Portal Failover Problem
Hi,
I've been trying to setup a multi-instance, weblogic clustered failover environment without any success for the last couple of days. Let me give you a bit of a background on the setup. I have a clustered weblogic setup in Solaris on which I have installed the portal server and the access manager server on the individual managed servers. After making a few configuration changes I was able to access the individual portal server and was able to login too.
After this I put a SunONE webserver server as a load balancer in front of the weblogic server and tried to access the portal server now through this load balancer. I was able to get to the anon user desktop and was able to login too. Till here everything is fine, the problem starts when I try to perform failover on this environment. For example, if I logged in to Portal 1 and then try to go to the other portal server I'm being asked to authenticate again(I've got the Failover setting turned on in the AMConfig's of both the portal servers).
Can anyone please let me know how failover works with portal server/access manager server?
[1133 byte] By [
howudoin] at [2007-11-25 20:07:34]
# 1
a) which version are you using ?
b) AM which holds the PS session doesn't use the HTTP session failover meachnisms of the appserver. It uses its own concept. As far as I remember there are also some persistent links from AM to the directory server which are cut off in case of an AM failure.
c) To my mind the AM failover for PS 6.3.x (I don't know what is planned for PS 7.x) uses the HADB but how this works together with Weblogic .. ?
/u
# 2
> a) which version are you using ?
> b) AM which holds the PS session doesn't use the HTTP
> session failover meachnisms of the appserver. It uses
> its own concept. As far as I remember there are also
> some persistent links from AM to the directory server
> which are cut off in case of an AM failure.
> c) To my mind the AM failover for PS 6.3.x (I don't
> know what is planned for PS 7.x) uses the HADB but
> how this works together with Weblogic .. ?
> /u
Hi,
Thanks for your response. I've now been able to perform user authentication failover, that is, when a user is logged into to one portal/am server and that server goes down, the user is automatically logged into the other portal/am server and is not asked to re-authenticate. I had to set the persistance cookie in the amconsole to get this done.
The problem I have now is that the session does not get carried over to the other portal instance. I've enabled session replication in the backend application server (weblogic 8.1) and session replication on the application server works. But when I try to set the session in the portal context and then retrieve the session from the new server, I can't do that.
I'm using SJES Portal Server 2005 Q2 on Weblogic Server 8.1.
Let me know if you can think of a way to make this work.
Cheers
# 3
> the user is automatically logged into the other portal/am server and is not asked to re-authenticate.
ok, then the second instance accepts the AM cookie
> The problem I have now is that the session does not get carried over to the other portal instance.
yes
> I've enabled session replication in the backend application server (weblogic 8.1) and session replication on the application server works.
So the http session is transferred - not more
> But when I try to set the session in the portal context and then retrieve the session from the new server, I can't do that.
The Portal session or better to say the AM(Identity Manager) session is independent from the Http session and so there is no chance to use Http-session failover of the underlying container :-(
> I'm using SJES Portal Server 2005 Q2 on Weblogic Server 8.1.
I don't know how this will be handled in the upcoming 7.x release .. hopefully in a different way
> Let me know if you can think of a way to make this work.
As far as I know with the 6.x release there is no chance for session failover, only service failover is possible, means login again and continue work but the original session is lost ..
I know this is not want you want to read .. sorry ..
/ulf
# 4
> As far as I know with the 6.x release there is no
> chance for session failover, only service failover is
> possible, means login again and continue work but the
> original session is lost ..
> I know this is not want you want to read .. sorry ..
> /ulf
Is this still the case with JES 2005Q4 if we install Portal (6) and Access Manager on Sun App Server EE 8.1 with HADB?
Greg
# 5
Just so you know, the "persistent cookie" feature is not really a solution for session failover. This feature is put in place as a sort of "remember me" function so that users don't need to log in again. If the user clicks logout, the persistent cookie is expired and another user may log in again. However, if the user does not log out then it simply appears as if their session never times out. The persistent cookie functionality is really separate from failover issues, although I can see where enabling the feature can make it "appear" as though it helps your problem.
- Jim
# 6
Since AM6.3 you can use AMs deployment container independent session failover mechanism which uses SJS Message Queue and Berkley DB - see AM docs.
