v0.7.3
Cloud Provider Azure v0.7.3
Major changes since v0.7.2
Cloud Provider Azure v0.7.3 supports out-of-tree node ipam controller, sharing the primary SLB with multiple vmSets, and a bunch of other features/bug fixes. The images are available at:
mcr.microsoft.com/oss/kubernetes/azure-cloud-node-manager:v0.7.3
mcr.microsoft.com/oss/kubernetes/azure-cloud-controller-manager:v0.7.3
Changes by Kind
Feature
- Chore: remove get file in CreateFileShare (#534, @andyzhangx)
- Feat: add ARM node provider (#580, @JesusAlvarezTorres)
- Feat: add disable RetentionPolicy parameter (#545, @andyzhangx)
- Feat: add update vm interface (#592, @andyzhangx)
- Feat: implement cloud allocator for vmas (#555, @nilo19)
- Feat: implement cloud cidr allocator for VMSS (#539, @nilo19)
- Feat: support sharing the primary slb with multiple vmSets (#578, @nilo19)
- Feat: support system tag (#558, @nilo19)
- Feat: upgrade azure sdk to v53.1.0 (#589, @andyzhangx)
Documentation
Bug or Regression
- Allow disabling AzureStackCloud API versions when using AzureStackCloud config on public cloud (#525, @feiskyer)
- Avoid caching the VMSS instances whose network profile is nil (#583, @feiskyer)
- Azure_storageaccount.go:99] found skip-matching tag for account %!s(*string=0xc000cbd7b0), skip matching (#529, @andyzhangx)
- Ensure only LoadBalancer rule is created when HA mode is enabled (#536, @feiskyer)
- Ensure service deleted when the Azure resource group has been deleted (#584, @feiskyer)
- Fix node public IP fetching from instance metadata service when the node is part of standard load balancer backend pool. (#540, @feiskyer)
- Fix: avoid panic when RouteTablePropertiesFormat is nil (#568, @feiskyer)
- Fix: not delete existing pip when service is deleted (#574, @nilo19)
- Fix: support sharing the primary slb when there are both external and internal load balancers in the cluster (#588, @nilo19)
- Fixed routes not created issues before Pod scheduling. When using kubenet, 1) cloud-node-manager supports “–wait-routes=true” to indicate a node would wait for route updates before accepting Pod scheduling and 2) route controller would wait a while for new routes to take effect (default is 30s). (#528, @feiskyer)
- Ignore not a VMSS error for VMAS nodes in reconcileBackendPools (#551, @CecileRobertMichon)
Other (Cleanup or Flake)
- Chore: move consts in azure_vmss.go to consts.go (#554, @nilo19)
- Chore: remove bazel support (#585, @nilo19)
- Chore: switch to network api 2020-08-01 (#569, @nilo19)
- Enable docker BuildKit and update Go to 1.15.8 (#548, @CecileRobertMichon)
- Update Azure compute API version to 2020-12-01 (#579, @feiskyer)
Uncategorized
- Fix availability set cache in vmss cache (#537, @CecileRobertMichon)
- Fix: check disk state before attach disk (#564, @andyzhangx)
- Fix: convert backend pool id to lower case before using it (#561, @nilo19)
Dependencies
Added
- github.com/gofrs/uuid: v4.0.0+incompatible
Changed
- github.com/Azure/azure-sdk-for-go: v51.2.0+incompatible → v53.1.0+incompatible