fix(quota): 回收的服务器保留资源规格,认领时按集群里的真实规格过配额并写回缓存

This commit is contained in:
Lemon-miaow committed 2026-09-26 23:28:18 +08:00
1 parent a60ff31ef6
commit d6e1a6464c
6 files changed
+169 -17

No files matched your search

+74 -5
View File
@@ -16,6 +16,7 @@ import (
"felis.lolicon.best/internal/apis/felis/v1alpha1" "felis.lolicon.best/internal/apis/felis/v1alpha1"
corev1 "k8s.io/api/core/v1" corev1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/api/resource"
) )
const testRoot = "mc.example.net" // neutral; never a deployment domain const testRoot = "mc.example.net" // neutral; never a deployment domain
@@ -47,6 +48,8 @@ type fakeRepo struct {
// UpdateServerResources records the write for assertions. // UpdateServerResources records the write for assertions.
serverResources map[string]ResourceSpec serverResources map[string]ResourceSpec
resourceUpdates map[string]ResourceSpec resourceUpdates map[string]ResourceSpec
// quotaChecked records the size every QuotaCheck was asked about, in order.
quotaChecked []ResourceSpec
audits []AuditEntry audits []AuditEntry
failAudit error // Audit fails with it (a store outage) failAudit error // Audit fails with it (a store outage)
// backupRequested mirrors the newest backup.create audit row per server, // backupRequested mirrors the newest backup.create audit row per server,
@@ -323,7 +326,8 @@ func (f *fakeRepo) ServerByName(_ context.Context, n string) (*ServerRecord, err
func (f *fakeRepo) IsLinked(_ context.Context, u string) (bool, error) { return f.linked[u], nil } func (f *fakeRepo) IsLinked(_ context.Context, u string) (bool, error) { return f.linked[u], nil }
func (f *fakeRepo) QuotaAvailable(_ context.Context, u string) (bool, error) { return f.quota[u], nil } func (f *fakeRepo) QuotaAvailable(_ context.Context, u string) (bool, error) { return f.quota[u], nil }
func (f *fakeRepo) QuotaCheck(_ context.Context, userID string, _ string, _ ResourceSpec) (bool, error) { func (f *fakeRepo) QuotaCheck(_ context.Context, userID string, _ string, incoming ResourceSpec) (bool, error) {
f.quotaChecked = append(f.quotaChecked, incoming)
// For hermetic tests, QuotaCheck delegates to the same QuotaAvailable // For hermetic tests, QuotaCheck delegates to the same QuotaAvailable
// store — tests that care about per-dimension checks should use // store — tests that care about per-dimension checks should use
// fakeQuotas with direct inspection. // fakeQuotas with direct inspection.
@@ -2400,7 +2404,7 @@ func TestClaimStateMachine(t *testing.T) {
t.Run("over quota -> 403", func(t *testing.T) { t.Run("over quota -> 403", func(t *testing.T) {
repo := newFakeRepo() repo := newFakeRepo()
repo.linked["u1"] = true repo.linked["u1"] = true
api := newTestAPI(repo, newFakeCluster()) api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user} api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil) w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" { if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" {
@@ -2415,7 +2419,7 @@ func TestClaimStateMachine(t *testing.T) {
repo.quota["u1"] = true repo.quota["u1"] = true
repo.claimOK["survival"] = true repo.claimOK["survival"] = true
repo.claimQuotaRefuse["survival"] = true repo.claimQuotaRefuse["survival"] = true
api := newTestAPI(repo, newFakeCluster()) api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user} api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil) w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" { if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" {
@@ -2427,7 +2431,7 @@ func TestClaimStateMachine(t *testing.T) {
repo.linked["u1"] = true repo.linked["u1"] = true
repo.quota["u1"] = true repo.quota["u1"] = true
repo.claimOK["survival"] = false // row exists but owner already set repo.claimOK["survival"] = false // row exists but owner already set
api := newTestAPI(repo, newFakeCluster()) api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user} api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil) w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusConflict || decodeErr(t, w) != "already_claimed" { if w.Code != http.StatusConflict || decodeErr(t, w) != "already_claimed" {
@@ -2439,7 +2443,7 @@ func TestClaimStateMachine(t *testing.T) {
repo.linked["u1"] = true repo.linked["u1"] = true
repo.quota["u1"] = true repo.quota["u1"] = true
repo.claimOK["survival"] = true repo.claimOK["survival"] = true
api := newTestAPI(repo, newFakeCluster()) api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user} api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil) w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusOK { if w.Code != http.StatusOK {
@@ -2449,6 +2453,71 @@ func TestClaimStateMachine(t *testing.T) {
t.Fatalf("audit not written as expected: %+v", repo.audits) t.Fatalf("audit not written as expected: %+v", repo.audits)
} }
}) })
// A world the reaper released used to have its resource cache zeroed, and the
// quota gate read the cache: the claim passed every resource cap and the server
// then counted as nothing. The claim reads the server's size off the cluster,
// gates on it, and writes it back to the cache.
t.Run("gated on the server's real size, whatever the cache says", func(t *testing.T) {
repo := newFakeRepo()
repo.linked["u1"] = true
repo.claimOK["survival"] = true
repo.serverResources["survival"] = ResourceSpec{}
api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" {
t.Fatalf("code = %d body %s, want 403 quota_exceeded", w.Code, w.Body.String())
}
want := ResourceSpec{CPUMilli: 2000, MemoryMB: 4096, StorageMB: 10240}
if len(repo.quotaChecked) != 1 || repo.quotaChecked[0] != want {
t.Errorf("quota checked against %+v, want [%+v]", repo.quotaChecked, want)
}
if got := repo.resourceUpdates["survival"]; got != want {
t.Errorf("resource cache = %+v, want %+v", got, want)
}
})
t.Run("the internal face is gated the same way", func(t *testing.T) {
repo := newFakeRepo()
repo.links[menuUUID] = "u1"
repo.claimOK["survival"] = true
api := newTestAPI(repo, claimCluster())
w := internalClaim(api, `{"mc_uuid":"`+menuUUID+`"}`)
if w.Code != http.StatusForbidden || decodeErr(t, w) != "quota_exceeded" {
t.Fatalf("code = %d body %s, want 403 quota_exceeded", w.Code, w.Body.String())
}
want := ResourceSpec{CPUMilli: 2000, MemoryMB: 4096, StorageMB: 10240}
if len(repo.quotaChecked) != 1 || repo.quotaChecked[0] != want {
t.Errorf("quota checked against %+v, want [%+v]", repo.quotaChecked, want)
}
if got := repo.resourceUpdates["survival"]; got != want {
t.Errorf("resource cache = %+v, want %+v", got, want)
}
})
t.Run("a server the cluster does not have -> 404", func(t *testing.T) {
repo := newFakeRepo()
repo.linked["u1"] = true
repo.quota["u1"] = true
repo.claimOK["survival"] = true
api := newTestAPI(repo, newFakeCluster())
api.External = staticExternal{p: user}
w := do(api.ExternalHandler(), "POST", "/api/v1/servers/survival/claim", "", nil)
if w.Code != http.StatusNotFound {
t.Fatalf("code = %d body %s, want 404", w.Code, w.Body.String())
}
if len(repo.audits) != 0 {
t.Fatalf("a refused claim was audited: %+v", repo.audits)
}
})
}
// claimCluster holds the survival server a claim test claims: 2 CPUs, 4Gi of memory
// and a 10Gi world.
func claimCluster() *fakeCluster {
cl := newFakeCluster()
cl.byName["survival"] = &ServerInfo{Name: "survival", StorageSize: "10Gi", Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{corev1.ResourceCPU: resource.MustParse("2"), corev1.ResourceMemory: resource.MustParse("4Gi")},
}}
return cl
} }
// ---- wake (autostartPolicy gate + cooldown) ---- // ---- wake (autostartPolicy gate + cooldown) ----
+1 -1
View File
@@ -34,7 +34,7 @@ func TestAccountLinkVertical(t *testing.T) {
repo.quota["u1"] = true repo.quota["u1"] = true
repo.claimOK["survival"] = true repo.claimOK["survival"] = true
api := newTestAPI(repo, newFakeCluster()) api := newTestAPI(repo, claimCluster())
api.External = staticExternal{p: user} api.External = staticExternal{p: user}
ih := api.InternalHandler() ih := api.InternalHandler()
eh := api.ExternalHandler() eh := api.ExternalHandler()
+4 -3
View File
@@ -251,10 +251,11 @@ func (a *API) handleInternalClaim(w http.ResponseWriter, r *http.Request) {
} }
// ② quota gate, evaluated before the ownership write (mirrors handleClaim). // ② quota gate, evaluated before the ownership write (mirrors handleClaim).
// All four dimensions (servers, CPU, memory, storage) are checked. // All four dimensions (servers, CPU, memory, storage) are checked, with the
res, err := a.Repo.ServerResources(r.Context(), name) // server at its real size.
res, err := a.claimResources(r.Context(), name)
if err != nil { if err != nil {
writeError(w, r, err) a.writeLookupError(w, r, err)
return return
} }
ok, err := a.Repo.QuotaCheck(r.Context(), userID, "", res) ok, err := a.Repo.QuotaCheck(r.Context(), userID, "", res)
+27 -4
View File
@@ -138,11 +138,12 @@ func (a *API) handleClaim(w http.ResponseWriter, r *http.Request) {
// ② quota gate, evaluated before the ownership write. All four dimensions // ② quota gate, evaluated before the ownership write. All four dimensions
// (servers, CPU, memory, storage) are checked against the user's quota caps // (servers, CPU, memory, storage) are checked against the user's quota caps
// using the PG-resident resource cache (spec §9.3, §22). The server being // (spec §9.3, §22), with the server at its real size (claimResources). The
// claimed has owner_id=NULL so it is not yet in the per-owner aggregate. // server being claimed has owner_id=NULL so it is not yet in the per-owner
res, err := a.Repo.ServerResources(r.Context(), name) // aggregate.
res, err := a.claimResources(r.Context(), name)
if err != nil { if err != nil {
writeError(w, r, err) a.writeLookupError(w, r, err)
return return
} }
ok, err := a.Repo.QuotaCheck(r.Context(), p.UserID, "", res) ok, err := a.Repo.QuotaCheck(r.Context(), p.UserID, "", res)
@@ -178,6 +179,28 @@ func (a *API) handleClaim(w http.ResponseWriter, r *http.Request) {
writeJSON(w, http.StatusOK, map[string]any{"name": name, "claimed": true}) writeJSON(w, http.StatusOK, map[string]any{"name": name, "claimed": true})
} }
// claimResources is the size a claim is gated on and then counted at: the server's
// spec as the cluster holds it, written through to the resource cache first. The
// cache is all the per-owner quota sums read, and a world the reaper released
// before it kept the cache holds zeros there: gated on those, a claim passed every
// resource cap and the server went uncounted for as long as its new owner kept it.
func (a *API) claimResources(ctx context.Context, name string) (ResourceSpec, error) {
info, err := a.Cluster.GetServer(ctx, name)
if err != nil {
return ResourceSpec{}, err
}
storage, _ := resource.ParseQuantity(info.StorageSize)
res := ResourceSpec{
CPUMilli: quantityToMilli(info.Resources.Limits[corev1.ResourceCPU]),
MemoryMB: quantityToMB(info.Resources.Limits[corev1.ResourceMemory]),
StorageMB: quantityToMB(storage),
}
if err := a.Repo.UpdateServerResources(ctx, name, res.CPUMilli, res.MemoryMB, res.StorageMB); err != nil {
return ResourceSpec{}, err
}
return res, nil
}
// handleStatus returns the CRD status view (spec §7 GET /servers/{name}/status). // handleStatus returns the CRD status view (spec §7 GET /servers/{name}/status).
func (a *API) handleStatus(w http.ResponseWriter, r *http.Request) { func (a *API) handleStatus(w http.ResponseWriter, r *http.Request) {
name := r.PathValue("name") name := r.PathValue("name")
+58
View File
@@ -5,6 +5,7 @@ package pgint
import ( import (
"context" "context"
"database/sql" "database/sql"
"errors"
"strings" "strings"
"testing" "testing"
"time" "time"
@@ -392,6 +393,63 @@ func TestBackupReadBack(t *testing.T) {
} }
} }
// TestReleasedWorldKeepsItsSize: a world the reaper released keeps its resource
// cache, so the next claim is gated on the server's size and counts it. Zeroed on
// release, it let a user whose quota is spent claim the server anyway and then
// counted it as nothing.
func TestReleasedWorldKeepsItsSize(t *testing.T) {
ctx := context.Background()
st := reaper.NewPGStore(db)
name := "released-" + suffix(t)
if _, err := db.ExecContext(ctx,
`INSERT INTO servers (name, cached_cpu_milli, cached_memory_mb, cached_storage_mb) VALUES ($1, 2000, 4096, 10240)`,
name); err != nil {
t.Fatalf("seed server: %v", err)
}
first := newUser(t, "user", "released-a")
if ok, err := repo.ClaimServer(ctx, name, first.ID); err != nil || !ok {
t.Fatalf("first claim = (%v, %v)", ok, err)
}
if err := st.ReleaseWorld(ctx, name, time.Now()); err != nil {
t.Fatalf("ReleaseWorld: %v", err)
}
var owner sql.NullString
var cpu, mem, stor int
if err := db.QueryRowContext(ctx,
`SELECT owner_id, cached_cpu_milli, cached_memory_mb, cached_storage_mb FROM servers WHERE name = $1`, name).
Scan(&owner, &cpu, &mem, &stor); err != nil {
t.Fatal(err)
}
if owner.Valid || cpu != 2000 || mem != 4096 || stor != 10240 {
t.Fatalf("after ReleaseWorld: owner=%v cache=%d/%d/%d; want no owner, cache 2000/4096/10240", owner, cpu, mem, stor)
}
// One CPU of quota does not cover a two-CPU server.
small := newUser(t, "user", "released-b")
oneCPU := 1000
if _, err := repo.SetQuotas(ctx, small.ID, api.QuotaInput{MaxCPUMilli: &oneCPU}, "pgint"); err != nil {
t.Fatalf("SetQuotas: %v", err)
}
if ok, err := repo.ClaimServer(ctx, name, small.ID); !errors.Is(err, api.ErrQuotaExceeded) {
t.Fatalf("claim over the CPU quota = (%v, %v), want ErrQuotaExceeded", ok, err)
}
// A claim that fits counts the server at its size.
fits := newUser(t, "user", "released-c")
if ok, err := repo.ClaimServer(ctx, name, fits.ID); err != nil || !ok {
t.Fatalf("claim within quota = (%v, %v)", ok, err)
}
var used int
if err := db.QueryRowContext(ctx,
`SELECT COALESCE(SUM(cached_cpu_milli), 0) FROM servers WHERE owner_id = $1 AND deleted_at IS NULL`, fits.ID).
Scan(&used); err != nil {
t.Fatal(err)
}
if used != 2000 {
t.Fatalf("the new owner's CPU use = %d, want 2000", used)
}
}
// TestRestartClock: an idle server with no world to reclaim gets its clock and // TestRestartClock: an idle server with no world to reclaim gets its clock and
// warnings reset, and keeps its owner and resource cache (data-durability-19). // warnings reset, and keeps its owner and resource cache (data-durability-19).
func TestRestartClock(t *testing.T) { func TestRestartClock(t *testing.T) {
+5 -4
View File
@@ -79,12 +79,13 @@ func (s *PGStore) InsertBackup(ctx context.Context, rec BackupRecord) error {
return err return err
} }
// ReleaseWorld releases ownership, zeros the resource cache, and resets the // ReleaseWorld releases ownership and resets the activity clock and warnings —
// activity clock and warnings — without deleting the row (red line ②). // without deleting the row (red line ②). The resource cache stays: the server
// keeps its spec, an ownerless row is in nobody's quota sum, and the next claim is
// gated on that size and counts it.
func (s *PGStore) ReleaseWorld(ctx context.Context, name string, at time.Time) error { func (s *PGStore) ReleaseWorld(ctx context.Context, name string, at time.Time) error {
const q = `UPDATE servers const q = `UPDATE servers
SET owner_id = NULL, cached_cpu_milli = 0, cached_memory_mb = 0, cached_storage_mb = 0, SET owner_id = NULL, last_active_at = $2, warned_3d_at = NULL, warned_1d_at = NULL
last_active_at = $2, warned_3d_at = NULL, warned_1d_at = NULL
WHERE name = $1 AND deleted_at IS NULL` WHERE name = $1 AND deleted_at IS NULL`
_, err := s.db.ExecContext(ctx, q, name, at) _, err := s.db.ExecContext(ctx, q, name, at)
return err return err