Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

apr_bucket_read() failed on weblogin server

Russ Allbery eagle at windlord.stanford.edu
Wed Feb 4 10:38:45 PST 2009


Guido Günther <agx at sigxcpu.org> writes:

> we're running Debian Etch's webauth 3.5.3-2 and everything is almost
> working fine. Sometimes out of a sudden when the user gets redirected to
> the weblogin server's login.fcgi after entering the password the dialog
> just stalls for a couple of seconds, then the login form is displayed
> again (without any error message displayed to the user). The weblogin
> server's logs then have:
>
> [Tue Feb 03 15:48:32 2009] [error] [client 192.168.0.1] (104)Connection reset by peer: ap_content_length_filter: apr_bucket_read() failed, referer: https://webauth.example.com/login/?RT=<base64encoded string>
>
> Restarting apache2 on the "application server" running the
> libapache2-webauth fixes the problem for a couple of days (often weeks)
> until it pops up again. Clients using SPNEGO to authenticate to the
> WebKDC never have any issues. Any ideas what could be causing this?

Doing some searches for that error, the problem seems most often to be
related to a CGI script not consuming all of its input before returning.
However, given that login.fcgi uses the CGI module, it really should be
quite well-behaved there, so I'm not sure what could be causing it in this
particular case.

I'm afraid I'm no help -- we're not seeing that message at all at
Stanford, although we're running a backported 3.6.0.  However, I can't
think of anything that would have changed in that version that would have
affected this.  We are, however, using mod_fastcgi.  Are you, or are you
running it as a traditional CGI script?

-- 
Russ Allbery <eagle at windlord.stanford.edu>
Technical Lead, ITS UNIX Systems and Applications, Stanford University



More information about the webauth-info mailing list