功能:完善云管助手与GPU故障恢复核对

This commit is contained in:
songyiming committed 2026-11-13 19:08:40 +08:00
1 parent b6026f9759
commit 9f7b01e4db
42 files changed
+3576 -120

No files matched your search

+65
View File
@@ -2,6 +2,7 @@ package proxmox
import (
"context"
"errors"
"net/http"
"net/http/httptest"
"strings"
@@ -58,6 +59,70 @@ func TestTLSAndTaskLifecycle(t *testing.T) {
t.Fatal("restore must never map to hard reset")
}
}
func TestObservationRetriesDoNotReplayPowerCommands(t *testing.T) {
for _, endpoint := range []string{"task", "power"} {
t.Run(endpoint, func(t *testing.T) {
reads, writes := 0, 0
ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
if r.Method == "POST" {
writes++
w.WriteHeader(http.StatusServiceUnavailable)
return
}
reads++
if reads < 3 {
w.WriteHeader(http.StatusServiceUnavailable)
return
}
fmtBody := `{"data":{"status":"running"}}`
if endpoint == "task" {
fmtBody = `{"data":{"status":"stopped","exitstatus":"OK"}}`
}
w.Write([]byte(fmtBody))
}))
defer ts.Close()
c := &Client{BaseURL: ts.URL, HTTP: ts.Client()}
var err error
if endpoint == "task" {
err = c.Wait(context.Background(), "node", "UPID:one", time.Millisecond)
} else {
_, err = c.Power(context.Background(), "node", 100)
}
if err != nil || reads != 3 {
t.Fatal("observation did not recover", reads, err)
}
if _, err = c.Action(context.Background(), "node", 100, "start"); err == nil || writes != 1 {
t.Fatal("command response loss must not replay POST", writes, err)
}
})
}
}
func TestObservationRetriesAreBoundedAndRespectCancellation(t *testing.T) {
for _, status := range []int{http.StatusForbidden, http.StatusServiceUnavailable} {
calls := 0
ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
calls++
w.WriteHeader(status)
}))
c := &Client{BaseURL: ts.URL, HTTP: ts.Client()}
_, err := c.Power(context.Background(), "node", 100)
want := 1
if status == http.StatusServiceUnavailable {
want = 3
}
if err == nil || calls != want {
t.Fatal("incorrect retry limit", calls, err)
}
ctx, cancel := context.WithCancel(context.Background())
cancel()
if _, err = c.Power(ctx, "node", 100); !errors.Is(err, context.Canceled) || calls != want {
t.Fatal("cancellation ignored", calls, err)
}
ts.Close()
}
}
func TestFailedPVEJobIsNotSuccess(t *testing.T) {
ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
w.Write([]byte(`{"data":{"status":"stopped","exitstatus":"ERROR: disk full"}}`))