NativeSPClustering- shared shibd

[Vedran Bartonicek]

>>Yes, and then it detects the failures and throws them out of the pool and
establishes new ones.

Thanks for the comment!
Did a bit more testing, and it appears to behave as you described-  after
shibd restart the pool seems to be gradually renewed.
As I try more times to connect, errors appear less frequently, then they
are gone after about 10 or so attempts.

On the bad side, those errors, when happening, they propagate to the user.
I suppose it is not possible to tune connection policy - maybe some kind of
polling to detect dead connections before the user gets the error?

-Vedran






On 12 September 2016 at 16:35, Cantor, Scott <cantor.2 at osu.edu> wrote:

> > I have setup clustering so that shibd runs on separate machine, and httpd
> > with mod_shib on separate machine, using TCP listener plugin.
> > When shibd is restarted, httpd must also be restarted, otherwise error
> > appears - like below from :
>
> Nobody has ever reported anything like that, and since the connection
> management is the same for TCP as for domain sockets and the Windows code
> uses TCP, and from the point of view of the client socket, localhost and a
> remote host are handled identically...no, I don't really buy that.
> Something else is wrong in your environment.
>
> > It looks like httpd tries to reuse dead connections to connect to shibd
> > (guessing here)
>
> Yes, and then it detects the failures and throws them out of the pool and
> establishes new ones.
>
> -- Scott
>
>
> --
> To unsubscribe from this list send an email to
> users-unsubscribe at shibboleth.net
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://shibboleth.net/pipermail/users/attachments/20160913/4af5e436/attachment-0001.html>