Here are two tests that I just performed (I had to go straight against the
servers server1.domain & server2.domain instead of against
loadbalanced.domain since we have sticky sessions).

Test 1:

1) Go to
https://server1.domain/cas/login/?service=http://server.domain/staticHtml(where
staticHtml is just a regular page that I can get the ST from)
2) Login, get redirected to staticHtml, get the ST from the query string
3) Go to
https://server2.domain/cas/serviceValidate/?service=http://server.domain/staticHtml&ticket=ST-FromTheQueryString
4) This results in getting redirected to
https://server2.domain/cas/login/?service=http://server.domain/staticHtml&ticket=ST-FromTheQueryString

In the browser, it is like server2 wants me to login even though I went to
serviceValidate

Test 2:

1) Go to
https://server1.domain/cas/login/?service=http://server.domain/staticHtml(where
staticHtml is just a regular page that I can get the ST from)
2) Login, get redirected to staticHtml, get the ST from the query string
3) Using a test ASP.NET web application and the DotNetCasClient
programmatically validate the ST retrieved in step 2.  This works and I get
the username back.

So what is the difference and why are we seeing this behavior.  Apparently I
was mistaken about getting an unrecognized ticket error, but I swear I have
gotten them before.  Our cluster configuration just seems shaky right now.
The reason I feel I need session replication is because of the documentation
here: http://www.ja-sig.org/wiki/display/CASUM/Clustering+CAS (which appears
to have been last modified by yourself Marvin, so I appreciate being able to
get help directly from you)

Thanks for your help on this issue.  I really would like to understand the
best way to get our servers clustered.



On Thu, Sep 24, 2009 at 6:56 PM, Marvin Addison <[email protected]>wrote:

> > if you get a ST from server A and try to
> > validate it on server B it returns "unrecognized ticket".
>
> Would you mind posting the relevant excerpt from the stack trace that
> results from either of your test scenarios?  I have a hunch about the
> cause of your problem, but can't be certain without further
> information.
>
> We have never seen any errors of this sort in months of testing our
> active-active setup that makes no provisions for either session
> replication or sessionless Web Flow.  Since the ticket request and
> validation steps have different sources, one would expect a 50% error
> rate in our case of a 2-node cluster if there were a fundamental
> requirement that Web Flow state be shared among nodes.  Our experience
> suggests there is no such requirement.
>
> M
>
> --
> 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
>

-- 
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