Shutdown Connection server?

[daleallenc]

I had a weird problem with a stand-alone (VM) Connection server. All of the files for it except for the -flat.vmdk file disappeared. This manifests itself with the machine name grayed-out and the comment "(inaccessible)" next to it in vCenter. However, the VM is still running, is still brokering desktops to users as required, and still allowing me to administer the environment. Restarting management services on the ESX host failed to bring it back up in vCenter.

All is not lost. There is a knowledgebase fix for this: Verifying ESX virtual machine file integrity (1003743) , which basically involves creating a new VM and changing file descriptors to match the original confugled machine.

Here's my dilemma. I created a Replica server and it replicated the ADAM database, and I can administer my View environment from it just fine. I need to shutdown the original Connection server in order to fix it's file structure. Now that I have the Replica, can I shut down the original Connection server without jeopardizing access to the rest of the environment? I know that the system was designed for fault tolerance in mind, but has anyone ever done it? And, if I can't revive the original Connection server, will the Replica server continue to run the environment by itself?

[lbourque · Accepted solution]

Yup. That's the idea behind having the replicas. Let's say someone decided that the Standard server would make a better Exchange server and forgot to tell you that they've yanked it and reprovisioned it into that role. As long as you have one replica, it has a full copy of the database. All replicas are peer level to each other (that is, there is master/slave or parent/child concept). The main difference is this: A standard View server starts off a View environment; a replica creates multiple avenues into an existing environment and allows it to scale.

So yes, you can hose it and continue the environment with the replica. You should take a look at the Admin Guide on how to backup the ADAM database regularly so you can restore it rather than just rebuilding.

[Troy_Clavell]

... but what happens if all your clients are pointing the the Main Connection server, if it goes down, no one will be able to connect. Assuming there is some sort of load balancing of your brokers, you will be fine.

We load balance with Round Robin DNS.

[lbourque]

Yup or you can use a web load balancer. The Admin Guide, IIRC, covers this as well. You can use F5 server or NLB or Hercules or BalanceNG.

[daleallenc]

Thanks, lbourque. I'll give it a shot at my next maintenance window - and keep my fingers crossed.