lerko
50f77da131
Replace inline numeric literals with named constants across 14 files: server timeouts/rate limits, cluster thresholds/intervals, DB pool sizes, alert/dial timeouts, TUI uptime thresholds, node status thresholds, and state history limits.
87 lines
2.0 KiB
Go
87 lines
2.0 KiB
Go
package cluster
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"net/http"
|
|
"strings"
|
|
"time"
|
|
|
|
"gitea.lerkolabs.com/lerkolabs/uptop/internal/monitor"
|
|
)
|
|
|
|
type Config struct {
|
|
Mode string // "leader" or "follower"
|
|
PeerURL string // URL of the Leader (e.g., http://primary:8080)
|
|
SharedKey string // Security Key
|
|
}
|
|
|
|
func Start(ctx context.Context, cfg Config, eng *monitor.Engine) {
|
|
if cfg.Mode == "leader" {
|
|
fmt.Println("Cluster: Running as LEADER (Active)")
|
|
if cfg.SharedKey != "" {
|
|
fmt.Println("WARNING: Cluster mode enabled. Ensure the HTTP server is behind a TLS-terminating proxy.")
|
|
}
|
|
eng.SetActive(true)
|
|
return
|
|
}
|
|
|
|
if cfg.Mode == "follower" {
|
|
fmt.Println("Cluster: Running as FOLLOWER (Passive)")
|
|
if cfg.PeerURL != "" && !strings.HasPrefix(cfg.PeerURL, "https://") {
|
|
fmt.Println("WARNING: Cluster peer URL is not HTTPS. Cluster secret will be sent in cleartext.")
|
|
}
|
|
eng.SetActive(false)
|
|
go runFollowerLoop(ctx, cfg, eng)
|
|
}
|
|
|
|
// "probe" mode is handled directly in main.go before cluster.Start is called
|
|
}
|
|
|
|
const (
|
|
followerTimeout = 2 * time.Second
|
|
leaderFailureThreshold = 3
|
|
followerRetryInterval = 5 * time.Second
|
|
)
|
|
|
|
func runFollowerLoop(ctx context.Context, cfg Config, eng *monitor.Engine) {
|
|
client := http.Client{Timeout: followerTimeout}
|
|
failures := 0
|
|
threshold := leaderFailureThreshold
|
|
|
|
for {
|
|
select {
|
|
case <-time.After(followerRetryInterval):
|
|
case <-ctx.Done():
|
|
return
|
|
}
|
|
|
|
req, _ := http.NewRequest("GET", cfg.PeerURL+"/api/health", nil)
|
|
if cfg.SharedKey != "" {
|
|
req.Header.Set("X-Uptop-Secret", cfg.SharedKey)
|
|
}
|
|
|
|
resp, err := client.Do(req)
|
|
isLeaderHealthy := false
|
|
|
|
if err == nil {
|
|
isLeaderHealthy = resp.StatusCode == http.StatusOK
|
|
_ = resp.Body.Close()
|
|
}
|
|
|
|
if isLeaderHealthy {
|
|
failures = 0
|
|
if eng.IsActive() {
|
|
eng.SetActive(false)
|
|
eng.AddLog("Cluster: Leader detected. Switching to PASSIVE.")
|
|
}
|
|
} else {
|
|
failures++
|
|
if failures >= threshold && !eng.IsActive() {
|
|
eng.SetActive(true)
|
|
eng.AddLog("Cluster: Leader Unreachable. Switching to ACTIVE.")
|
|
}
|
|
}
|
|
}
|
|
}
|