-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

About six weeks ago I started a new job and with it I inherrited a
server running CAS 3.3.5 on Tomcat 6.0.18.  The whole thing is running
behind pound 2.4.3 which provides an SSL front end so we don't have to
muck with SSL in Tomcat.  The whole thing runs on Ubuntu 9.04 (amd64)
(I've considered upgrading to 9.10 to see if that fixes my problem).

About 3 weeks ago my users started complaining that intermittently
(about 5% of the time) the connection gets interrupted before the login
web page would load.  A simple page refresh will load the page and
they're able to log in without a problem.  At the same time I started
getting errors in the pound log such as:

> Mar 18 10:08:21 login pound: (7fd1efddd950) error read from 199.8.237.239: 
> Connection timed out
> Mar 18 10:44:57 login pound: (7fd1ed622950) error read from 199.8.239.245: 
> Connection timed out
> Mar 18 10:45:04 login pound: (7fd1efd5b950) error read from 199.8.239.245: 
> Connection timed out
> Mar 18 10:47:07 login pound: (7fd1ed622950) error read from 199.8.239.245: 
> Connection timed out
> Mar 18 10:47:14 login pound: (7fd1efe1e950) error read from 199.8.239.245: 
> Connection timed out
> Mar 18 10:54:32 login pound: (7fd1ed622950) error read from 208.157.149.188: 
> Connection timed out
> Mar 18 10:54:40 login pound: (7fd1efe1e950) error read from 208.157.149.188: 
> Connection timed out
> Mar 18 10:55:38 login pound: (7fd1ed622950) error read from 208.157.149.188: 
> Connection timed out
> Mar 18 10:58:08 login pound: (7fd1efe1e950) error read from 208.157.149.188: 
> Connection timed out
> Mar 18 11:22:36 login pound: (7fd1efddd950) error read from 199.8.238.59: 
> Connection timed out
> Mar 18 12:32:59 login pound: (7fd1ed622950) error read from 199.8.239.245: 
> Connection timed out

The timed out connections never show up in the CAS log but just before
they should have shown up (according to the timestamps) there is
_usually_ an entry in the CAS log such as:

> 2010-03-18 15:59:46,719 INFO 
> [org.jasig.cas.services.DefaultServicesManagerImpl] - Reloading registered 
> services.
> 2010-03-18 15:59:46,720 INFO 
> [org.jasig.cas.services.DefaultServicesManagerImpl] - Loaded 0 services.

These two messages show up at other points in the CAS log and don't seem
to cause a problem.  I'm not using CAS's service manager so not loading
any services shouldn't be a problem.

I am unable to reliably replicate the problem but the logs indicate that
the problem reliably occurs for someone every 3-7 minutes.  The CAS logs
indicate that services are reloaded every 2-4 minutes.

I've been working with the pound mailing list and the only thing we can
figure out is that CAS stops responding for a split second every time it
reloads services and anyone lucky enough to time a request during that
interval is screwed.  So I have two questions: 1) Does this theory make
any sense and, if so, how can I prevent this from happening? and 2) What
else could cause this behavior?

Thanks for the help.

- --
David King
Goshen College ITS
[email protected]
574-535-7726

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkuiihQACgkQH+/Vg7DylXaLcgCgpeDgm3FO6l3zCEj+JvCdpsnJ
DsgAn3dNCmVCBkUYxk2KTr45dOWXfvjs
=jYPf
-----END PGP SIGNATURE-----

-- 
You are currently subscribed to [email protected] as: 
[email protected]
To unsubscribe, change settings or access archives, see 
http://www.ja-sig.org/wiki/display/JSG/cas-user

Reply via email to