Should Undertow Timeout Kill IO Thread Process? (Wildfly 8.2.0)

[mcowan]

Hi,

I am currently using Wildfly 8.2.0 and have a question around how bean processes are handled that I couldn't find documentation for.

What I am working on is adding limits to our stack to start rejecting calls when we near resource exhaustion (thinking CPU, memory etc).

We have an Nginx proxy in front of Wildfly that we are using to limit the number of concurrent requests (I am aware this can also be enforced in Undertow) and increased task-max-threads in urn:jboss:domain:io:1.1 worker.

I also wanted to limit the time a request can put load on the service, so added a read-timeout and write-timeout to urn:jboss:domain:undertow:1.2 http-listener.

This all works fine. The problem is that when the http-listener times out, the java process that it spawns is not interrupted, which is what I expected.

Clearly I have the wrong idea here but wondered how can you set a TTL for the Java process created for the http request?

Thanks for any help you can give.

[ctomc]

Thread just returns back to worker/io pool.

Threads are almost never destroyed.

[mcowan]

Thanks for the reply

So is it simply the responsibility of the servlet programmer to ensure no request lasts forever?

[ctomc]

Mike Cowan wrote:

Thanks for the reply

So is it simply the responsibility of the servlet programmer to ensure no request lasts forever?

By default there is some timeout for how long the servlet can execute before server "kills" it.

Not sure what default timeout is, but there is.

[mcowan]

I have seen threads live over 30 mins, which is a bit of a worry to me.

I did check the documentation for a worker thread timeout but didn't find one. I tried transaction timeout, session timeout and listener timeout (some out of desperation) but none seemed to kill the worker threads.

Looking at the XNIO documentation for worker threads, Im beginning to think there might not be a time out at all.