It's not a published API method, but there is a way to get the list of vCenter servers in a linked mode group. I have some explanation and sample code: http://www.virtuin.com/2012/12/querying-vcenter-linked-mode-with.html
In terms of authentication, if your vCenter instances are using the same permissions (which is recommended), then it's just a matter of authenticating to each vCenter. This is what the native client does as well (if you ever setup vCenters with different permissions you'll see login prompts for each vCenter).
If you look at the blog post I linked earlier, you'll se an example of authenticating to each vCenter as well. Basically you just call Util::connect() twice, once for each of your vCenter URLs. However, you do need to clear the VIM global that the SDK toolkit uses, or it will default to the last vCenter. I show this as well in the sample script: LinkedModeExample.pl
In terms of combining queries....not so easy. You'll have to call into each vCenter with $vim->find_entity_views(). Now, on a more advanced note, the API is quite powerful for doing customized queries against the inventory, so you can generally pre-fetch and keep updates into your objects.
What I would recommend as a simple approach is two calls to find_entity_views, then push them into a hash or array. You'll just want to use the vCenter UUID:
$vc_uuid = $vim1->get_service_content()->{'about'}->{'instanceUuid'};
The instanceUuid was added in vCenter 4.x and later (won't have it in older vCenter 2.5 instances). Then use this instanceUuid + moref as hash keys. This is important because morefs can be duplicated between vCenters.
So you can then build a hash:
my %vm_hash;
foreach my $entity (@$vm_views) {
my $moref = $entity->{'mo_ref'}->{'value'}; # will be vm-# for virtual machines
$vm_hash{$vc_uuid . "+". $moref} = $entity;
}
Then you can quickly lookup specific VMs and keep them separate between vCenters.
You'll find this works for simple data queries, but as you start to get more and more inventory objects (say for an inventory report), the relationships get more complex and the number of objects get larger. I've done work with pushing these values into a database and excel files (including running it on 5s or less intervals with WaitForUpdatesEx), which works well if you have the time to build the DB schemas and object to database work. Just use instanceUuid + Moref value as a primary key to prevent any cross vCenter moref value re-use.
The other option is that the vim reference is added to each object, so you can just get data from each entity:
foreach my $vm (@$vm_views) {
print "VM: " . $vm->{'name'} . "\n";
print " VC InstanceUuid: " . $vm->{'vim'}->{'service_content'}->{'about'}->{'instanceUuid'} . "\n";
print " VC ServiceUrl: " . $vm->{'vim'}->{'service_url'} . "\n";
}
I have a project I tinker with on my free time (which is rare these days) that is meant to aggregate multiple vCenters into a single database for queries and reporting, but I haven't gotten it far enough to share it (very rough, but works for a subset of data currently, but is very efficient at bulk collection on <5s intervals).
You could also look into VMware Orchestrator. Orchestrator can connect to multiple vCenters and will cache data from each for access by automation. Depending on your WAN performance, it may work well from one geographical site and provide you that aggregated data you require (just watch the moref, entity names and other possible duplication between vCenters).
Reuben Stump | http://www.virtuin.com | @ReubenStump
