Login stops working until cookies are cleared

11 messages Options
Embed this post
Permalink
masonoise () Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
We've been having a strange problem, which still remains after upgrading to the latest 3.3.1 version. It doesn't happen all the time, but every so often one of us will try to login to our Plone installation and the login page will just reload after entering name/password. There's no error or warning, no sign of what's going on; the login just silently fails and sends the user back to the login page again. If the user goes into their browser and clears the cookies used by Plone (__k*), then the login will work again for a while. Then after a day or two, or sometimes longer, it will stop working again until cookies are cleared.

Has anyone else encountered this or have any ideas about how to fix it? It's driving everyone crazy, understandably.

Thanks for any suggestions.
Dieter Maurer () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
masonoise wrote at 2009-11-5 09:07 -0800:

>
>We've been having a strange problem, which still remains after upgrading to
>the latest 3.3.1 version. It doesn't happen all the time, but every so often
>one of us will try to login to our Plone installation and the login page
>will just reload after entering name/password. There's no error or warning,
>no sign of what's going on; the login just silently fails and sends the user
>back to the login page again. If the user goes into their browser and clears
>the cookies used by Plone (__k*), then the login will work again for a
>while. Then after a day or two, or sometimes longer, it will stop working
>again until cookies are cleared.
>
>Has anyone else encountered this or have any ideas about how to fix it? It's
>driving everyone crazy, understandably.

I have not encountered such a situation.

Your description suggests that there is some cookie which contains
a problematic value that is not reset properly.

PlonePAS usually stores the information about a successful authentication
via its so called "session plugin". Despite its name, this plugin
does not use the normal Plone session object but instead uses
its own cookie, properly secured, to indicate a successful authentication.
I had to fight a bit with this plugin -- therefore, I do not fully trust it.
Maybe, you check whether its cookie uses a name you must manually clear.

Otherwise, you can make a larger search over the Plone sources
to locate this cookie. When you find it, this will give you
a valuable hint about the component causing your problem.

And as usual: debugging may allow to finally understand the problem -- although
it may be difficult to debug a productive instance.



--
Dieter

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
PriusGeek () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
I have encountered this on a 3.2.1 installation. I have a corporate
internet site and multiple plone customer portals running on the same
machine. The internet site is accessed via http (http://
www.domain.com) and the plone site via https (using VHM,
https://www.domain.com/plonesite) to the same machine. The internet
site is tied into google analytics via some plugins from
Salesforce.com. (note that everything worked without a problem until I
implemented Salesforce.com web lead generation on the internet site)

There is a page on the Internet site that allows users to select their
Plone portal and link to it for login. What I think was happening was
a cookie conflict between the Google cookies and the Plone login
cookies. It appeared that it happened if the referring domain from the
Internet site was the same as the Plone site. If I put in the URL from
a machine outside of the hosting domain, I think it worked OK. What I
found was that if I deleted the google cookies, plone worked fine,
until I tried it a second time.

I have no idea exactly what was going on, but I was able to work
around the problem by creating a CNAME in DNS for the plone site with
a different domain name from the internet site. This sounds
convoluted, but basically, http://www.domain.com and https://domain.com
both pointed to the same box. I created a new domain called domain1
and pointed it to the same machine. If there were any links on the
internet site that linked to the plone site, I forced it to use the
https://domain1.com form and the problem went away. In my case, it
appeared to have something to do with Google analytics and the
referring site/url being in the same domain as the plone site. As soon
as I made them different, the problem went away.

You probably have nothing like this, but maybe this will help trigger
something similar. If you have google analytics and/or Salesforce.com
web lead integration running, try turning it off and see if the
problem changes.

Good Luck!

Gray



On Nov 9, 1:41 am, "Dieter Maurer" <[hidden email]> wrote:

> masonoise wrote at 2009-11-5 09:07 -0800:
>
>
>
> >We've been having a strange problem, which still remains after upgrading to
> >the latest 3.3.1 version. It doesn't happen all the time, but every so often
> >one of us will try to login to our Plone installation and the login page
> >will just reload after entering name/password. There's no error or warning,
> >no sign of what's going on; the login just silently fails and sends the user
> >back to the login page again. If the user goes into their browser and clears
> >the cookies used by Plone (__k*), then the login will work again for a
> >while. Then after a day or two, or sometimes longer, it will stop working
> >again until cookies are cleared.
>
> >Has anyone else encountered this or have any ideas about how to fix it? It's
> >driving everyone crazy, understandably.
>
> I have not encountered such a situation.
>
> Your description suggests that there is some cookie which contains
> a problematic value that is not reset properly.
>
> PlonePAS usually stores the information about a successful authentication
> via its so called "session plugin". Despite its name, this plugin
> does not use the normal Plone session object but instead uses
> its own cookie, properly secured, to indicate a successful authentication.
> I had to fight a bit with this plugin -- therefore, I do not fully trust it.
> Maybe, you check whether its cookie uses a name you must manually clear.
>
> Otherwise, you can make a larger search over the Plone sources
> to locate this cookie. When you find it, this will give you
> a valuable hint about the component causing your problem.
>
> And as usual: debugging may allow to finally understand the problem -- although
> it may be difficult to debug a productive instance.
>
> --
> Dieter
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Plone-Users mailing list
> [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Steven H () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
In reply to this post by masonoise
Hi

We recently rebuilt part of our Plone 3.1.2 cluster under Debian Lenny
64 bit and used the Debian supplied Python. This has been showing a
similar cookie-related issue with failed logins. The other systems,
running Debian Etch 32 bit and a local Python, have not.

I'm in the process of investigating. The issue seemed to be related to
a WebTrends tracking cookie (WT_FPC) with an almost 10 year life. The
cookie was being attached to a higher level domain (le.ac.uk, while
Plone is at www2.le.ac.uk) by another web server. The cookie doesn't
always cause the failure, but every failure I've investigated had the
cookie present and deleting it cleared the problem.

Steven





------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Dieter Maurer () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
Steven H wrote at 2009-11-10 04:06 -0800:

> ...
>We recently rebuilt part of our Plone 3.1.2 cluster under Debian Lenny
>64 bit and used the Debian supplied Python. This has been showing a
>similar cookie-related issue with failed logins. The other systems,
>running Debian Etch 32 bit and a local Python, have not.
>
>I'm in the process of investigating. The issue seemed to be related to
>a WebTrends tracking cookie (WT_FPC) with an almost 10 year life. The
>cookie was being attached to a higher level domain (le.ac.uk, while
>Plone is at www2.le.ac.uk) by another web server. The cookie doesn't
>always cause the failure, but every failure I've investigated had the
>cookie present and deleting it cleared the problem.

As it is unlikely that Plone uses a cookie of this name,
this would indicate a severe cookie bug in either Zope or the browser.
I would try to check the "cookies" request header (using either
a TCP logger or "firebug") when the problem occurs to find out
which cookie information the browser sends and whether this is wrong
or interpreted badly by Zope.



--
Dieter

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Steven H () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
I have now seen cases where the WebTrends tracking cookie (WT_FPC) was
not present, but Google Analytics tracking cookies were. Removing the
Google cookies made the problem go away.

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Dieter Maurer () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
Steven H wrote at 2009-11-12 02:07 -0800:
>I have now seen cases where the WebTrends tracking cookie (WT_FPC) was
>not present, but Google Analytics tracking cookies were. Removing the
>Google cookies made the problem go away.

A long while back, an Apache cookie problem had caused similar
effects (aparently, Apache lost some cookies for requests containing
too many cookie definitions). I do not expect that you see
now the Apache problem (because the bug in Apache was fixed long ago)
but maybe, the cause is similar (cookie information lost somewhere
under some conditions). You may try to determine the exact requests
arriving at Plone by means of a tcp logger (e.g. "etherreal", "wireshark",
"tcpdump", ...) and verify whether the problem is in the cookie definitions
arriving at Plone.


--
Dieter

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Steven H () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
Thanks Dieter for the helpful hints.

I'm in the middle of a debug session where I've managed to capture
this bug. So far it looks as if the  HTTP_COOKIES header is sometimes
missing from the request. I'm trying to determine if it's going
missing in early ZOPE processing or before ZOPE in Apache, Squid or
Pound.

It would be worth comparing versions - we're using Debian Lenny 64 bit
with the packaged versions of Apache 2 (2.2.9-10+lenny4), Squid
(2.7.STABLE3-4.1) and Pound (2.4.3-1) except that Squid has a patch to
prevent it hanging with certain HTTP headers.

Steven

On Nov 14, 7:24 am, "Dieter Maurer" <[hidden email]> wrote:

> Steven H wrote at 2009-11-12 02:07 -0800:
>
> >I have now seen cases where the WebTrends tracking cookie (WT_FPC) was
> >not present, but Google Analytics tracking cookies were. Removing the
> >Google cookies made the problem go away.
>
> A long while back, an Apache cookie problem had caused similar
> effects (aparently, Apache lost some cookies for requests containing
> too many cookie definitions). I do not expect that you see
> now the Apache problem (because the bug in Apache was fixed long ago)
> but maybe, the cause is similar (cookie information lost somewhere
> under some conditions). You may try to determine the exact requests
> arriving at Plone by means of a tcp logger (e.g. "etherreal", "wireshark",
> "tcpdump", ...) and verify whether the problem is in the cookie definitions
> arriving at Plone.
>
> --
> Dieter
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Plone-Users mailing list
> [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Steven H () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink

I this case, it looks as if it's pound that's not sending the cookie
header through.

Steven

On Nov 18, 3:51 pm, Steven H <[hidden email]> wrote:

> Thanks Dieter for the helpful hints.
>
> I'm in the middle of a debug session where I've managed to capture
> this bug. So far it looks as if the  HTTP_COOKIES header is sometimes
> missing from the request. I'm trying to determine if it's going
> missing in early ZOPE processing or before ZOPE in Apache, Squid or
> Pound.
>
> It would be worth comparing versions - we're using Debian Lenny 64 bit
> with the packaged versions of Apache 2 (2.2.9-10+lenny4), Squid
> (2.7.STABLE3-4.1) and Pound (2.4.3-1) except that Squid has a patch to
> prevent it hanging with certain HTTP headers.
>
> Steven
>
> On Nov 14, 7:24 am, "Dieter Maurer" <[hidden email]> wrote:
>
>
>
> > Steven H wrote at 2009-11-12 02:07 -0800:
>
> > >I have now seen cases where the WebTrends tracking cookie (WT_FPC) was
> > >not present, but Google Analytics tracking cookies were. Removing the
> > >Google cookies made the problem go away.
>
> > A long while back, an Apache cookie problem had caused similar
> > effects (aparently, Apache lost some cookies for requests containing
> > too many cookie definitions). I do not expect that you see
> > now the Apache problem (because the bug in Apache was fixed long ago)
> > but maybe, the cause is similar (cookie information lost somewhere
> > under some conditions). You may try to determine the exact requests
> > arriving at Plone by means of a tcp logger (e.g. "etherreal", "wireshark",
> > "tcpdump", ...) and verify whether the problem is in the cookie definitions
> > arriving at Plone.
>
> > --
> > Dieter
>
> > ------------------------------------------------------------------------------
> > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> > trial. Simplify your report design, integration and deployment - and focus on
> > what you do best, core application coding. Discover what's new with
> > Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> > _______________________________________________
> > Plone-Users mailing list
> > [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Plone-Users mailing list
> [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Steven H () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink

The version of pound we were using has a reduced buffer size from
earlier versions (1024 rather than 2048) - see discussion at
http://www.apsis.ch/pound/pound_list/archive/2008/2008-02/1203715523000/index_html?fullMode=1#1203952046000

When the cookies header became too big, it was no longer forwarded to
Plone.

A configure option allows the buffer size to be adjusted (we have set
it to 4096). When the package is rebuilt, all works as it should.

Clearly, the more tracking cookies that are in use the more likely you
are to exceed the limit. We had Google and Yahoo tracking cookies in
active use, plus WebTrends from some old configuration, and a cookie
to handle Google Search Appliance integration. Plone authentication
and cut and paste add  further cookies.

Steven

On Nov 19, 9:47 am, Steven H <[hidden email]> wrote:

> I this case, it looks as if it's pound that's not sending the cookie
> header through.
>
> Steven
>
> On Nov 18, 3:51 pm, Steven H <[hidden email]> wrote:
>
>
>
> > Thanks Dieter for the helpful hints.
>
> > I'm in the middle of a debug session where I've managed to capture
> > this bug. So far it looks as if the  HTTP_COOKIES header is sometimes
> > missing from the request. I'm trying to determine if it's going
> > missing in early ZOPE processing or before ZOPE in Apache, Squid or
> > Pound.
>
> > It would be worth comparing versions - we're using Debian Lenny 64 bit
> > with the packaged versions of Apache 2 (2.2.9-10+lenny4), Squid
> > (2.7.STABLE3-4.1) and Pound (2.4.3-1) except that Squid has a patch to
> > prevent it hanging with certain HTTP headers.
>
> > Steven
>
> > On Nov 14, 7:24 am, "Dieter Maurer" <[hidden email]> wrote:
>
> > > Steven H wrote at 2009-11-12 02:07 -0800:
>
> > > >I have now seen cases where the WebTrends tracking cookie (WT_FPC) was
> > > >not present, but Google Analytics tracking cookies were. Removing the
> > > >Google cookies made the problem go away.
>
> > > A long while back, an Apache cookie problem had caused similar
> > > effects (aparently, Apache lost some cookies for requests containing
> > > too many cookie definitions). I do not expect that you see
> > > now the Apache problem (because the bug in Apache was fixed long ago)
> > > but maybe, the cause is similar (cookie information lost somewhere
> > > under some conditions). You may try to determine the exact requests
> > > arriving at Plone by means of a tcp logger (e.g. "etherreal", "wireshark",
> > > "tcpdump", ...) and verify whether the problem is in the cookie definitions
> > > arriving at Plone.
>
> > > --
> > > Dieter
>
> > > ------------------------------------------------------------------------------
> > > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> > > trial. Simplify your report design, integration and deployment - and focus on
> > > what you do best, core application coding. Discover what's new with
> > > Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> > > _______________________________________________
> > > Plone-Users mailing list
> > > [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users
>
> > ------------------------------------------------------------------------------
> > Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> > trial. Simplify your report design, integration and deployment - and focus on
> > what you do best, core application coding. Discover what's new with
> > Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> > _______________________________________________
> > Plone-Users mailing list
> > [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
> trial. Simplify your report design, integration and deployment - and focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Plone-Users mailing list
> [hidden email]://lists.sourceforge.net/lists/listinfo/plone-users

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users
Dieter Maurer () Re: Login stops working until cookies are cleared
Reply Threaded More More options
Print post
Permalink
In reply to this post by Steven H
Steven H wrote at 2009-11-18 07:51 -0800:
>Thanks Dieter for the helpful hints.
>
>I'm in the middle of a debug session where I've managed to capture
>this bug. So far it looks as if the  HTTP_COOKIES header is sometimes
>missing from the request. I'm trying to determine if it's going
>missing in early ZOPE processing or before ZOPE in Apache, Squid or
>Pound.

The correct name in "REQUEST.environ" is "HTTP_COOKIE" (without "S").
It is unlikely, that Zope loses this special entry unless
it loses all other "HTTP_*" entries as well, because "HTTP_COOKIE"
is not treated specially from other HTTP request headers -- only
at a later stage, "HTTP_COOKIE" is translated into "request.cookies".



--
Dieter

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day
trial. Simplify your report design, integration and deployment - and focus on
what you do best, core application coding. Discover what's new with
Crystal Reports now.  http://p.sf.net/sfu/bobj-july
_______________________________________________
Plone-Users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/plone-users