micahyoung / bosh-vmrun-cpi-release Goto Github PK

Ideas:

• Naive ping
• vmrun list | vmrun getGuestIpAddress
• Explore whether the API supports this

• Support bosh create-env after director is hard-shutdown
• Support bosh create-env when VMX files have been deleted
• Support bosh deploy for existing deployment after VMs are shutdown
• Support bosh recreate for existing deployment after VMs are shutdown
• Support bosh deploy when VMX files have been deleted

Would floating IPs make sense?

Task 5 | 22:24:29 | Preparing deployment: Preparing deployment (00:00:00)
Task 5 | 22:24:29 | Preparing package compilation: Finding packages to compile (00:00:00)
Task 5 | 22:24:29 | Compiling packages: 7zip/c1dc6c69e6d29a7e39752711fa1eeb74c17aac65
Task 5 | 22:24:29 | Compiling packages: golang-windows/cd8520750605724aa618cfa7228f0794a6993fbc
Task 5 | 22:24:31 | Compiling packages: 7zip/c1dc6c69e6d29a7e39752711fa1eeb74c17aac65 (00:00:02)
L Error: CPI error 'Bosh::Clouds::CloudError' with message 'Running '/Applications/VMware Fusion.app//Contents/Library/vmrun clone /Users/micah/workspace/vmfiles/vm-store-path/cs-4006ae40-590b-49ba-6cec-0600ad7fb1d4/cs-4006ae40-590b-49ba-6cec-0600ad7fb1d4.vmx /Users/micah/workspace/vmfiles/vm-store-path/vm-c0284a54-7cca-4c4a-5020-a5bd2cc72fae/vm-c0284a54-7cca-4c4a-5020-a5bd2cc72fae.vmx linked -cloneName=vm-c0284a54-7cca-4c4a-5020-a5bd2cc72fae: Error: Cannot open VM: /Users/micah/Box/workspace/vmfiles/vm-store-path/cs-4006ae40-590b-49ba-6cec-0600ad7fb1d4/cs-4006ae40-590b-49ba-6cec-0600ad7fb1d4.vmx, The file already exists
': exit status 255' in 'create_vm' CPI method

Reproduce

1. bosh deploy
2. vmrun addSharedFolder ...
3. bosh recreate
4. Error:
   Windows:

                    L Error: CPI error 'Bosh::Clouds::CloudError' with message 'Running 'c:\Program Files (x86)\VMware\VMware Workstation\vmrun.exe start E:\vmfiles\vm-store\vm-72200b02-1f70-407a-723b-56edbe34b7d3\vm-72200b02-1f70-407a-723b-56edbe34b7d3.vmx nogui: Error: Cannot open VM: E:\vmfiles\vm-store\vm-72200b02-1f70-407a-723b-56edbe34b7d3\vm-72200b02-1f70-407a-723b-56edbe34b7d3.vmx, Cannot read the virtual machine configuration file
': exit status 4294967295' in 'detach_disk' CPI method (CPI request ID: 'cpi-227537')

Background:
create_vm does not include any job name info - instead this is only supplied in set_vm_metadata

Idea: tags https://bosh.io/docs/manifest-v2/#tags

tags:
  project: cf

• Fail fast so it doesn't have to wait for "VM is in use" from vmrun

Maybe each stemcell/VM directory should get an identifier file that could later be used to find outdated VMs and Stemcells.

Use tar reader and gzip reader to avoid unpacking all stemcells to find manifests

Example: https://gist.github.com/indraniel/1a91458984179ab4cf80

Requirements:

• Modify jobs/packages/vmrun_cpi/cpi.erb to act as a shim
  • On hypervisor: Make a copy of the cpi-{platform}, config.json and stemcell store on the hypervisor VM.
  • On hypervisor: Ensure authorized_keys has entry for restricted,command="<cpi> -configPath <config.json>" ssh-rsa...
  • On director: call actual cpi-{platform} remotely over SSH using hypervisor machine
• Change create_stemcell to optionally ignore imagePath and instead use a stemcell store on hypervisor machine of unextracted pristine stemcells.
  • Would use cloudProperty metadata then scan all hypervisor stemcells' stemcell.MF manifests to find match.
  • Works around hypervisor CPI-over-ssh not having access to director's imagePath location
  • This could also enable light stemcells for use both in create-env / upload-stemcell / deploy for significant speed up by not verifying stemcells every time.
  • Change to be idempotent

WIP: branch: bosh-deployment

restrict,command="~/vmfiles/vm-store-path/cpi ${SSH_ORIGINAL_COMMAND#* }",port-forwarding,permitopen="*:6868",permitopen="*:25555" ssh-rsa ...

• Workstation/Fusion must be closed while running + Fusion Force Quit workaround

I think this causes a failure with bosh create-env --recreate --recreate-perisistent-disks on an existing director