https://bz.apache.org/bugzilla/show_bug.cgi?id=64080

--- Comment #9 from Mark Thomas <ma...@apache.org> ---
(In reply to carbattles from comment #8)
> > 'allow requests to complete if
> > possible' is a more invasive change and is not consistent with,
> > historically, the behaviour Tomcat users have asked for on shutdown. 
> 
> Isn't this what unloadDelay is for? My testing shows, this is exactly what
> it does, it provides time for requests that have reached the servlet to
> finish processing. If the servlets do not finish their processing, and the
> time runs out, the in-flight requests are terminated.

No. unloadDelay applies only once the request has been passed to the Servlet.

There are opportunities to respond with a 503 before the request is passed
to the Servlet and this is what Tomcat does because that is what users have
previously requested.

> I also think this bug is being over-compilicated. I'm not asking for new
> functionality,

Yes you are.

> just the shutdown to also respect connections that have
> already been established (but not yet reached the servlets) and allow them
> time to complete based on unloadDelay.

That *is* new behaviour and runs contrary to the current design. That is what
makes the change more invasive / complicated / risky / etc.

> If it is any help, our setup is 12 tomcat servers, with load balancer in
> front. Our build process will restart the tomcat servers in sequence. We
> expect that not a single request is dropped, however may process slower.
> This should have been fixed in bug 62515, but as I understand Mark, a new
> bug was introduced with the introduction of non blocking IO.

The bug has always existed. Non-blocking I/O just makes it more likely the
connection is closed immediately rather than a 503 response being written
and then the connection being closed.

-- 
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

Reply via email to