claus Ibsen resolved CAMEL-25447.
---------------------------------
    Resolution: Fixed

Fixed by https://github.com/apache/camel/pull/29727 (merged as c9c4f3fd7e78 on 
main, 4.24.0).

_Claude Code on behalf of davsclaus_

> camel-cli-connector - answer actions in about 100 ms instead of once a second
> -----------------------------------------------------------------------------
>
>                 Key: CAMEL-25447
>                 URL: https://issues.apache.org/jira/browse/CAMEL-25447
>             Project: Camel
>          Issue Type: Improvement
>          Components: camel-jbang
>            Reporter: Claus Ibsen
>            Assignee: Claus Ibsen
>            Priority: Minor
>             Fix For: 4.24.0
>
>
> The file transport of camel-cli-connector looks for action files once per 
> loop, and the loop runs every second (fixed delay 1000 ms) after writing the 
> status file. So every action to a running integration waits up to a second 
> before the integration sees it, whatever the client: camel get / camel cmd, 
> the camel-jbang monitor, the MCP server, camel ask.
> Measured (CAMEL-25438): an action takes about 1 s whatever its size 
> (thread-dump, bean, startup-recorder, type-converters of 41 KB); the status 
> file is rewritten every ~1015 ms.
> Proposal: check for action files on a schedule of its own, about every 100 
> ms, on the same single connector thread, and keep the status file (and the 
> trace task) at its current interval. Debug mode already runs the whole loop 
> at 100 ms.
> * Cost: a scan of ~/.camel for <pid>-action*.json, about 0.04 ms with 127 
> files there: negligible at 10 per second. No extra thread, no TCP port 
> (unlike the websocket transport).
> * Gain: an action drops from ~1 s to ~0.1 s for every client.
> * Care: the connector runs inside every integration that has it (Camel Main, 
> Spring Boot, Quarkus), so this is a runtime change. A slow action (jfr, up to 
> 30 s) still holds up the loop, as today.



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to