VAULT-11829: Add cluster status handler (#18351)

* go get link proto @vault-11829-meta-get-cluster-status

* add HA status

* add HAEnabled method

* add raft config

* allocate HA nodes based on actual count

* add raft autopilot status

* add raft quorum warnings

* add ClusterID method

* add StorageType

* add ClusterID

* update github.com/hashicorp/vault/vault/hcp_link/proto

* add changelog entry

* fix raft config panic

* remove "Warning" quorum message prefix

* add error wrapping

* add Core.HAStateWithLock method

* reduce quorum warnings to single string

* fix HCP_API_HOST test env var check

* Revert "fix HCP_API_HOST test env var check"

This reverts commit 97c73c4.

changelog/18351.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+hcp/status: Add cluster-level status information 
+```

go.mod

@@ -138,7 +138,7 @@ require (
 	github.com/hashicorp/vault/api/auth/approle v0.1.0
 	github.com/hashicorp/vault/api/auth/userpass v0.1.0
 	github.com/hashicorp/vault/sdk v0.6.1
-	github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20221209165735-a2eed407e08d
+	github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230106203127-9eaf26716342
 	github.com/influxdata/influxdb1-client v0.0.0-20200827194710-b269163b24ab
 	github.com/jackc/pgx/v4 v4.15.0
 	github.com/jcmturner/gokrb5/v8 v8.4.2

go.sum

@@ -1173,6 +1173,16 @@ github.com/hashicorp/vault-testing-stepwise v0.1.2 h1:3obC/ziAPGnsz2IQxr5e4Ayb7t
 github.com/hashicorp/vault-testing-stepwise v0.1.2/go.mod h1:TeU6B+5NqxUjto+Zey+QQEH1iywuHn0ciHZNYh4q3uI=
 github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20221209165735-a2eed407e08d h1:U692VbDl6ww5GQsNFClJVFJDaPeuqtDt1Mwqf21KYek=
 github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20221209165735-a2eed407e08d/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20221213220056-b0613b59f419 h1:yl6f//YTaTTGKJwyOpRe7v1DDPrzP+NErwgnef6qx7A=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20221213220056-b0613b59f419/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230103211812-c28545e74f94 h1:Rx4Q2/mOPqJuanzwZYttDkWjdibPv3UpvsvKmOkl6h4=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230103211812-c28545e74f94/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230105183308-048241517ffb h1:PgXcBszV61BvxD0wZzm4QCz9btgTWX74NO4be6S2afU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230105183308-048241517ffb/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230106184443-96cfe11e7051 h1:cMQoRbIUMhbM0NsmP6hH3S3ZmAPVgic3g3L8Z55rXCI=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230106184443-96cfe11e7051/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230106203127-9eaf26716342 h1:9cMwZnaAV/lKs8EZsvBF00wPt350wD3sg/xqWGeN4gM=
+github.com/hashicorp/vault/vault/hcp_link/proto v0.0.0-20230106203127-9eaf26716342/go.mod h1:a2crHoMWwY6aiL8GWT8hYj7vKD64uX0EdRPbnsHF5wU=
 github.com/hashicorp/vic v1.5.1-0.20190403131502-bbfe86ec9443 h1:O/pT5C1Q3mVXMyuqg7yuAWUg/jMZR1/0QTzTRdNR6Uw=
 github.com/hashicorp/vic v1.5.1-0.20190403131502-bbfe86ec9443/go.mod h1:bEpDU35nTu0ey1EXjwNwPjI9xErAsoOCmcMb9GKvyxo=
 github.com/hashicorp/yamux v0.0.0-20180604194846-3520598351bb/go.mod h1:+NfK9FKeTrX5uv1uIXGdwYDTeHna2qgaIlx54MXqjAM=

vault/cluster.go

@@ -388,3 +388,7 @@ func (c *Core) SetClusterListenerAddrs(addrs []*net.TCPAddr) {
 func (c *Core) SetClusterHandler(handler http.Handler) {
 	c.clusterHandler = handler
 }
+
+func (c *Core) ClusterID() string {
+	return c.clusterID.Load()
+}

vault/core.go

@@ -688,6 +688,13 @@ func (c *Core) HAState() consts.HAState {
 	}
 }
 
+func (c *Core) HAStateWithLock() consts.HAState {
+	c.stateLock.RLock()
+	c.stateLock.RUnlock()
+
+	return c.HAState()
+}
+
 // CoreConfig is used to parameterize a core
 type CoreConfig struct {
 	entCoreConfig
@@ -3699,3 +3706,26 @@ func (c *Core) GetHCPLinkStatus() (string, string) {
 
 	return status, resourceID
 }
+
+func (c *Core) HAEnabled() bool {
+	return c.ha != nil && c.ha.HAEnabled()
+}
+
+func (c *Core) GetRaftConfiguration(ctx context.Context) (*raft.RaftConfigurationResponse, error) {
+	raftBackend := c.getRaftBackend()
+
+	if raftBackend == nil {
+		return nil, nil
+	}
+
+	return raftBackend.GetConfiguration(ctx)
+}
+
+func (c *Core) GetRaftAutopilotState(ctx context.Context) (*raft.AutopilotState, error) {
+	raftBackend := c.getRaftBackend()
+	if raftBackend == nil {
+		return nil, nil
+	}
+
+	return raftBackend.GetAutopilotServerState(ctx)
+}

vault/hcp_link/capabilities/meta/meta.go

@@ -4,12 +4,14 @@ import (
 	"context"
 	"fmt"
 	"math"
+	"os"
 	"sync"
 	"time"
 
 	"github.com/hashicorp/go-hclog"
 	scada "github.com/hashicorp/hcp-scada-provider"
 	"github.com/hashicorp/vault/helper/namespace"
+	"github.com/hashicorp/vault/sdk/helper/consts"
 	"github.com/hashicorp/vault/vault"
 	"github.com/hashicorp/vault/vault/cluster"
 	"github.com/hashicorp/vault/vault/hcp_link/capabilities"
@@ -23,7 +25,7 @@ import (
 type hcpLinkMetaHandler struct {
 	meta.UnimplementedHCPLinkMetaServer
 
-	wrappedCore   internal.WrappedCoreListNamespacesMounts
+	wrappedCore   internal.WrappedCoreMeta
 	scadaProvider scada.SCADAProvider
 	logger        hclog.Logger
 
@@ -129,7 +131,7 @@ func (h *hcpLinkMetaHandler) ListNamespaces(ctx context.Context, req *meta.ListN
 func (h *hcpLinkMetaHandler) ListMounts(ctx context.Context, req *meta.ListMountsRequest) (*meta.ListMountsResponse, error) {
 	mountEntries, err := h.wrappedCore.ListMounts()
 	if err != nil {
-		return nil, err
+		return nil, fmt.Errorf("unable to list secret mounts: %w", err)
 	}
 
 	var mounts []*meta.Mount
@@ -140,7 +142,7 @@ func (h *hcpLinkMetaHandler) ListMounts(ctx context.Context, req *meta.ListMount
 		if nsID != namespace.RootNamespaceID {
 			ns, err := h.wrappedCore.NamespaceByID(ctx, entry.NamespaceID)
 			if err != nil {
-				return nil, err
+				return nil, fmt.Errorf("unable to get namespace associated with secret mount: %w", err)
 			}
 
 			path = ns.Path + path
@@ -161,7 +163,7 @@ func (h *hcpLinkMetaHandler) ListMounts(ctx context.Context, req *meta.ListMount
 func (h *hcpLinkMetaHandler) ListAuths(ctx context.Context, req *meta.ListAuthsRequest) (*meta.ListAuthResponse, error) {
 	authEntries, err := h.wrappedCore.ListAuths()
 	if err != nil {
-		return nil, err
+		return nil, fmt.Errorf("unable to list auth mounts: %w", err)
 	}
 
 	var auths []*meta.Auth
@@ -172,7 +174,7 @@ func (h *hcpLinkMetaHandler) ListAuths(ctx context.Context, req *meta.ListAuthsR
 		if nsID != namespace.RootNamespaceID {
 			ns, err := h.wrappedCore.NamespaceByID(ctx, entry.NamespaceID)
 			if err != nil {
-				return nil, err
+				return nil, fmt.Errorf("unable to get namespace associated with auth mount: %w", err)
 			}
 
 			path = ns.Path + path
@@ -189,3 +191,113 @@ func (h *hcpLinkMetaHandler) ListAuths(ctx context.Context, req *meta.ListAuthsR
 		Auths: auths,
 	}, nil
 }
+
+func (h *hcpLinkMetaHandler) GetClusterStatus(ctx context.Context, req *meta.GetClusterStatusRequest) (*meta.GetClusterStatusResponse, error) {
+	if h.wrappedCore.HAStateWithLock() != consts.Active {
+		return nil, fmt.Errorf("node not active")
+	}
+
+	hostname, err := os.Hostname()
+	if err != nil {
+		return nil, fmt.Errorf("unable to fetch hostname: %w", err)
+	}
+
+	haEnabled := h.wrappedCore.HAEnabled()
+	haStatus := &meta.HAStatus{
+		Enabled: haEnabled,
+	}
+
+	if haEnabled {
+		leader := &meta.HANode{
+			Hostname: hostname,
+		}
+
+		peers := h.wrappedCore.GetHAPeerNodesCached()
+
+		haNodes := make([]*meta.HANode, len(peers)+1)
+		haNodes[0] = leader
+
+		for i, peerNode := range peers {
+			haNodes[i+1] = &meta.HANode{
+				Hostname: peerNode.Hostname,
+			}
+		}
+
+		haStatus.Nodes = haNodes
+	}
+
+	raftStatus := &meta.RaftStatus{}
+	raftConfig, err := h.wrappedCore.GetRaftConfiguration(ctx)
+	if err != nil {
+		return nil, fmt.Errorf("unable to get Raft configuration: %w", err)
+	}
+
+	if raftConfig != nil {
+		raftServers := make([]*meta.RaftServer, len(raftConfig.Servers))
+
+		var voterCount uint32
+		for i, srv := range raftConfig.Servers {
+			raftServers[i] = &meta.RaftServer{
+				NodeID:          srv.NodeID,
+				Address:         srv.Address,
+				Voter:           srv.Voter,
+				Leader:          srv.Leader,
+				ProtocolVersion: srv.ProtocolVersion,
+			}
+
+			if srv.Voter {
+				voterCount++
+			}
+		}
+
+		raftStatus.RaftConfiguration = &meta.RaftConfiguration{
+			Servers: raftServers,
+		}
+
+		evenVoterMessage := "Vault should have access to an odd number of voter nodes."
+		largeClusterMessage := "Very large cluster detected."
+		var quorumWarning string
+
+		if voterCount == 1 {
+			quorumWarning = "Only one server node found. Vault is not running in high availability mode."
+		} else if voterCount%2 == 0 && voterCount > 7 {
+			quorumWarning = evenVoterMessage + " " + largeClusterMessage
+		} else if voterCount%2 == 0 {
+			quorumWarning = evenVoterMessage
+		} else if voterCount > 7 {
+			quorumWarning = largeClusterMessage
+		}
+
+		raftStatus.QuorumWarning = quorumWarning
+	}
+
+	raftAutopilotState, err := h.wrappedCore.GetRaftAutopilotState(ctx)
+	if err != nil {
+		return nil, fmt.Errorf("unable to get Raft Autopilot state: %w", err)
+	}
+
+	if raftAutopilotState != nil {
+		autopilotStatus := &meta.AutopilotStatus{
+			Healthy: raftAutopilotState.Healthy,
+		}
+
+		autopilotServers := make([]*meta.AutopilotServer, 0)
+		for _, srv := range raftAutopilotState.Servers {
+			autopilotServers = append(autopilotServers, &meta.AutopilotServer{
+				ID:      srv.ID,
+				Healthy: srv.Healthy,
+			})
+		}
+
+		raftStatus.AutopilotStatus = autopilotStatus
+	}
+
+	resp := &meta.GetClusterStatusResponse{
+		ClusterID:   h.wrappedCore.ClusterID(),
+		HAStatus:    haStatus,
+		RaftStatus:  raftStatus,
+		StorageType: h.wrappedCore.StorageType(),
+	}
+
+	return resp, nil
+}

vault/hcp_link/internal/wrapped_hcpLink.go

@@ -4,6 +4,7 @@ import (
 	"context"
 
 	"github.com/hashicorp/vault/helper/namespace"
+	"github.com/hashicorp/vault/physical/raft"
 	"github.com/hashicorp/vault/sdk/helper/consts"
 	"github.com/hashicorp/vault/sdk/logical"
 	"github.com/hashicorp/vault/vault"
@@ -30,14 +31,21 @@ type WrappedCoreHCPToken interface {
 
 var _ WrappedCoreHCPToken = &vault.Core{}
 
-type WrappedCoreListNamespacesMounts interface {
+type WrappedCoreMeta interface {
 	NamespaceByID(ctx context.Context, nsID string) (*namespace.Namespace, error)
 	ListNamespaces(includePath bool) []*namespace.Namespace
 	ListMounts() ([]*vault.MountEntry, error)
 	ListAuths() ([]*vault.MountEntry, error)
+	HAEnabled() bool
+	HAStateWithLock() consts.HAState
+	GetHAPeerNodesCached() []vault.PeerNode
+	GetRaftConfiguration(ctx context.Context) (*raft.RaftConfigurationResponse, error)
+	GetRaftAutopilotState(ctx context.Context) (*raft.AutopilotState, error)
+	StorageType() string
+	ClusterID() string
 }
 
-var _ WrappedCoreListNamespacesMounts = &vault.Core{}
+var _ WrappedCoreMeta = &vault.Core{}
 
 type WrappedCoreHCPLinkStatus interface {
 	WrappedCoreStandbyStates

vault/request_handling.go

@@ -796,7 +796,7 @@ func (c *Core) handleCancelableRequest(ctx context.Context, req *logical.Request
 	}
 
 	if walState.LocalIndex != 0 || walState.ReplicatedIndex != 0 {
-		walState.ClusterID = c.clusterID.Load()
+		walState.ClusterID = c.ClusterID()
 		if walState.LocalIndex == 0 {
 			if c.perfStandby {
 				walState.LocalIndex = LastRemoteWAL(c)
@@ -2343,7 +2343,7 @@ func (c *Core) checkSSCTokenInternal(ctx context.Context, token string, isPerfSt
 		return plainToken.Random, nil
 	}
 
-	requiredWalState := &logical.WALState{ClusterID: c.clusterID.Load(), LocalIndex: plainToken.LocalIndex, ReplicatedIndex: 0}
+	requiredWalState := &logical.WALState{ClusterID: c.ClusterID(), LocalIndex: plainToken.LocalIndex, ReplicatedIndex: 0}
 	if c.HasWALState(requiredWalState, isPerfStandby) {
 		return plainToken.Random, nil
 	}