Recently, I was on an engagement where use of VMware Capacity Planner was required, to get an estimate of consolidation ratios.  This was done in order to build a business case for virtualization.  Now I know that this is not a topic that many are interested in as it’s not a “core” virtualization component but I thought I should “log” some of the things I went through as there isn’t much available on the web for it.  I will assume throughout that if you’re using the software, you have access to all the binaries, documents and the Capacity Planner Dashboard.  This post is more about tips so basic knowledge is assumed or please refer to the documentation for detailed instructions.

From what I can see, the process can be broken down into four main stages:

  1. Installation
  2. Discovery
  3. Collection
  4. Reporting

Putting them altogether, the whole process looks something like this:

Capacity Planner Flowchart

Let’s take each stage in turn and see the important considerations:

1. Installation

Excellent documentation already exists for VMware Capacity Planner so I am not going to repeat that here.  That said, there are a few things that you should consider for a deployment:

  • Machine Builds: The “Getting Started” or “Installation” Guides give a minimum hardware specification for the server.  Don’t go for that!  Instead, I’ve found four cores to be about the right amount of resource on a typical collector machine.  RAM is not much of an issue so 2 – 4 GB should be enough, depending on the OS running the software.  Typically, one server of the specification mentioned is required for every 200 – 250 machines.  It depends on many factors e.g. speed of network and existing traffic on it but most importantly, it depends on age and power of the target machines and how busy they comparatively are at the time of scanning.
  • VMware Capacity Planner Data Manager Install: The next step is to install the Data Manager Software.  The install is a straight-forward process.  The only thing to say here is that I normally use a dedicated named account for it.

Once the install is done, “Setup Wizard” should be run.  For a fresh install, I use the following settings:

    • Set Collection Settings: Servers
    • Set Collection Detail Level: Complete Collection
    • Collector/Dashboard Synchronization Options: Tick “Allow changing some collector configuration settings remotely using the Dashboard”.  Leave the other two unticked but depending on privacy issues, one might want to prevent sending the domain/machine information etc.
    • Logging Details: Leave as default
    • Add Accounts: Enter details for a dedicated account that has admin access to most machines you wish to scan.  Please note: This is NOT the account that was used during installation.  That one is used to run the service.  This one accesses the machines to scan.
    • Ready to Find Systems: Select “Import Systems from file” and upload the file already in place (see below).  Untick “Run discover domains when finished” and click “Finish”

Once the wizard is finished, go to “Admin” and “Options…”.  Click on the “Jobs” tab and make sure the scheduled is suspended.  If not, tick the box “Suspend Scheduler”. Do NOT try to link up the collectors to VMware Capacity Planner Dashboard site just yet.

  • Discovery Method: VMware Capacity Planner is capable of finding systems using a discovery scan given a range of subnets and it uses various techniques to reliably detect systems.  However, due to its thoroughness, that method could take a substantial amount of time, especially if there are sizeable gaps in present systems on a given subnet.  Another way is to run a discovery tool in advance on the subnet. Once a list of responding target machines is decided, that list can be imported into Capacity Planner to run the scans.  Another benefit of the latter method is that it can be done while the machines for collection are being provisioned.  I prefer the latter, if possible.

2. Discovery

Once the collectors are ready and a list of target servers (or subnets) is decided, the next step is to discover those systems and see if there are any access issues.  The steps go like this:

  • Discovery or Import of Systems: Depending on the method chosen in the previous section, the next step is to implement that method.  Only use one method of the two.  This is not just to avoid confusion later but there also are significant differences in the way collection is done.
  • Run “Test Collection”: Once the machines are imported into the system, a “Test Collection” is run against all of them.  This is a robust test to check if the target machines are accessible on the network, are resolvable and WMI/Remote Registry access is OK.
  • Iterative Troubleshooting: The previous step comes back with some results. We could be fortunate enough to get all successes but that’s not generally the case. So, I normally export the results into a spreadsheet and filter out the failures and their causes.  Following that, I go through an iterative process to fix access issues.  That usually takes some time, depending on the number of target systems.  There is excellent documentation from VMware about Capacity Planner troubleshooting.  I also did a post on a particular issue here.  Please note: We’re still not at the stage where collectors should be joined to the Dashboard site.
  • Manual Inventory/Performance Scans: Once all access issues are fixed, a “Manual” Inventory scan of imported machines should be done.  Please make sure “Suspend Scheduler” is still selected before scanning.  This process will show if the number of machines being scanned from this collector are too many or not.  If a successful inventory can’t be run before the process times out (after 16 hours), then adjustments have to be made to get a complete inventory run.  As inventory is fine if collected once properly, one way to get this part working is to divide them up into smaller groups and run inventory scan on those individual groups.  I’ve also run parallel inventory scan threads that way and this is where the extra cores on the collectors help most.  Once that stage is reached and complete inventory is in place, a manual performance scan should be run, which also needs to complete for all machines in the list before timing out (2 hours). Successful completion of both means the collector is ready for the next stage.

This post is getting long now so it would be best if I take a break here and write a next one to cover the rest of the process.  See you in the next post!

Update: Part II is out now!