Deploying Pivotal Cloud Foundry in Your Own Home Lab
When I began this post I thought the install would be easy. I found I had to work through a few challenges. The following summarizes some of the key things you need to know before getting started.
- The OVF template VM is pedantic about what datastores it installs to. I found they needed to be formatted as VMFS 5.60. The original datastores I had were 5.58 and created as part of the original ESX 5.1 install which I upgraded to 5.5. I ended up creating a new 5.60 formatet datastore. Note that the NFS volume would not work either.
- For your system domain create a domain on your DNS server with something like *.cf.mattzwol.com and point it at an IP address that you will use for your HA Proxy. The Pivotal CF creates the HA Proxy, you just need to specify an IP for it. I created an “A” type record of *.cf.mattzwol.com and pointed it at IP 10.0.0.183. Note that the address range I used for the Pivotal CF VMs started at 10.0.0.180. I found a blog suggesting to use the 3rd address after that.
- Exclude IP ranges that are used by your DHCP server / router, Ops Manager. The IP used for the HA Proxy should be part of the included IP address range, but make sure you leave the first two IPs of the included range for Pivotal CF services.
- You will receive an error stating “Cluster ‘MyCLuster’ has 16 CPU Cores. Installation requires 36 cores.” Ignore this and continue.
- Sometimes running the install a few times without changing anything fixes problems.
- If the install goes wrong, try deleting elastic runtime from Pivotal Ops Manager and starting again. If you need to unwind it totally or cannot delete it, manually delete the VMs, files and templates in your ESX environment associated with the Pivotal CF install.
- It can take an hour or two for the install to complete. I upgraded the storage to the Synology with SSDs to speed up the install.
- I kept receiving a failure while installing Bosh. It turned out I had an issue or corruption in my vcenter server. If this happens you may need to do what I did and reinstall the vcenter server appliance.
- I couldn’t ping my DNS VM, which is SimpleDNS running on a Windows 8 machine. I had to run this command to open ports “netsh firewall set icmpsetting 8 enable” so the DNS would work.
- While I was having issues, I broke the install down into smaller parts to work out where my problems were coming from. I did this by just installing Ops Manager Director first, then when this was successful I did the elastic runtime, followed by the MySQL service individually.
- Make sure you are running in Automatic mode for DRS. Otherwise you will get the error: /home/tempest-web/tempest/web/vendor/bundle/ruby/2.0.0/gems/bosh_vsphere_cpi-1.2366.0/lib/cloud/vsphere/client.rb:159:in `power_on_vm’: Recommendations were detected, you may be running in Manual DRS mode. Aborting. (RuntimeError)
- Here is a screen shot of how DRS should be configured.
- If you receive an error stating “the operation is not allowed in the current connection state of the host”, have a look at “Recent Tasks” pane in your vCenter console. If you see a failure relating to cloning a virtual machine, try this. In vCenter, right click each one of your servers and click disconnect. Wait a moment and then reconnect them.