Releases: cloudfoundry/bosh-azure-cpi-release
v33
New Features:
-
Support Availability Zones #331
-
Support a new CPI method
calculate_vm_cloud_properties
#336With this feature,
vm_resources
can be specified in the deployment manifest.
Fixes:
-
Sleep 30 seconds before attaching data disk #340
This is a workaround for the issue #280.
-
Vendors ruby-2.4-r3 from bosh-packages/ruby-release #338
-
Change the format of user image name #342
In old format, the user image name length may exceed Azure limits (80) in some region (e.g.
Australia Southeast
). The new format has a shorter name. -
Fix a bug of using ip configuration properties to check whether the network interface is primary #337
Documents:
- Update the guidance for the bosh-setup template v3.0.0 #343
v32
Fixes:
-
Allow users to control the "keep failed vms" functionality via CPI global configuration #332
The functionality was introduced since
v24
, which is for troubleshooting the failed VM during creating. Inv32+
, you can control the functionality via thekeep_failed_vms
property in Azure CPI's global configuration. The default value isfalse
. More details are in #330.
Development:
- Add a flag to enable/disable the azure application security groups tests in CI #333
v31
New Features:
-
Support application security groups #325
Fixes:
-
Fix an issue introduced by azure-storage-ruby 0.14.0-preview #323
Background: In azure-storage-ruby 0.14.0-preview, timestamp of the entity is a String instead of a Date. v30 will fail when using multiple storage accounts.
-
Fix the issue of weak password generation in create-service-principal #324
-
Use the latest version of the REST API to create NIC to avoid possible timeout when using Azure Application Gateway #326
v30
Notices:
Please DONOT use this version if you are using multiple storage accounts. A new issue is introduced by the change in azure-storage-ruby 0.14.0-preview that start_time = entities[0]['Timestamp']
in azure-storage-ruby 0.14.0-preview is a String, but in azure-storage-ruby 0.12.1-preview it is a Time so below error will be thrown when using multiple storage accounts.
A hot fix is coming...
External CPI got response: {"result":null,"error":{"type":"Unknown","message":"can't convert String into an exact number","ok_to_retry":false},"log":"Rescued Unknown: can't convert String into an exact number. backtrace: /var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/stemcell_manager.rb:153:in `-'\n/var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/stemcell_manager.rb:153:in `wait_stemcell_copy'\n/var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/stemcell_manager.rb:100:in `handle_stemcell_in_different_storage_account'\n/var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/stemcell_manager.rb:59:in `has_stemcell?'\n/var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/cloud.rb:167:in `block in create_vm'\n/var/vcap/packages/bosh_azure_cpi/vendor/bundle/ruby/2.4.0/gems/bosh_common-1.3262.24.0/lib/common/thread_formatter.rb:49:in `with_thread_name'\n/var/vcap/packages/bosh_azure_cpi/lib/cloud/azure/cloud.rb:132:in `create_vm'\n/var/vcap/packages/bosh_azure_cpi/vendor/bundle/ruby/2.4.0/gems/bosh_cpi-2.4.1/lib/bosh/cpi/cli.rb:82:in
Fixes:
-
Refine keyword for the error Hostname not known
Background: After upgrading Ruby to 2.4.2, the error message for DNS resolve failure changes so the retry logic for this error in v29 does not work.
-
Bump azure-storage-ruby to 0.14.0-preview #317
Please see changes from 0.12.1-preview to 0.14.0-preview in azure-storage-ruby Release Notes.
Documents:
Development:
v29
Fixes:
-
Fix #314: Bump ruby to 2.4.2
Background: Proxy does not work because of one bug #1513 in Ruby 2.4.1. You can see more details in the issue #314. If you must use CPI with a proxy, you must avoid using v27 and v28. About when to use CPI with a proxy, please reference here.
Documents:
- Update the detailed steps to use application gateway #313
v28
v27
New Features:
- Support CA cert in Azure Stack so CPI can use HTTPS to communicate with Azure Stack #302
Fixes:
- Delete the empty availability set when CPI deletes a VM and the VM is the last one in the availability set #306
Improvements:
-
Retry policy handles connection refused error #298
-
Redact sensitive data in CustomData in logs #298
-
Remove the limitation of max disk size in CPI since the max disk size on Azure is still changing #310
-
Bump ruby=>2.4.1, yaml=>0.1.7, rubygems=>2.6.2, bundler=>1.15.4 #311
Development:
Documents:
v26
New Features:
-
Support multiple resource group for VM resources #286. Please see more details here
-
Support managed disks in Azure National Cloud (Germany, China and USGovernment) #288
-
Support large disk, and add new available vm size #289
-
Enable boot diagnostics when debug mode is on #295
Fixes:
- Set the correct OS disk size for Windows #290
Improvements:
- Refine locks #287
Development:
v25
New Features:
- VM is created in the same location with the virtual network instead of the storage account. #274
Fixes:
-
Handle the connection reset error to azure blob storage. #268
-
Retry once and only once if CPI receives AzureConflictError when deleting a managed disk. #269
-
Specify the disk size when attaching an unmanaged disk. Otherwise, the disk size can't be shown by Azure Portal or Azure CLI. #275
-
Uses the same SCSI host device ID for AzureStack as Azure so that it works correctly with the stemcell
v3232.5+
. #276 -
Update resource provider API versions for AzureStack. #285
Improvements:
-
Bump bosh_cpi version from
1.3262.4.0
to2.1.1
. #273 -
Upgrade azure-storage-ruby to
v0.12.1-preview
. #268- Please see the release notes of v0.12.1-preview and v0.12.0-preview.
Docs:
- Provide a checklist before migrating to managed disks. #271
Development:
v24
New Features:
- Support to set idle timeout in minutes for dynamic public IPs by configuring
pip_idle_timeout_in_minutes
in global Azure properties.- The default idle timeout of Azure public IPs is 4 minutes. The available value is [4, 30]. It must be integer.
- If the interval that your applications send
keep-alive
is longer than the idle timeout, the TCP connections from/to the public IPs will be closed by Azure. So you need to increase the idle timeout of public IPs or decrease the interval in your applications. - If you set a too big value as the idle timeout, all ports associated with public IPs may be exhausted if your applications do not close TCP connections properly.
- Keep failed VM if the VM fails in provisioning.
- Sometimes, the user needs to keep the VM for troubleshooting when the VM provisioning fails due to some platform or stemcell issues.
Fixes:
- Fix that some functions eat exceptions.
Docs:
- Add a doc about how to backup and restore Cloud Foundry