Clustered Peoplesoft

Hi Folks,

We're currently testing clustering with Peoplesoft Enterrpsie One 8.0. We're on W2K using MSCS and Oracle 9i Rel 2 database. We've the DB on clustered servers and an App server also clustered. We've prety much identified what settings we'll be using in our TNSNAMES.ORA on our App servers and our Clients but I would be keen to hear what configs other folks are using. Our experience is that how the client "re-acts" to a failover is very much dependant on [1] What app you are using during the failover, [2] If you are actively using the app during failover and [3] How you react to the pop up screen that inform you of loss of connectivity.

Howdy CNC_Kid,

My general experience with a clustered JDE/PeopleSoft environment (ours has been live on a cluster for over two years) is that if the cluster rolls-over, kick your users out. Have the DBA(s) test the database for integrity. Have the CNCs test for all servers functioning. Have your developers or key users test the app. Then let your users back in.

Gregg Larkin
Praxair, Inc.
North American PeopleSoft
Enterprise One System Administrator

We are clustering two HP-UX machines with MC/SG. It works great in 8.9. I've tested the application during a forced failover and notice the "Waiting on Clustered Server" message pops up, and then everything returns to normal a few seconds later. I'm sure it does depend on which application you are in, and of course whether or not you are in the middle of any jdenet communication or db writes during the failover operation. More testing is needed before we get to production.

Charles

Howdy Greg,

If you are going to "stop" and performs the checks you suggest does this not really defeat the primary benefit of having a cluster configuration?

If you have actually had to perform these checks I'd be interested in hearing some more about them. Would you be willing to share more information on them backchannel?

Howdy Sean,

When we have done our failover testing in the past, whatever application the user had open at the time of the failover is broken until the user logs out and back in. There is no seemless 24/7 solution to JDE. The purpose of a clustered server is to minimize the downtime. If the node that is running your JDE enterprise code goes down, your users are hosed. Especially if you have logic servers. Once the enterprise server is back up, the logic servers have to have JDE bounced in order to re-establish connectivity. If SQL goes down, JDE crashes. Same deal as before. If SQL goes down during the day, transactions were occurring. The DBAs need to verify that the database is not corrupted and that partial transactions were rolled back.

We have no set procedure for what to do when the system goes down. What we do have is a good team that springs into action, analyzes the situation, brings it back up with integrity as quickly as possible. We also have set procedures for notifying the helpdesk and the business that the system is down and our estimate of when it will be back up.


Gregg Larkin
Praxair, Inc.
North American PeopleSoft
Enterprise One System Administrator

I was sent a question off-line that I'll share my response on-line. The question -

I have a few questions for you regarding your clustered solution. We are
also SQL Server and will be moving to a clustered solution. In fact, we are
going one step further and doing data replication to an offsite location,
which is also clustered.

We are planning on doing a cluster that contains one database server and one
application server. We have web HTML clients and FAT clients. We don't
have any Citrix servers.

I would be interested in knowing some additional information on your
clustered solution. What is clustered? What type of servers are you using?
Are you using SQL Server Enterprise? Are you using AWE?

Similar to you, we have EnterpriseOne 8.9 in our development network and are
scheduled to migrate to this version. Thanks.

My response -

What is clustered? The enterprise server. It is an active active cluster. JDE and SQL are running on one node. I have the PD databases, central objects and system running on node 1. On node two we have a second instance of SQL running the PY and DV databases. Enterprise servers are eight-way Compaqs. They attached to a SAN. The servers are in two separate buildings one mile apart. The SAN is also split between both buildings and is replicated in real time.

What type of servers are you using? Compaq servers. Win 2000, SQL 2000


Are you using SQL Server Enterprise? Yes

Are you using AWE? no, but I am in awe that our complex environment works!

The terminal servers are spilt between the two buildings.

I have several logic servers. One for processing UBEs. One processing BSFNs for the web server and the Webmethods (XPIe) servers. I have a third test logic server and a fourth logic server coming on-line later this summer for processing invoices. Again, we have those servers distributed between the two buildings.

We also replicate data to Cognos servers for Management Reporting. Data is replicated to servers in Mexico, Brazil, Italy, and Korea. We have Webmethods interfaces in development to our mainframe and Lotus Notes.

Gregg Larkin
Praxair, Inc.
North American PeopleSoft
Enterprise One System Administrator