cmd/{k8s-proxy,containerboot,k8s-operator},kube: add health check and metrics endpoints for k8s-proxy (#16540)

* Modifies the k8s-proxy to expose health check and metrics
endpoints on the Pod's IP.

* Moves cmd/containerboot/healthz.go and cmd/containerboot/metrics.go to
  /kube to be shared with /k8s-proxy.

Updates #13358

Signed-off-by: David Bond <davidsbond93@gmail.com>
This commit is contained in:
David Bond
2025-07-22 17:07:51 +01:00
committed by GitHub
parent 22a8e0ac50
commit 4494705496
8 changed files with 196 additions and 82 deletions

View File

@@ -1,57 +0,0 @@
// Copyright (c) Tailscale Inc & AUTHORS
// SPDX-License-Identifier: BSD-3-Clause
//go:build linux
package main
import (
"fmt"
"log"
"net/http"
"sync"
"tailscale.com/kube/kubetypes"
)
// healthz is a simple health check server, if enabled it returns 200 OK if
// this tailscale node currently has at least one tailnet IP address else
// returns 503.
type healthz struct {
sync.Mutex
hasAddrs bool
podIPv4 string
}
func (h *healthz) ServeHTTP(w http.ResponseWriter, r *http.Request) {
h.Lock()
defer h.Unlock()
if h.hasAddrs {
w.Header().Add(kubetypes.PodIPv4Header, h.podIPv4)
if _, err := w.Write([]byte("ok")); err != nil {
http.Error(w, fmt.Sprintf("error writing status: %v", err), http.StatusInternalServerError)
}
} else {
http.Error(w, "node currently has no tailscale IPs", http.StatusServiceUnavailable)
}
}
func (h *healthz) update(healthy bool) {
h.Lock()
defer h.Unlock()
if h.hasAddrs != healthy {
log.Println("Setting healthy", healthy)
}
h.hasAddrs = healthy
}
// registerHealthHandlers registers a simple health handler at /healthz.
// A containerized tailscale instance is considered healthy if
// it has at least one tailnet IP address.
func registerHealthHandlers(mux *http.ServeMux, podIPv4 string) *healthz {
h := &healthz{podIPv4: podIPv4}
mux.Handle("GET /healthz", h)
return h
}

View File

@@ -121,7 +121,9 @@ import (
"tailscale.com/client/tailscale"
"tailscale.com/ipn"
kubeutils "tailscale.com/k8s-operator"
healthz "tailscale.com/kube/health"
"tailscale.com/kube/kubetypes"
"tailscale.com/kube/metrics"
"tailscale.com/kube/services"
"tailscale.com/tailcfg"
"tailscale.com/types/logger"
@@ -232,13 +234,13 @@ func run() error {
}
defer killTailscaled()
var healthCheck *healthz
var healthCheck *healthz.Healthz
ep := &egressProxy{}
if cfg.HealthCheckAddrPort != "" {
mux := http.NewServeMux()
log.Printf("Running healthcheck endpoint at %s/healthz", cfg.HealthCheckAddrPort)
healthCheck = registerHealthHandlers(mux, cfg.PodIPv4)
healthCheck = healthz.RegisterHealthHandlers(mux, cfg.PodIPv4, log.Printf)
close := runHTTPServer(mux, cfg.HealthCheckAddrPort)
defer close()
@@ -249,12 +251,12 @@ func run() error {
if cfg.localMetricsEnabled() {
log.Printf("Running metrics endpoint at %s/metrics", cfg.LocalAddrPort)
registerMetricsHandlers(mux, client, cfg.DebugAddrPort)
metrics.RegisterMetricsHandlers(mux, client, cfg.DebugAddrPort)
}
if cfg.localHealthEnabled() {
log.Printf("Running healthcheck endpoint at %s/healthz", cfg.LocalAddrPort)
healthCheck = registerHealthHandlers(mux, cfg.PodIPv4)
healthCheck = healthz.RegisterHealthHandlers(mux, cfg.PodIPv4, log.Printf)
}
if cfg.egressSvcsTerminateEPEnabled() {
@@ -438,8 +440,8 @@ authLoop:
)
// egressSvcsErrorChan will get an error sent to it if this containerboot instance is configured to expose 1+
// egress services in HA mode and errored.
var egressSvcsErrorChan = make(chan error)
var ingressSvcsErrorChan = make(chan error)
egressSvcsErrorChan := make(chan error)
ingressSvcsErrorChan := make(chan error)
defer t.Stop()
// resetTimer resets timer for when to next attempt to resolve the DNS
// name for the proxy configured with TS_EXPERIMENTAL_DEST_DNS_NAME. The
@@ -644,7 +646,7 @@ runLoop:
}
if healthCheck != nil {
healthCheck.update(len(addrs) != 0)
healthCheck.Update(len(addrs) != 0)
}
if cfg.ServeConfigPath != "" {

View File

@@ -1,79 +0,0 @@
// Copyright (c) Tailscale Inc & AUTHORS
// SPDX-License-Identifier: BSD-3-Clause
//go:build linux
package main
import (
"fmt"
"io"
"net/http"
"tailscale.com/client/local"
"tailscale.com/client/tailscale/apitype"
)
// metrics is a simple metrics HTTP server, if enabled it forwards requests to
// the tailscaled's LocalAPI usermetrics endpoint at /localapi/v0/usermetrics.
type metrics struct {
debugEndpoint string
lc *local.Client
}
func proxy(w http.ResponseWriter, r *http.Request, url string, do func(*http.Request) (*http.Response, error)) {
req, err := http.NewRequestWithContext(r.Context(), r.Method, url, r.Body)
if err != nil {
http.Error(w, fmt.Sprintf("failed to construct request: %s", err), http.StatusInternalServerError)
return
}
req.Header = r.Header.Clone()
resp, err := do(req)
if err != nil {
http.Error(w, fmt.Sprintf("failed to proxy request: %s", err), http.StatusInternalServerError)
return
}
defer resp.Body.Close()
for key, val := range resp.Header {
for _, v := range val {
w.Header().Add(key, v)
}
}
w.WriteHeader(resp.StatusCode)
if _, err := io.Copy(w, resp.Body); err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
}
}
func (m *metrics) handleMetrics(w http.ResponseWriter, r *http.Request) {
localAPIURL := "http://" + apitype.LocalAPIHost + "/localapi/v0/usermetrics"
proxy(w, r, localAPIURL, m.lc.DoLocalRequest)
}
func (m *metrics) handleDebug(w http.ResponseWriter, r *http.Request) {
if m.debugEndpoint == "" {
http.Error(w, "debug endpoint not configured", http.StatusNotFound)
return
}
debugURL := "http://" + m.debugEndpoint + r.URL.Path
proxy(w, r, debugURL, http.DefaultClient.Do)
}
// registerMetricsHandlers registers a simple HTTP metrics handler at /metrics, forwarding
// requests to tailscaled's /localapi/v0/usermetrics API.
//
// In 1.78.x and 1.80.x, it also proxies debug paths to tailscaled's debug
// endpoint if configured to ease migration for a breaking change serving user
// metrics instead of debug metrics on the "metrics" port.
func registerMetricsHandlers(mux *http.ServeMux, lc *local.Client, debugAddrPort string) {
m := &metrics{
lc: lc,
debugEndpoint: debugAddrPort,
}
mux.HandleFunc("GET /metrics", m.handleMetrics)
mux.HandleFunc("/debug/", m.handleDebug) // TODO(tomhjp): Remove for 1.82.0 release.
}