v21

[Note: It's been reported that multi-homed networks are not properly configured. v22 will fix this.]

• Add CPU & memory hot add enable option (PR #6)

v20

Improvements:

• Removed 60 minute timeout on tasks to avoid exiting prematurely when migrating large disks
  • Instead will be printing out periodic "still waiting" message in the logs
• Added vcenter.connection_options.ca_cert property to specify custom CA certs for the vCenter

Dev:

• Updated Ruby to 2.2.4
• Removed several unnecessary gem dependencies

v19

Dev:

• Deprecate configure_networks CPI call

v18

• Automatically rename and move disks to their expected locations after storage vMotion migrations.
  • Note that this only happens if disk is still attached to any VM.
• Added few safety checks to make sure that persistent disk has correct disk mode (independent persistent) when it's being detached from a VM.

v14

• The deployment manifest accepts empty hash for cluster configuration

Now allows:

clusters: 
- cluster-name: { }

which behaves the same as

clusters:
- 'cluster-name'

v13

Improvements:

• User should see an error message if other VM has a requested IP address
  • Due to large number of users experiencing this problem either due to misconfigured IP ranges or run away VMs, CPI will now query vCenter for VMs with a requested IP address and raise an error if it finds any other VM using it. We have run a few benchmarks and found that such queries did not impact performance of the CPI.
  • Example error message: Cannot create new VM because of IP conflicts with other VMs on the same networks: [{:vm_name="vm-f3f6a5fb-0936-4aed-95d9-c079a44e4f64", :network_name="salsa", :ip="10.85.33.95"}]
• User should be able to switch back and forth from using vSphere resource pools
  • We started querying vCenter for all VMs to find a specific VM by name. We have run a few benchmarks and found that such queries did not impact performance of the CPI.

v12

• Supports thick, thin, preallocated, and eager zeroed thick types for persistent disks

v11

• Renamed job cpi to vsphere_cpi
  • Note: This breaks current bosh-init deployment manifests! Be sure to change director.cpi_job, cloud_provider.template, and templates properties to specify vsphere_cpi.

v10

• Update to use Ruby 2
• Backfill missing blobstore options for S3

v9

• Fixed ability to reboot
• Improved several error messages to include more details
• Improve VM power on failure message
• Support disk migration from one datastore to another when user changes size of the persistent disk, or does bosh deploy --recreate
• Confirm that ephemeral disks and stemcells can be placed on storage that's only connected to each one of the hosts