December 15, 2020
Support for Kubernetes 1.17.9 has been added.
CCE supports hybrid cluster creation by using Kubernetes 1.17.9 version.
EulerOS 2.5 and CentOs 7.7 are also supported as node Operating System.
Kubernetes 1.17 support SFS Turbo Attachment
SFS Turbo offers by demand, fast supply, flexible expansion, convenience and flexibility，it’s suitable for：DevOps，Container microservices，corporate office etc.
Now we support importing SFS Turbo storage volumes under different VPCs * Compliant with standard file protocols, users can mount the file system to the server, just like using a local file directory.
- Data sharing, multiple servers can mount the same file system, operation and access can be shared on data
- Private network, data access must be in the internal network of the cloud
- It directly uses existing IAAS services in the cloud to build exclusive cloud file storage, providing tenants with data isolation protection and IOPS performance guarantee, mainly for DevOps, container microservices, enterprise office and other scenarios.
- Suitable for use with a variety of workloads (Deployment/StatefulSet) and common tasks (Job) in ReadWriteMany scenarios.
Currently，It’s not supported to create a SFS Turbo directly，please follow the user guide to create it in the SFS Turbo console then user SFS Turbo service in CCE
Kubernetes 1.17 support OBS Attachment
CCE supports the creation of OBS object storage volumes that can be mounted in a path under a container, making object storage suitable for cloud workloads, data analytics, content analytics, and hot objects.
CCE supports container-mounted OBS object storage, which mounts OBS buckets as shared storage on nodes based on s3fs. This approach is suitable for single save and multiple read scenarios for file objects of different sizes, not for scenarios where random changes are often made to saved files. If you want better access performance, we recommend using the OBS SDK directly. Please refer to the s3fs website for details, object storage buckets mounted based on s3fs cannot provide the same performance or semantics as the local file system, the limitations are. * Random writing or adding to the file requires rewriting the entire file
- Metadata operations such as listing directories will cause poor performance due to network delays
- Eventually consistent design may temporarily produce outdated data
- No atomic renaming of files or directories
- There is no mutual coordination mechanism between multiple clients that mount the same bucket
- Hard links are not supported
Support p2s ECS (NVIDIA v100 GPU)
CCE now supports ECS p2s as compute nodes (NVIDIA v100 GPU)
Support Hibernate management by APIs
When you do not need to use the cluster temporarily, you can set the cluster to sleep. After the cluster sleeps, node instances are stopped and are not charged anymore. Resources such as workloads cannot be created and managed on the cluster.
The sleeping cluster can be waked up quickly, and the waked up cluster can then be used normally.
Support node pool management by API
User can use APIs to create and delete node pool，node pool is a group of nodes with same configuration，one node pool could have one or multiple nodes
Support plugins management (at cluster creation) by API
CCE supports plugins management when user create cluster by API.
Advanced function plug-ins can be selectively installed according to user needs.
Support Kubernetes upgrade to version 1.15
CCE supports upgrading clusters of 1.13 version to v1.15 version: * During the cluster upgrade process, the upgrade may fail. For your data security, it is recommended to back up your data before performing the upgrade operation.
- During a cluster upgrade, you can upgrade a user node only by resetting the node. This upgrade mode is simple and rarely fails, and also supports OS upgrade. However, services will be interrupted during the upgrade. In addition, the node OS will be reinstalled, and data in the system disk and data disk will be cleared.
- User nodes can be upgraded in batches to minimize service interruption.