https://bz.apache.org/bugzilla/show_bug.cgi?id=64080
Mark Thomas <ma...@apache.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Severity|normal |enhancement
OS| |All
--- Comment #1 from Mark Thomas <ma...@apache.org> ---
I'll note that:
- there is a similar situation with requests in the keep-alive state;
- the issue was introduced largely as a result of the switch from blocking IO
to non-blocking IO.
Are you expecting Tomcat to process these requests normally (within some
configured timeout or set of timeouts for each stage) or is cleanly returning a
503 an acceptable response?
Looking at the existing code, processing the requests normally is a much more
invasive change that returning a 503.
A likely issue with a 503 is that clients won't read it until they have sent
the entire request so the Connector's maxSwallowSize configuration becomes a
factor as well.
It looks like changes would be required;
- in the Processor.service() implementations to always process the request on
the first iteration around the processing loop
- in Endpoint.stopInternal() to add a delay / wait between pause() (which stops
the Acceptor) and stopping the Poller (NIO/APR) / closing the active
connections (NIO2)
I'm moving this to an enhancement
--
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
