|
|
@@ -6,6 +6,7 @@ import (
|
|
|
"fmt"
|
|
|
"net/http"
|
|
|
"os"
|
|
|
+ "os/signal"
|
|
|
"sort"
|
|
|
"time"
|
|
|
|
|
|
@@ -15,74 +16,104 @@ import (
|
|
|
|
|
|
func NewClusterHealthCommand() cli.Command {
|
|
|
return cli.Command{
|
|
|
- Name: "cluster-health",
|
|
|
- Usage: "check the health of the etcd cluster",
|
|
|
- Flags: []cli.Flag{},
|
|
|
+ Name: "cluster-health",
|
|
|
+ Usage: "check the health of the etcd cluster",
|
|
|
+ Flags: []cli.Flag{
|
|
|
+ cli.BoolFlag{Name: "forever", Usage: "forever check the health every 10 second until CTRL+C"},
|
|
|
+ },
|
|
|
Action: handleClusterHealth,
|
|
|
}
|
|
|
}
|
|
|
|
|
|
func handleClusterHealth(c *cli.Context) {
|
|
|
+ forever := c.Bool("forever")
|
|
|
+ if forever {
|
|
|
+ sigch := make(chan os.Signal, 1)
|
|
|
+ signal.Notify(sigch, os.Interrupt)
|
|
|
+
|
|
|
+ go func() {
|
|
|
+ <-sigch
|
|
|
+ os.Exit(0)
|
|
|
+ }()
|
|
|
+ }
|
|
|
+
|
|
|
tr, err := getTransport(c)
|
|
|
if err != nil {
|
|
|
handleError(ExitServerError, err)
|
|
|
}
|
|
|
|
|
|
+ // TODO: update members when forever is set.
|
|
|
mi := mustNewMembersAPI(c)
|
|
|
ms, err := mi.List(context.TODO())
|
|
|
if err != nil {
|
|
|
+ fmt.Println("cluster may be unhealthy: failed to list members")
|
|
|
handleError(ExitServerError, err)
|
|
|
}
|
|
|
-
|
|
|
cl := make([]string, 0)
|
|
|
for _, m := range ms {
|
|
|
cl = append(cl, m.ClientURLs...)
|
|
|
}
|
|
|
|
|
|
- // check the /health endpoint of all members first
|
|
|
-
|
|
|
- ep, rs0, err := getLeaderStatus(tr, cl)
|
|
|
- if err != nil {
|
|
|
- fmt.Println("cluster may be unhealthy: failed to connect", cl)
|
|
|
- os.Exit(1)
|
|
|
- }
|
|
|
+ for {
|
|
|
+ // check the /health endpoint of all members first
|
|
|
|
|
|
- time.Sleep(time.Second)
|
|
|
-
|
|
|
- // are all the members makeing progress?
|
|
|
- _, rs1, err := getLeaderStatus(tr, []string{ep})
|
|
|
- if err != nil {
|
|
|
- fmt.Println("cluster is unhealthy")
|
|
|
- os.Exit(1)
|
|
|
- }
|
|
|
-
|
|
|
- if rs1.Commit > rs0.Commit {
|
|
|
- fmt.Printf("cluster is healthy: raft is making progress [commit index: %v->%v]\n", rs0.Commit, rs1.Commit)
|
|
|
- } else {
|
|
|
- fmt.Printf("cluster is unhealthy: raft is not making progress [commit index: %v]\n", rs0.Commit)
|
|
|
- }
|
|
|
- fmt.Printf("leader is %v\n", rs0.Lead)
|
|
|
+ ep, rs0, err := getLeaderStatus(tr, cl)
|
|
|
+ if err != nil {
|
|
|
+ fmt.Println("cluster may be unhealthy: failed to connect", cl)
|
|
|
+ if forever {
|
|
|
+ time.Sleep(10 * time.Second)
|
|
|
+ continue
|
|
|
+ }
|
|
|
+ os.Exit(1)
|
|
|
+ }
|
|
|
|
|
|
- var prints []string
|
|
|
+ time.Sleep(time.Second)
|
|
|
|
|
|
- for id, pr0 := range rs0.Progress {
|
|
|
- pr1, ok := rs1.Progress[id]
|
|
|
- if !ok {
|
|
|
- fmt.Println("Cluster configuration changed during health checking. Please retry.")
|
|
|
+ // are all the members makeing progress?
|
|
|
+ _, rs1, err := getLeaderStatus(tr, []string{ep})
|
|
|
+ if err != nil {
|
|
|
+ fmt.Println("cluster is unhealthy")
|
|
|
+ if forever {
|
|
|
+ time.Sleep(10 * time.Second)
|
|
|
+ continue
|
|
|
+ }
|
|
|
os.Exit(1)
|
|
|
}
|
|
|
- if pr1.Match <= pr0.Match {
|
|
|
- prints = append(prints, fmt.Sprintf("member %s is unhealthy: raft is not making progress [match: %v->%v]\n", id, pr0.Match, pr1.Match))
|
|
|
+
|
|
|
+ if rs1.Commit > rs0.Commit {
|
|
|
+ fmt.Printf("cluster is healthy: raft is making progress [commit index: %v->%v]\n", rs0.Commit, rs1.Commit)
|
|
|
} else {
|
|
|
- prints = append(prints, fmt.Sprintf("member %s is healthy: raft is making progress [match: %v->%v]\n", id, pr0.Match, pr1.Match))
|
|
|
+ fmt.Printf("cluster is unhealthy: raft is not making progress [commit index: %v]\n", rs0.Commit)
|
|
|
+ }
|
|
|
+ fmt.Printf("leader is %v\n", rs0.Lead)
|
|
|
+
|
|
|
+ var prints []string
|
|
|
+
|
|
|
+ for id, pr0 := range rs0.Progress {
|
|
|
+ pr1, ok := rs1.Progress[id]
|
|
|
+ if !ok {
|
|
|
+ // TODO: forever should handle configuration change.
|
|
|
+ fmt.Println("Cluster configuration changed during health checking. Please retry.")
|
|
|
+ os.Exit(1)
|
|
|
+ }
|
|
|
+ if pr1.Match <= pr0.Match {
|
|
|
+ prints = append(prints, fmt.Sprintf("member %s is unhealthy: raft is not making progress [match: %v->%v]\n", id, pr0.Match, pr1.Match))
|
|
|
+ } else {
|
|
|
+ prints = append(prints, fmt.Sprintf("member %s is healthy: raft is making progress [match: %v->%v]\n", id, pr0.Match, pr1.Match))
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ sort.Strings(prints)
|
|
|
+ for _, p := range prints {
|
|
|
+ fmt.Print(p)
|
|
|
+ }
|
|
|
+
|
|
|
+ if !forever {
|
|
|
+ return
|
|
|
}
|
|
|
- }
|
|
|
|
|
|
- sort.Strings(prints)
|
|
|
- for _, p := range prints {
|
|
|
- fmt.Print(p)
|
|
|
+ time.Sleep(10 * time.Second)
|
|
|
}
|
|
|
- os.Exit(0)
|
|
|
}
|
|
|
|
|
|
type raftStatus struct {
|