IM2/Slurm: Untested fixes for several issues

#4746 and #4747

Author: DanThrane

provider-integration/im2/pkg/im/controller/connection.go

@@ -12,13 +12,13 @@ import (
 
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promauto"
+	"ucloud.dk/pkg/im/ipc"
 	"ucloud.dk/shared/pkg/apm"
 	db "ucloud.dk/shared/pkg/database"
-	"ucloud.dk/pkg/im/ipc"
 
-	"ucloud.dk/shared/pkg/client"
 	cfg "ucloud.dk/pkg/im/config"
 	"ucloud.dk/pkg/im/gateway"
+	"ucloud.dk/shared/pkg/client"
 	"ucloud.dk/shared/pkg/log"
 	"ucloud.dk/shared/pkg/util"
 )
@@ -176,27 +176,47 @@ func RegisterConnectionCompleteEx(username string, uid uint32, notifyUCloud bool
 	return err
 }
 
-func RemoveConnection(uid uint32, notifyUCloud bool) error {
+type RemoveConnectionFlag int
+
+const (
+	RemoveConnectionNotify RemoveConnectionFlag = 1 << iota
+	RemoveConnectionTrulyRemove
+)
+
+func RemoveConnection(uid uint32, flags RemoveConnectionFlag) error {
 	ucloud, ok, _ := MapLocalToUCloud(uid)
 	if !ok {
 		return fmt.Errorf("unknown user supplied: %v", uid)
 	}
 
 	db.NewTx0(func(tx *db.Transaction) {
-		db.Exec(
-			tx,
-			`
-				update connections
-				set expires_at = now()
-				where uid = :uid
-		    `,
-			db.Params{
-				"uid": uid,
-			},
-		)
+		if flags&RemoveConnectionTrulyRemove == 0 {
+			db.Exec(
+				tx,
+				`
+					update connections
+					set expires_at = now()
+					where uid = :uid
+				`,
+				db.Params{
+					"uid": uid,
+				},
+			)
+		} else {
+			db.Exec(
+				tx,
+				`
+					delete from connections
+					where uid = :uid
+				`,
+				db.Params{
+					"uid": uid,
+				},
+			)
+		}
 	})
 
-	if notifyUCloud {
+	if flags&RemoveConnectionNotify != 0 {
 		type Req struct {
 			Username string `json:"username"`
 		}
@@ -450,7 +470,7 @@ func controllerConnection(mux *http.ServeMux) {
 					}
 				}
 
-				err := RemoveConnection(local, false)
+				err := RemoveConnection(local, 0)
 				sendResponseOrError(w, util.EmptyValue, err)
 			}),
 		)

provider-integration/im2/pkg/im/external/gpfs/mock_server.go

@@ -7,12 +7,12 @@ import (
 	"io"
 	"net/http"
 	"os"
-	"os/user"
 	"path/filepath"
 	"strconv"
 	"strings"
 	"sync"
 	"time"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/shared/pkg/log"
 	"ucloud.dk/shared/pkg/util"
 )

provider-integration/im2/pkg/im/external/user/00_module.go

@@ -0,0 +1,87 @@
+package user
+
+import (
+	"fmt"
+	"os"
+	"strings"
+	"ucloud.dk/shared/pkg/util"
+)
+
+// NOTE(Dan): This package provides a "getent" based alternative to Go's builtin user package. The API is largely
+// compatible, but doesn't support all fields. This alternative is needed because of Go's alternative implementation
+// when CGO is disabled for a given build. In these cases, Go will fall back to reading from /etc/passwd and /etc/group
+// which is not sufficient for our use-cases, in particular with the Slurm integration. In these cases we often need
+// to read from NSS. AS a result, we use the "getent" executable which is, more or less, omnipresent on the systems we
+// are targeting.
+//
+// This API differs mostly by the fact that structs are returned directly instead of an interface (reference type). This
+// means that it is not possible to check if a UserInfo or GroupInfo is nil. Instead, you must look at the error being
+// returned. This, however, has the added benefit that you are very unlikely to accidentally do a nil-dereference.
+
+type UserInfo struct {
+	Uid      string
+	Gid      string
+	Username string
+}
+
+type GroupInfo struct {
+	Gid  string
+	Name string
+}
+
+func Current() (UserInfo, error) {
+	return LookupId(fmt.Sprintf("%d", os.Getuid()))
+}
+
+func Lookup(username string) (UserInfo, error) {
+	return lookupPasswd(username)
+}
+
+func LookupId(uid string) (UserInfo, error) {
+	return lookupPasswd(uid)
+}
+
+func lookupPasswd(query string) (UserInfo, error) {
+	stdout, _, ok := util.RunCommand([]string{getent, "passwd", query})
+	if !ok {
+		return UserInfo{}, fmt.Errorf("unknown user: %s", query)
+	}
+
+	toks := strings.Split(stdout, ":")
+	if len(toks) != 7 {
+		return UserInfo{}, fmt.Errorf("unable to read output of getent: %s", stdout)
+	}
+
+	return UserInfo{
+		Uid:      toks[2],
+		Gid:      toks[3],
+		Username: toks[0],
+	}, nil
+}
+
+func LookupGroup(groupName string) (GroupInfo, error) {
+	return lookupGroup(groupName)
+}
+
+func LookupGroupId(gid string) (GroupInfo, error) {
+	return lookupGroup(gid)
+}
+
+func lookupGroup(query string) (GroupInfo, error) {
+	stdout, _, ok := util.RunCommand([]string{getent, "group", query})
+	if !ok {
+		return GroupInfo{}, fmt.Errorf("unknown user: %s", query)
+	}
+
+	toks := strings.Split(stdout, ":")
+	if len(toks) != 4 {
+		return GroupInfo{}, fmt.Errorf("unable to read output of getent: %s", stdout)
+	}
+
+	return GroupInfo{
+		Gid:  toks[0],
+		Name: toks[2],
+	}, nil
+}
+
+const getent = "getent"

provider-integration/im2/pkg/im/launcher/module.go

@@ -4,8 +4,8 @@ import (
 	"fmt"
 	"net/http"
 	"os"
-	"os/user"
 	"path/filepath"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/im/migrations"
 
 	"github.com/jmoiron/sqlx"

provider-integration/im2/pkg/im/services/idfreeipa/projects.go

@@ -2,11 +2,11 @@ package idfreeipa
 
 import (
 	"fmt"
-	"os/user"
 	"strconv"
-	fnd "ucloud.dk/shared/pkg/foundation"
 	ctrl "ucloud.dk/pkg/im/controller"
 	"ucloud.dk/pkg/im/external/freeipa"
+	"ucloud.dk/pkg/im/external/user"
+	fnd "ucloud.dk/shared/pkg/foundation"
 	"ucloud.dk/shared/pkg/log"
 	"ucloud.dk/shared/pkg/util"
 )

provider-integration/im2/pkg/im/services/slurm/account_mapper.go

@@ -3,16 +3,16 @@ package slurm
 import (
 	"fmt"
 	"net/http"
-	"os/user"
 	"regexp"
 	"strconv"
 	"time"
-	"ucloud.dk/shared/pkg/apm"
-	db "ucloud.dk/shared/pkg/database"
 	cfg "ucloud.dk/pkg/im/config"
 	ctrl "ucloud.dk/pkg/im/controller"
 	slurmcli "ucloud.dk/pkg/im/external/slurm"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/im/ipc"
+	"ucloud.dk/shared/pkg/apm"
+	db "ucloud.dk/shared/pkg/database"
 	"ucloud.dk/shared/pkg/log"
 	"ucloud.dk/shared/pkg/util"
 )

provider-integration/im2/pkg/im/services/slurm/cli_allocations.go

@@ -5,17 +5,17 @@ import (
 	"fmt"
 	"net/http"
 	"os"
-	"os/user"
 	"regexp"
 	"strings"
 	"time"
-	"ucloud.dk/shared/pkg/apm"
 	"ucloud.dk/pkg/cli"
-	db "ucloud.dk/shared/pkg/database"
-	fnd "ucloud.dk/shared/pkg/foundation"
 	cfg "ucloud.dk/pkg/im/config"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/im/ipc"
 	"ucloud.dk/pkg/termio"
+	"ucloud.dk/shared/pkg/apm"
+	db "ucloud.dk/shared/pkg/database"
+	fnd "ucloud.dk/shared/pkg/foundation"
 	"ucloud.dk/shared/pkg/util"
 )
 

provider-integration/im2/pkg/im/services/slurm/cli_connect.go

@@ -3,9 +3,9 @@ package slurm
 import (
 	"fmt"
 	"os"
-	"os/user"
 	"time"
 	ctrl "ucloud.dk/pkg/im/controller"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/termio"
 	"ucloud.dk/shared/pkg/util"
 )

provider-integration/im2/pkg/im/services/slurm/cli_drives.go

@@ -6,15 +6,15 @@ import (
 	"fmt"
 	"net/http"
 	"os"
-	"os/user"
 	"regexp"
 	"strings"
 	"ucloud.dk/pkg/cli"
-	db "ucloud.dk/shared/pkg/database"
 	ctrl "ucloud.dk/pkg/im/controller"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/im/ipc"
-	orc "ucloud.dk/shared/pkg/orchestrators"
 	"ucloud.dk/pkg/termio"
+	db "ucloud.dk/shared/pkg/database"
+	orc "ucloud.dk/shared/pkg/orchestrators"
 )
 
 func HandleDrivesCommand() {

provider-integration/im2/pkg/im/services/slurm/cli_jobs.go

@@ -6,17 +6,17 @@ import (
 	"fmt"
 	"net/http"
 	"os"
-	"os/user"
 	"regexp"
 	"slices"
 	"strings"
 	"time"
 	"ucloud.dk/pkg/cli"
-	db "ucloud.dk/shared/pkg/database"
 	slurmcli "ucloud.dk/pkg/im/external/slurm"
+	"ucloud.dk/pkg/im/external/user"
 	"ucloud.dk/pkg/im/ipc"
-	orc "ucloud.dk/shared/pkg/orchestrators"
 	"ucloud.dk/pkg/termio"
+	db "ucloud.dk/shared/pkg/database"
+	orc "ucloud.dk/shared/pkg/orchestrators"
 )
 
 func HandleJobsCommand() {