Shibboleth.Listener errors

[Cantor, Scott]

I don't know what you're asking, but those are just informational details. It shuts down its worker threads when the select loop for a given socket detects the socket was closed.

It's not a pool on the server end. Each connection opened gets its own new worker thread with a dedicated connected socket and keeps calling select/recv for new messages until the socket closes because the the calling process shuts down or just stops making new calls.

There are no known issues with that aside from resource consumption under heavy loads when using improper agent models like Apache prefork that cause too many threads to be created. The process is fast enough most often to keep things moving without thread counts growing all that high when the number of calling processes and threads is reasonably bounded to under 50 or 100.

> ·         If the socket remains closed or is not properly re-
> established, shibd may stop listening on its configured port,
> resulting in "port dead" or unresponsive state.
>        Persistent or repeated socket resets can cause shibd to
> crash or hang, making the port unavailable until the process is
> restarted.

Says who?

I've seen SPs run for weeks and months and not exhibit any problems continuing to accept new connections and recycle the worker threads internally.

-- Scott