shashank created CAMEL-27209:
--------------------------------

             Summary: camel-stream - a stream:http consumer with scanStream 
stops reading for good when the HTTP server cannot be reached or the connection 
is lost, also with retry=true
                 Key: CAMEL-27209
                 URL: https://issues.apache.org/jira/browse/CAMEL-27209
             Project: Camel
          Issue Type: Bug
          Components: camel-stream
            Reporter: shashank
            Assignee: shashank


h3. What happens

The {{stream}} consumer reads in one task on its own thread. With 
{{scanStream=true}} that task loops for as long as the consumer runs, and with 
{{retry=true}} it opens the stream again at its end. For {{stream:http}} an 
{{IOException}} ends the task:

* opening the URL fails (the server cannot be reached, or answers with an error 
status such as 503): at the start of a scanning consumer (it does not open the 
stream in {{doStart}}, to be lenient), or when it opens the stream again 
({{retry=true}}) after the server closed the response;
* reading a line fails (the connection is lost in the middle of a response).

The exception reaches the exception handler once (a WARN), the task ends, and 
the consumer stays started without reading anything any more, until the route 
is restarted. A server that restarts, or a network glitch, is enough. 
{{stream:file}} is not affected for a file that does not exist (it is opened 
again later), and the raw mode ({{readLine=false}}) ignores read errors but 
ends the same way when the stream cannot be opened again.

h3. How to reproduce

{{StreamHttpScanRetryTest}} (camel-stream, a minimal HTTP server): the server 
answers the first request, fails the second one with 503 (or resets the 
connection in the middle of a body) and answers again afterwards; the consumer 
never routes the data of the later answers.

h3. Fix

In scan mode a stream that cannot be opened or read is reported (the first 
failure in a row to the exception handler, the next ones at DEBUG) and treated 
as the end of the stream: with {{retry=true}} it is opened again, after at 
least one second when it failed, so a server that is down is not asked in a 
loop. Without {{retry}} the behaviour is that of a stream that ended. The 
consumer that does not scan is unchanged.

_Filed with Claude Code on behalf of allthingssecurity._



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

Reply via email to