Releases: foundation-model-stack/multi-nic-cni
Releases · foundation-model-stack/multi-nic-cni
v1.0.5
v1.0.4
Changes:
- fix scalability issue due to API server
- apply linter code analysis removing unused function, handling errors
- allow changing reconciler time and log level on the fly from config.multinic
- allow changing synchronization ticker period (TickerIntervalKey) and maximum size of daemon pod watching queue (MAX_QSIZE) from environment variable
- change logger time encoder to ISO 8601
more detail:
https://foundation-model-stack.github.io/multi-nic-cni/release/v1.0.4/
v1.0.3
Stable version for minimum requirement:
- Auto discovery of Host and Interface
- IPVLANL3 with Host-Interface CIDR computation, auto route configuration
- Number-based policy selection
- Tolerate to node failure
- Dynamic CIDR changes corresponding to Host/Interface changes
- Scaled (up to 200 node scale, 8 host block)
Changes:
- updated CRDs: cidrs. multinicnetwork, config
- separate log level on controller manager
- log CNI message on host
- CI tests for daemon, CNI and scalability
- several bug fixes
more details:
https://foundation-model-stack.github.io/multi-nic-cni/release/v1.0.3/
v1.0.2
Core Features:
- Host-interface auto-discovery
- Single definition for multiple secondary network attachments
- Multi-NIC IPAM (CIDR computation, IP allocation/deallocation) for multiple secondary subnets
- L3 configurations on host neighbour route table corresponding to ipvlan CNI plugin with l3 mode
Supplementary Features:
- NIC selection based on specific requested number or specific interface name list
- Dynamic CIDR updates when
- detecting added/removed hosts at creation/deletion of multi-nic daemon
- (periodically) discovering added/removed secondary interfaces
- Fault tolerance in scale (tested upto 100 nodes x 2 secondary interfaces) with
- initial synchronization of CIDR, IPPool after controller restarted
- periodic synchronization of L3 routes for hosts which were restarted and lost the configuration