ICM and URLS

HCL Domino Domino Forum
1

I have an issue generating urls from the ICM for our intranet, they refer back to Server. I am also using LTPA Tokens and two icms in the cluster. It is a 6.5.1 server.

when I enter “htp://www.apps.com/names.nsf” in the browser it returns “htp://servera.com/names.nsf” or “htp://serverb.com/names.nsf”

Any ideas??? - Everything else works perfectly…

Round robin DNS Entries

www.apps.com → 192.192.192.1

www.apps.com → 192.192.192.2

Server A Document Settings

Fully qualified Internet host name> servera.com

Http Settings:

Internet Host names(s)> servera.com

Bind to Host name> Enable

Domino Web Engine Settings:

Session authentication> Multiple Server (SSO)

Web SSO Configuration> LtpaToken

ICM Server Task:

ICM notes Port> ICMPORT (seperate ip address 192.192.192.3/ addded to

Server ports as well)

Get configuration from> this server Document

ICM hostname> www.apps.com

Server B Document Setings

Fully qualified Internet host name> serverb.com

Http Settings:

Internet Host names(s)> serverb.com

Bind to Host name> Enable

Domino Web Engine Settings:

Session authentication> Multiple Server (SSO)

Web SSO Configuration> LtpaToken

ICM Server Task:

ICM notes Port> ICMPORT (seperate ip address 192.192.192.4/ addded to

Server ports as well)

Get configuration from> this server Document

ICM hostname> www.apps.com

2

Subject: ICM and URLS

Hi, I have not worked on this in R6 but under R5 we experienced similiar issues…

Users going to intranet.company.com (which was the ICM) and then the ICM redirecting them to either intranet1.company.com or intranet2.company.com

This in turn gave us problems if a server went down, because by this time the users were in their mail file or application they were using the URLs for the given server not the ICM so they were never redirected to serverB which was happily still running, unless they entered the URL of the ICM again. We did a lot of development work editing the apps to ensure that they always referenced the ICM URL but this was messy and involved lots of code.

In the end we choose a different tack and implemented a windows cluster which worked much better for our requirements, as follows:

  • serverA and serverB using a shared disk

  • windows runs the Notes server on either using a ‘cluster IP and name’

Therefore users always access intranet.company.com no matter what physical server it was running on. We also found that if the server failed over web users rarely noticed because by the time they next requested a page the server was up and running on the other physical server, also because it had the same URL etc. all LTPA tokens were valid and they did not need to logon again etc. etc. etc.

Hopefully this information helps. We were fortunate to be a fairly small site so could get away with this option in terms of load. We had a powerful serverA and a less powerful serverB with the business understanding that in the event of a failover the systems would be online again in 30-50seconds but would be slower, which they accepted happily compared to the issues we had experienced trying to use the ICM and Notes clusters.

Some discussion with IBM at the time resulted in larger plans for websphere, edge servers and more which were out of our price range and overkill for 500 users.

3

Subject: RE: ICM and URLS

OK. Thank you for your response…

I will do some research into Windows Clustering… Weird that a Microsoft Product would intergrate better with the Domino HTTP task than the Domino ICM Server task.

4

Subject: RE: ICM and URLS

Yes its strange that the Windows clustering works better! Although to be fair I think that the ICM is working as designed but not necessarily doing what is actually needed!

One other point I didnt mention was about using the round robin IP addresses. We thought this would work great having both entries so if the one server was down then the next request would get the alternate IP. However we found that Win2000 PC’s cached the DNS lookup and so never picked a different IP address which was annoying, the unix boxes were fine as they always looked up every time.

Just something to consider also. You can test this using a simple ping, if the round robin is working each ping command should return a different IP for the host. As said we found the windows clients only ‘got’ a different address after a reboot.

Be interested in how you get on! Good luck.