Looking for in-depth information on Agent for VMware

[Cwooddell]

So for planning out my environment, I need to understand more in depth what is going on "under the hood" with the Backup Exec Agent for VMware but I am having trouble locating the information I need.

Specificially, I need to understand the information flows - what data stream goes where during a backup operation.

For simplicity of discussion's sake, let's say I have the following minimal HA cluster configuration:

• Two hosts, ESXi1 and ESXi2.
• On those hosts, only a single virtual server is found. VM1 and VM2 (each on thier respective hosts)
• A single vCenter server appliance is running on ESXi1
• The media server is a physical machine. The vCenter server is configured in BE, and all jobs are configured as jobs underneath the vCenter server.
• The storage back-end used by ESXi is a SAN consisting of two storage controllers. Each host connects with and communicates with all SANs. The protocol used by the ESXi hosts to communicate with the SAN is NFS.
• The backup destination is an iSCSI SAN not attached or related to the SAN used by the ESXi hosts. I don't think they're germane to the discussion, but I'm listing them for the sake of completeness.

The first question I'm trying to answer is, "When backing up a Virtual Machine via a vCenter server, what is/are the data path(s)?"

So say I tell BE to back up VM2. VM2 is on a different host than the vCenter server, so which of the following is the true picture of how is BE getting it's data stream?

1. Is BE pulling its data through the vCenter server, and vCenter in turn pulling from ESXi2 via ESXi1, which in turn pulls data from the SAN(s)? Path: BE <- ESXi1 <- vCenter (on ESXi1) <-> ESXi2 <- SAN
2. Is BE using the vCenter server simply as an window into the vSphere world, but is actually cutting out one of the middle men and using VAAI to pull data from ESXi2 directly? Path: BE <- ESXi2 <- SAN
3. Is BE using the Remote Agent for Windows and pulling data directly from the virtual the same as it would from a physical? Path: BE <- ESXi2 <- VM2

I suspect it's #2, but I can't find much in the way of documentation to piece together a definitive answer.

The second question, which is even deeper, and a follow up to #1, is which vmKernel port types are being used? For instance:

1. If the answer is #1 above, is the vCenter to ESXi2 communication using the VM network ports, or the management vmKernel port?
2. If the answer is #2 above, is the BE <-> ESXi2 commuication using the VM network ports, the management vmKernel port, or the vmKernel port used for NFS traffic?
3. More questions based on what's moving where over what links.

There is a dearth of information about BE 2012 beyond the surface level. Hopefully someone out there can shed some light...

[Cwooddell]

Bah I misspelled vmKernel twice and I can't figure out how to edit it.

Meta-edit: I figured it out and edited the OP. The edit button is waaay at the top of the screen. :D

[Colin_Weaver]

if you are using SAN Transport then BE directly accesses the Snapshot in the Datastore which would make the data path

LUN Containing Datastore --> via SAN to Media Server --> via SAN back to your B2D storage

BE will send control metadata to the vCenter (over the LAN) to ask for the list of VM's and to send the create Snapshot requests. It will also get back some metadata about the snapshot(s) from the vCenter so that the API on the media server knows the identity of the snapshot(s) in order to mount it/them over the SAN.

If using NBD it wil be a similar mechanism, although I think we have to attach to the snapshot via the current ESX host, so once we have got the snapshot details form the vCenter the data transfer paths switch to the ESX host (this is a guess I am not 100% sure of this last point)

With regards comms to the vCenter and/or hosts we inistially use Port 902 (not sure what the translates to in terms of kernels)