Network deployment server -single point of failure

Jaise James

Reputable Poster
Hello

People who are using Network deployment , How are you ensuring that the ND is not a single point of failure. Are you clustering it

1) are you using vertical or horizontal clustering.
2) How you are handling upgrade of tool release ( do you bring your entire infrasture down or can you upgrade one webserver at a time with horrizontal cloning)
3) DO you have your Http server seprate from websphere ( then how are you ensuring Http is not a single point of failure)
4) How are you load balacing logic /App server Windows cluster have a limitation os 2 nodes ( as per oracle) Has any one used any other solution

Any info, suggestions will be be highly appreciated
 
I'm not sure how the ND is a single point of failure. It just ties all of the individual nodes together. In fact, the ND should be considered the redundancy in many regards.

If you are worried about losing the ND piece, you can always back up your config and have a secondary ND installed that you can restore to at any time.
 
You may wish to spend some time reading about the various technologies as you are being led astray by some folks on a number of technical questions.

The network deployment server does not play a role in the actual processing functions of the WebSphere system. It exists to administer the nodes and therefore its failure will have no impact on system availability.

Load balancing logic servers does not make use of MS Clustering. If it did Oracle is still incorrect in stating that you are limited to two nodes. The method of clustering logic servers involves an IP Sprayer-type device and is not limited in the number of nodes. Darryl Shakespeare and Jon Steele have each written a whitepaper describing the method for load balancing logic servers.

Although IBM markets WAS as a full uptime solution I have found that session handling for E1 does not allow for rippling, or taking down one server at a time. I'll be glad to hear from others who can.

Having the HTTP server remote will not make your system any more or less failure prone. I have designed systems both ways and prefer from a management standpoint to have remote HTTP servers.

Take a look at the attachment. This is what I consider a "reasonably redundant" solution.



[ QUOTE ]
Hello

People who are using Network deployment , How are you ensuring that the ND is not a single point of failure. Are you clustering it

1) are you using vertical or horizontal clustering.
2) How you are handling upgrade of tool release ( do you bring your entire infrasture down or can you upgrade one webserver at a time with horrizontal cloning)
3) DO you have your Http server seprate from websphere ( then how are you ensuring Http is not a single point of failure)
4) How are you load balacing logic /App server Windows cluster have a limitation os 2 nodes ( as per oracle) Has any one used any other solution

Any info, suggestions will be be highly appreciated

[/ QUOTE ]
 

Attachments

  • 94416-WAS Architecture.jpg
    94416-WAS Architecture.jpg
    118.6 KB · Views: 195
Thanks for your reply,

I am little confused.

In our document, if if the ND goes down, how would the client know which node to go to. I thought this is maintianed by ND.

Regarding the Logic server, we are looking into f5 to the load balancing the different servers. I see you have worked with cisco, do you know of any gotcha we should be looking for.

Thanks
 
The load balancing is handled by the WAS HTTP plugin, not the network deployment server.



[ QUOTE ]
Thanks for your reply,

I am little confused.

In our document, if if the ND goes down, how would the client know which node to go to. I thought this is maintianed by ND.

Regarding the Logic server, we are looking into f5 to the load balancing the different servers. I see you have worked with cisco, do you know of any gotcha we should be looking for.

Thanks

[/ QUOTE ]
 
Back
Top