fanhuadesenlinnn/kube-vip
1
0
Fork 0
mirror of https://hubproxy.babadafafafafa.cn/https://github.com/kube-vip/kube-vip.git synced 2026-09-20 08:03:47 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
a51944b89d641ae65a6e5e6c630bdac9226d286b
kube-vip/CHANGELOG.md
Justin Cichra 2b126dceed Fix bgp_attach_ip_to_interface not applied to Service VIPs (#1744) 
shouldAddServiceIP() checks BGPAttachIPToInterface on a per-service
config struct. NewInstance() builds that struct fresh for each Service,
copying over most fields from the global config, missing this one. As
a result the flag has no effect regardless of its value, and BGP-mode
Service VIPs are never bound to the interface.

Copy the field at both construction sites where it's built, and add
a regression test covering the propagation.

Signed-off-by: Justin Cichra <jrcichra@yahoo.com>
2026-08-31 00:20:08 +02:00

4.5 KiB
Raw Blame History

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[Unreleased]

Fixed

  • Propagate bgp_attach_ip_to_interface into per-service config so it attaches BGP-mode Service VIPs to the interface as configured.
  • Add a configurable kube-vip instance name and use it to isolate internal nftables egress tables, persist table ownership on Services, and migrate per-Service chains without affecting other deployments. Fixes #1634.
  • Retry on 403 Forbidden and 401 Unauthorized in ServicesWatcher at startup with exponential backoff. Fixes #1464.
  • Reintroduce BGP config via node annotations. Fixes #1488.
  • Fail fast in runtime manager and service paths when legacy vip_address is used without vip_subnet in control-plane ARP, BGP, or Routing Table mode.
  • Cancel the mode context on init or configuration failure before waiting on goroutines during shutdown.

Added

  • Configurable control-plane health check for BGP mode without leader election
    • Polls a configurable HTTP(S) endpoint (e.g. https://localhost:6443/livez) to verify the exposed service is healthy (usually the local kube-apiserver)
    • Withdraws the BGP route after a configurable number of consecutive failures, removing the unhealthy node from the ECMP set
    • Re-announces the route automatically once the endpoint recovers
    • Gracefully withdraws the route on shutdown (SIGTERM)
    • Supports custom CA certificates for TLS verification
    • Configuration via environment variables or CLI flags:
      • control_plane_health_check_address / --controlPlaneHealthCheckAddress: URL to poll
      • control_plane_health_check_period_seconds / --controlPlaneHealthCheckPeriodSeconds: interval between checks (default: 5)
      • control_plane_health_check_timeout_seconds / --controlPlaneHealthCheckTimeoutSeconds: per-request timeout (default: 3)
      • control_plane_health_check_failure_threshold / --controlPlaneHealthCheckFailureThreshold: consecutive failures before withdrawal (default: 3)
      • control_plane_health_check_ca_path / --controlPlaneHealthCheckCAPath: CA cert for HTTPS verification
  • SIGUSR1 signal handler for runtime configuration dumps (#1301)
    • Send SIGUSR1 to kube-vip process to dump current configuration to stdout
    • Configuration dump includes:
      • Basic configuration (VIP, interface, port, namespace settings)
      • BGP configuration (enabled status, AS number, router ID, peers)
      • ARP/NDP configuration (enabled status, broadcast rate)
      • Services configuration (enabled status, load balancer settings)
      • Network interfaces status
      • Leader election configuration (type, lease details)
      • Runtime statistics (load balancer, Prometheus, health check settings)
    • Output format: Human-readable plaintext via fmt.Printf()
    • Thread-safe implementation using mutex protection
    • Non-disruptive: Process continues running after configuration dump
    • Added comprehensive unit tests for all dump methods
    • Added E2E tests for signal handling
  • Opt-in support for endpointless LoadBalancer services with externalTrafficPolicy: Cluster
    • Annotation: kube-vip.io/allow-reconcile-without-endpoints: "true"
    • Starts service handling path for opted-in endpointless Cluster services while preserving default endpoint-gated behavior for non-opt-in services and Local policy
    • Added endpoint behavior tests and README usage documentation
    • Added support in ipoib interfaces in ARP mode. Fixes #694

Changed

  • BGP mode now honours enable_leader_election for services: a single global services leader advertises the service VIPs instead of every node advertising them. Deployments that enabled enable_leader_election for the control plane and relied on ECMP/multipath for services must unset it (or switch to enable_service_election) to keep the previous datapath. kube-vip logs a warning on startup when this path is taken.
  • Updated signal handlers in manager_arp.go, manager_bgp.go, manager_wireguard.go, and manager_table.go to use switch statement pattern for handling multiple signals (SIGUSR1, SIGINT, SIGTERM)
  • wireguard.go now manages a complete wireguard interface on the current network namespace
  • manager_wireguard.go uses the new wireguard.go implementation

[v1.0.1] - Previous Release

Previous changes

  • See git history for changes prior to CHANGELOG.md introduction