Client Management Suite

ok so we have a site that they powered down a mess of servers to move to a different building. This is going to be this way for a couple of weeks. My client machines are now not able to download any software or patches. 

I thought if the package server setup in Site management wasnt reachable, the clients would automatically go to the main NS?

I'm not sure it works that way unless you specifically set up the NS as a Package Server. I think you may need to delete the offline Package Server in Site Management until it comes back, you could maybe try and create a dummy site and just assign it to that but deletion is probably the easiest option. You can then reassign it once it come back up.

interesting as one machine did actually switch. We watched it and it downloaded the packages. I thought the purpose of package servers was to alleviate any bandwith issues but can point back to the main NS... 

We have a separate site server setup in the same subnet as the NS.  In the case where a site server goes down in a branch office, we usually have to "reset agent' on the client task tab for the machine to pickup a different site server without rebooting the machine or restarting the SMA service.  We can use an Altiris job to push this to the machines or the RAAD tool to test. 

Task and Package server allocation is separate AFAIK.

Previous testing I've done, has shown that a computer in a site, will be bound to this site for package server assignment. Thus if you have a site with a package server, a client is assigned through a subnet or whatnot, and the package server goes down, assuming the client is not covered by other site servers it will NOT get a codebase for new packages, and the NS will provide codebases for old packages to the defunct/unavaialble package server, as it still believes it's available to service packages.

In fact, if you create a site, add no site server, then add a client to this, previous testing show it getting no codebases.

I'll be honest and say, that I haven't tested this for a little while, so it could be functionality has changed. I can also add, that various people in- and outside of Symantec have different expectations of what is supposed to happen... The testing I performed however, showed you'd see the result you are seeing where clients will essentially be unable to download anything until the assigned site server comes back and can provide the expected download location for the issued codebases.

/Rasmus

Some other food for thought...

You could try and cascade your package servers in a way, that you have one or more "upstream" site servers in sites with just one local package sever. Assuming you can handle the potential load to the "upstream" site server if it's elected based on ping-/speed-tests.

This would allow you a fall-back site server for packages, while under normal circumstances service packages locally. There is obviously more to this than meets the eye, but wanted to add this as an additional design idea.

In addition to the above if you have manually assigned endpoints to a site they will only communicate with that package server.  

It's not a good idea to delete the site server from the console.  When it comes back online package service and task service will uninstall and you will need to re-replicate and assign subnets etc... Think of this as a DR situation.  Just reroute the traffic.

If your network bandwidth will allow the additional traffic you can remove the manually assigned machines and add the subnet to another site and the endpoints will retrieve codebase from the newly assigned package server.  

You should think about adding a main site server on the same subnet as your NS and SQL servers.  This allows any endpoint without a listed subnet to download from the main site server.  Just make sure to bandwidth throttle all your endpoints and site servers outside your NS and SQL subnet.  

If you are using CEM there are many other things you need to look at.