vault/sdk/plugin/grpc_backend_client.go
Alexander Scheel c042e4dae3
Add path based primary write forwarding (PBPWF) - OSS (#18735)
* Add WriteForwardedStorage to sdk's plugin, logical in OSS

This should allow backends to specify paths to forward write
(storage.Put(...) and storage.Delete(...)) operations for.

Notably, these semantics are subject to change and shouldn't yet be
relied on.

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Collect paths for write forwarding in OSS

This adds a path manager to Core, allowing tracking across all Vault
versions of paths which could use write forwarding if available. In
particular, even on OSS offerings, we'll need to template {{clusterId}}
into the paths, in the event of later upgrading to Enterprise. If we
didn't, we'd end up writing paths which will no longer be accessible
post-migration, due to write forwarding now replacing the sentinel with
the actual cluster identifier.

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Add forwarded writer implementation to OSS

Here, for paths given to us, we determine if we need to do cluster
translation and perform local writing. This is the OSS variant.

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Wire up mount-specific request forwarding in OSS

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Clarify that state lock needs to be held to call HAState in OSS

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Move cluster sentinel constant to sdk/logical

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Expose ClusterID to Plugins via SystemView

This will let plugins learn what the Cluster's ID is, without having to
resort to hacks like writing a random string to its cluster-prefixed
namespace and then reading it once it has replicated.

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

* Add GRPC ClusterID implementation

For any external plugins which wish to use it.

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>

Signed-off-by: Alexander Scheel <alex.scheel@hashicorp.com>
2023-01-20 16:36:18 -05:00

300 lines
7.7 KiB
Go

package plugin
import (
"context"
"errors"
"math"
"sync/atomic"
log "github.com/hashicorp/go-hclog"
"github.com/hashicorp/go-plugin"
"github.com/hashicorp/vault/sdk/helper/pluginutil"
"github.com/hashicorp/vault/sdk/logical"
"github.com/hashicorp/vault/sdk/plugin/pb"
"google.golang.org/grpc"
"google.golang.org/grpc/codes"
"google.golang.org/grpc/status"
)
var (
ErrPluginShutdown = errors.New("plugin is shut down")
ErrClientInMetadataMode = errors.New("plugin client can not perform action while in metadata mode")
)
// Validate backendGRPCPluginClient satisfies the logical.Backend interface
var _ logical.Backend = &backendGRPCPluginClient{}
// backendPluginClient implements logical.Backend and is the
// go-plugin client.
type backendGRPCPluginClient struct {
broker *plugin.GRPCBroker
client pb.BackendClient
versionClient logical.PluginVersionClient
metadataMode bool
system logical.SystemView
logger log.Logger
// This is used to signal to the Cleanup function that it can proceed
// because we have a defined server
cleanupCh chan struct{}
// server is the grpc server used for serving storage and sysview requests.
server *atomic.Value
doneCtx context.Context
}
func (b *backendGRPCPluginClient) Initialize(ctx context.Context, _ *logical.InitializationRequest) error {
if b.metadataMode {
return nil
}
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
reply, err := b.client.Initialize(ctx, &pb.InitializeArgs{}, largeMsgGRPCCallOpts...)
if err != nil {
if b.doneCtx.Err() != nil {
return ErrPluginShutdown
}
// If the plugin doesn't have Initialize implemented we should not fail
// the initialize call; otherwise this could halt startup of vault.
grpcStatus, ok := status.FromError(err)
if ok && grpcStatus.Code() == codes.Unimplemented {
return nil
}
return err
}
if reply.Err != nil {
return pb.ProtoErrToErr(reply.Err)
}
return nil
}
func (b *backendGRPCPluginClient) HandleRequest(ctx context.Context, req *logical.Request) (*logical.Response, error) {
if b.metadataMode {
return nil, ErrClientInMetadataMode
}
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
protoReq, err := pb.LogicalRequestToProtoRequest(req)
if err != nil {
return nil, err
}
reply, err := b.client.HandleRequest(ctx, &pb.HandleRequestArgs{
Request: protoReq,
}, largeMsgGRPCCallOpts...)
if err != nil {
if b.doneCtx.Err() != nil {
return nil, ErrPluginShutdown
}
return nil, err
}
resp, err := pb.ProtoResponseToLogicalResponse(reply.Response)
if err != nil {
return nil, err
}
if reply.Err != nil {
return resp, pb.ProtoErrToErr(reply.Err)
}
return resp, nil
}
func (b *backendGRPCPluginClient) SpecialPaths() *logical.Paths {
reply, err := b.client.SpecialPaths(b.doneCtx, &pb.Empty{})
if err != nil {
return nil
}
if reply.Paths == nil {
return nil
}
return &logical.Paths{
Root: reply.Paths.Root,
Unauthenticated: reply.Paths.Unauthenticated,
LocalStorage: reply.Paths.LocalStorage,
SealWrapStorage: reply.Paths.SealWrapStorage,
WriteForwardedStorage: reply.Paths.WriteForwardedStorage,
}
}
// System returns vault's system view. The backend client stores the view during
// Setup, so there is no need to shim the system just to get it back.
func (b *backendGRPCPluginClient) System() logical.SystemView {
return b.system
}
// Logger returns vault's logger. The backend client stores the logger during
// Setup, so there is no need to shim the logger just to get it back.
func (b *backendGRPCPluginClient) Logger() log.Logger {
return b.logger
}
func (b *backendGRPCPluginClient) HandleExistenceCheck(ctx context.Context, req *logical.Request) (bool, bool, error) {
if b.metadataMode {
return false, false, ErrClientInMetadataMode
}
protoReq, err := pb.LogicalRequestToProtoRequest(req)
if err != nil {
return false, false, err
}
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
reply, err := b.client.HandleExistenceCheck(ctx, &pb.HandleExistenceCheckArgs{
Request: protoReq,
}, largeMsgGRPCCallOpts...)
if err != nil {
if b.doneCtx.Err() != nil {
return false, false, ErrPluginShutdown
}
return false, false, err
}
if reply.Err != nil {
return false, false, pb.ProtoErrToErr(reply.Err)
}
return reply.CheckFound, reply.Exists, nil
}
func (b *backendGRPCPluginClient) Cleanup(ctx context.Context) {
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
b.client.Cleanup(ctx, &pb.Empty{})
// This will block until Setup has run the function to create a new server
// in b.server. If we stop here before it has a chance to actually start
// listening, when it starts listening it will immediately error out and
// exit, which is fine. Overall this ensures that we do not miss stopping
// the server if it ends up being created after Cleanup is called.
<-b.cleanupCh
server := b.server.Load()
if server != nil {
server.(*grpc.Server).GracefulStop()
}
}
func (b *backendGRPCPluginClient) InvalidateKey(ctx context.Context, key string) {
if b.metadataMode {
return
}
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
b.client.InvalidateKey(ctx, &pb.InvalidateKeyArgs{
Key: key,
})
}
func (b *backendGRPCPluginClient) Setup(ctx context.Context, config *logical.BackendConfig) error {
// Shim logical.Storage
storageImpl := config.StorageView
if b.metadataMode {
storageImpl = &NOOPStorage{}
}
storage := &GRPCStorageServer{
impl: storageImpl,
}
// Shim logical.SystemView
sysViewImpl := config.System
if b.metadataMode {
sysViewImpl = &logical.StaticSystemView{}
}
sysView := &gRPCSystemViewServer{
impl: sysViewImpl,
}
// Register the server in this closure.
serverFunc := func(opts []grpc.ServerOption) *grpc.Server {
opts = append(opts, grpc.MaxRecvMsgSize(math.MaxInt32))
opts = append(opts, grpc.MaxSendMsgSize(math.MaxInt32))
s := grpc.NewServer(opts...)
pb.RegisterSystemViewServer(s, sysView)
pb.RegisterStorageServer(s, storage)
b.server.Store(s)
close(b.cleanupCh)
return s
}
brokerID := b.broker.NextId()
go b.broker.AcceptAndServe(brokerID, serverFunc)
args := &pb.SetupArgs{
BrokerID: brokerID,
Config: config.Config,
BackendUUID: config.BackendUUID,
}
ctx, cancel := context.WithCancel(ctx)
quitCh := pluginutil.CtxCancelIfCanceled(cancel, b.doneCtx)
defer close(quitCh)
defer cancel()
reply, err := b.client.Setup(ctx, args)
if err != nil {
return err
}
if reply.Err != "" {
return errors.New(reply.Err)
}
// Set system and logger for getter methods
b.system = config.System
b.logger = config.Logger
return nil
}
func (b *backendGRPCPluginClient) Type() logical.BackendType {
reply, err := b.client.Type(b.doneCtx, &pb.Empty{})
if err != nil {
return logical.TypeUnknown
}
return logical.BackendType(reply.Type)
}
func (b *backendGRPCPluginClient) PluginVersion() logical.PluginVersion {
reply, err := b.versionClient.Version(b.doneCtx, &logical.Empty{})
if err != nil {
if stErr, ok := status.FromError(err); ok {
if stErr.Code() == codes.Unimplemented {
return logical.EmptyPluginVersion
}
}
b.Logger().Warn("Unknown error getting plugin version", "err", err)
return logical.EmptyPluginVersion
}
return logical.PluginVersion{
Version: reply.GetPluginVersion(),
}
}
func (b *backendGRPCPluginClient) IsExternal() bool {
return true
}