Service Desk Practitioners Forum

Clustered and Single Servers connecting to One DB instance

Go to solution
Super Contributor.

Clustered and Single Servers connecting to One DB instance

Hi All,

Our environment has recently changed from a clustered (3 app servers with load balancing and fail-over) setup to this same cluster, plus 2 seperate app servers in a DMZ pointing to the same DB instance as our internal cluster.

We have internal people using our internal cluster, and external people using the 2 standalone application servers in the DMZ.

Our expected result was that internal people would be load-balanced between the 3 internal clustered app servers ONLY and the external people would try to connect to 1 of the app servers in the DMZ, and if it failed, they would manually connect to the other.

What we are experiencing is messages saying that the application server has failed, as if the "hp service desk server" service is not running. Both internal and external people are getting this message, when refreshing their session or when logging on. What seems to be happening is that attempts are being made to switch internal people onto the external servers and vice versa. The firewall is not allowing this to happen therefore the SD client is reporting the message that the application server has failed.

Has anyone implemented a configuration like this and overcome this issue, or can you offer me any advice on things I could try? I have found instances of people have a cluster of app servers and a standalone app server poiting to the one DB instance, but there was no DMZ involved. Ideally it shouldn't matter that there is a firewall as user's shouldn't be getting switched onto standalone app servers and user's who have tried connecting to a standalone app server should not be getting directed to a clustered array.

Any help or thoughts would be greatly appreciated.
Ramaprasad N
Outstanding Contributor.

Re: Clustered and Single Servers connecting to One DB instance

Hi Brendon,

From the above description, I assume that you have added 2 more app servers for the use of external people. If you want the external people to connect only to the external non-clustered app servers (and internal people to connect to the internal clustered app server only), then it's better to break the app server connections. Remove the non-clustered app servers located in DMZ, from the multiple servers environment (un-check the 'join multiple servers' option from the server settings editor, for the external app servers). With this, external people will have to use the specific SD app server name (located in DMZ) to connect to OVSD, where as internal people will have to use the another name (clustered app server) to connect to OVSD.

Will this help???

Jasper Verweij
Acclaimed Contributor.

Re: Clustered and Single Servers connecting to One DB instance


I recently learned that app servers are trying to do a heartbeat check, even if app servers are in single server mode!

I learned this when installing an app server in a DMZ environment as well (because of nasty web-api problem with uploading attachments...)

This customer had an app server running as well in internal network in single server mode (as well as DMZ App server)

We experienced that the list of active app servers and sessions was cleared each 2-3 minutes. After feedback of hp support I learned that each app server is trying to create an ITP session with the other app server and if it fails it will clear appserver and its linked sessions (due to constraint). This also called "health check"

This might be the case at your site as well. I asked the customer to open up port 30999 from internal network to DMZ and vice versa, but I haven't heard from customer if this solves the issue... Another workaround might be to alter the hosts file to fool the app server when he is trying to connect the other server, but I haven't tested this.

According to hp support "join multiple servers" is only intended for load-balancing and has nothing to do with this heartbeat polling mechanism.


Super Contributor.

Re: Clustered and Single Servers connecting to One DB instance

Hi Ramaprasad and Jasper,

Thank you both for responding.

Ramaprasad, we checked the "join multiple servers" option for the internal clustered servers but the external servers in the DMZ do not have this checked. We have set it up exactly as you suggested.

Jasper, before this post, I too had found some information regarding placing an app server in the DMZ, see ( I am getting in touch with my local hp support to see what port(s) the Server Health Monitor Process uses as I too suspect that this is causing the drama. We too have noticed the server sessions list is always inaccurate. I think I can explain (logically anyway) what is happening and I think that as long as all the servers can see each other, they will keep that list correct, and therefore allow users to connect to their prefered server (or in the case of a cluster, whichever clustered server is deemed appropriate).

I will post back when I know more.
Super Contributor.

Re: Clustered and Single Servers connecting to One DB instance

Further investigation of the application server logs showed that the servers were trying to communicate but couldn't. The firewall logs also showed this. I got our network specialists to allow the application servers on both sides to use port 30999 both ways, and now the active application list in the admin console is correct and so far there is no trace errors in the log file for the applicaton servers and users aren't getting the error messages due to their connections being cleared. Thanks for replying to this post.