Web Timeout Issue

Hello All
We have a very weird issue and would like to see if someone here can help us resolve it.

Since last couple of weeks, our users are getting kicked out of JDE in the middle of transactions. The users get two different messages when they get logged off
1. They get a dialog box which says "Your session has expired. Please refresh your browser to login again."
2. They also get a message in the middle of IE window "You have beeb logged out of this application. The JD Edwards EnterpriseOne session for this application has timed out. If you were in the middle of processing a transaction that transaction will need to be reentered. To log back in, please click Sign Out link, or close this window and restart your application in a new browser."

We are on EnterpriseOne 8.12, 8.97.1.1 WAS 6.0.2.13
Windows 2003 SP2, SQL Server 2005 SP2

We have a case open with ORacle but they are stumped too. We have tried changing our IBM HTTP timeout settings to 120 Min, IBM App server/ Enterprise Application timeout settings to 60 min and JAS timeout setting to be 120 Min.

We have also changed the user's IE settings to allow session cookies.

Has anyone else experienced this issue? If yes, can you help out with the resolution?

Any suggestions/ help would be most welcome. I am attaching our JAS.INI with this post.

Thanks

Try and see the comparison report (between my jas.ini and yours).
Maybe you find something

Hello,
Could you send the following files:
jas.ini/jdbj.ini/http.conf
Also, print screen from the following at WAS

Hi
please find all the things you requested in the ZIP file.

Thanks
Jaya

Thanks, I looked through it and the sections are up and down but the settings look almost same.

Any chance you're using DSI? I see this sometimes when DSI decides not to play nice with JDE. If you shutdown DSI, JDE is all better - for a while u ntil DSI decides not to play nice again.

If you don't have DSI do you have any kind of third party interface similar that may not be getting along with JDE at times?

HI...we dont have DSI but we do have two web servers which are in DNS round robin with a generic site like
https://erp.XXXXXXXX.com/jde/e1menuf.maf to connect to JDE.

Do you think DNS Round robin might be affecting this?

I would be worth trying it without it just to elminiate it as a variable.

Round robin is a bad idea for any kind of app server (HTTP, java app server, db, etc.). Better to use a VIP and a sticky session. This is likely the problem. JAS will report it as a timeout because the session cache is usually not configured for replication between WAS instances. The timeout message is a red herring. It is just not finding the session cache. The login is valid because the id / pwd are cached at the browser level.

FWIW, there are 2 places to set the timeout in Websphere: at the app server level (the container) and at the web module level. Be sure that you are changing it in both places.

Hi

We are also experiencing the same issue and use round robin on our OAS servers. Did you try this without it enabled and is it any better? The RL is stumped with our issue and the users are getting very frustrated.

Thanks
Kieran
E812 8.96.2.0

Kieran
In our case the issue turned out to be the DNS round robin because it doesnt hold session informations. As mentioned in one of the above posts, the users were getting logged off because the session id was getting lost. We are now trying to find other solutions which hold the session information.

We tried to get Oracle to answer why when they support Windows DNS, we were having issues but we were informed that its outside the scope of GSC.

You might want to look at network appliances or IBM Edge server or VIP as suggested in other posts.

We are having the exact same issue and we had a consultant here for three days with no progress. We've had a call open with Oracle since 10/17/08 and they are stumped as well. We are currently going through Open Enrollment for 19,000 users and they are not happy. We also notice we're getting various errors at random with this issue.

1. Big Red Letters saying 'No response from Application Web Server'.
2. A yellow bar across the screen saying 'You have been logged out of this application'.
3. A pop up that says 'Request could not be processed HTTP error: 503: Service Unavailable Please contact your System Administrator'.
4. A pop up that says 'Your Session has expired. Please refresh your browser to login again'.

We have WebCache set up to load balance our OAS instances. We are using round robin as well but also have VIPs and sticky sessions configured. We have plenty of OAS instances to handle the traffic and our Webcache servers can handle 1400 sessions coming in. We have only hit 1000 users once in the past week. We made sure the IE settings were correct. One thing that's special about us is we use SSL.

We do use DClink but I'm not sure how that would affect our JAS sessions. We only use it in one area for barcoding.

Any ideas?

Thanks,
Lacey
8.10 Tools 8.97.2.0

Why spend money on a round robin load balancer? A simple ASP script will do just fine.

I attached the one I use.

How can you have round robin DNS with a VIP and OAS load-balancing? That is illogical if not impossible. The purpose of the VIP is to create a single IP for 2 (or more) independent interfaces.

We built ours to be simple. There is 1 address that users connect to which checks the table to see which servers are online and which was the last picked. It then redirects the user to the next server. User sessions are not moved between servers.

Just a basic round robin to pick the server. Works for us and is free.

The original poster was using round robin DNS which is not the same as random redirection via script (which is what it sounds like you have). The original poster was having problems persisting a JAS session. This was the result of forcing DNS to resolve the same v-host onto multiple servers.

Definition of Sticky Session

My guess is that unless you are working with a really cheap switch, you too could run a VIP and Apache load-balancing (which is free too).

Understood.

Just saying that sometimes a simple setup might be less headache to maintain.

All depends on your requirements.

So was the issue related to round robin? We are setting up in a similar fashion and are experiencing the exact same symptoms. However when we login directly to the web server we still are facing the same timeouts which tells me that it is not DNS related.
Any help on this will be very appreciated

I am not sure I understand your comment about "logging into the server directly". You are not proxying thru a DNS server.

For any given web app, you have to make the HTTP request with your browser and get a DNS resolution to the V-Host name. If that V-Host is in your local DNS cache (ipconfig /displaydns), then that V-Host has been resolved previously. That V-Host corresponds to an IP which is where your HTTP request will be sent. If for some reason, your local DNS cache is updated (ie. TTL expiration), you may not get a DNS resolution or you may not get the same IP for the very same V-Host from the same DNS server. Hence, round-robin DNS. If the V-Host is resolved to a new and different IP, the second/other web server isn't going to know anything about what you were doing on the first. The only exception is if there is session replication configured on the web app server side _and_ the business application supports session failover. My experience was that JDE JAS once supported session replication. It presently does not.

If you have a VIP, OTOH, and you can configure DNS to have a fixed A record for the VIP IP, you will always get the very same IP address. The switch recognizes your session identifier and directs your request to the server associated with the related server in its session database. This effectively persists your HTTP session to the same HTTP server on the same box indefinitely.

I agree with what you say. However to keep it simple we connect directly to the web server rather than going to the DNS round robin. Say for e.g you have jas1 and jas2 as your web server and the users connect to jdejas and under the covers it resolves to jas1 and jas2 via dns round robin.
What I meant was we logged directly into jas1 and jas2 directly and still got the timeouts.
Hope this helps.