mirror of https://hubproxy.babadafafafafa.cn/https://github.com/kube-vip/kube-vip.git synced 2026-09-20 08:03:47 +08:00

Files

Guillermo Gaston dcd3236925 Add configurable health check in BGP mode

In BGP mode without leader election, every kube-vip instance announces
the same VIP. ECMP distributes traffic evenly, but if a node's
kube-apiserver goes down, kube-vip (a static pod tied to kubelet) keeps
advertising the route — sending a fraction of all API traffic into a
black hole.

An optional health check loop now polls the local kube-apiserver and
withdraws the BGP route after a configurable number of consecutive
failures, removing the dead node from the ECMP set. The route is
re-announced once the API server recovers.

Signed-off-by: Guillermo Gaston <ggastonloren@nvidia.com>

2026-05-04 20:04:45 +00:00

3.2 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

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

Changed

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

3.2 KiB Raw Blame History