Fix v18 tablets removed from healthcheck when topo server GetTablet c…

…all fails

Signed-off-by: Brendan Dougherty <[email protected]>

go/vt/discovery/topology_watcher.go

@@ -190,6 +190,17 @@ func (tw *TopologyWatcher) loadTablets() {
 			topologyWatcherOperations.Add(topologyWatcherOpGetTablet, 1)
 			<-tw.sem // Done; enable next request to run
 			if err != nil {
+				if !topo.IsErrType(err, topo.NoNode) {
+					// We failed to get the tablet, but it may still exist.
+					// We don't want this tablet to be removed from the tw.tablets map or the healthcheck,
+					// so we fill the gap in the newTablets map using the existing tablet.
+					tw.mu.Lock()
+					aliasStr := topoproto.TabletAliasString(alias)
+					if val, ok := tw.tablets[aliasStr]; ok {
+						newTablets[aliasStr] = val
+					}
+					tw.mu.Unlock()
+				}
 				topologyWatcherErrors.Add(topologyWatcherOpGetTablet, 1)
 				select {
 				case <-tw.ctx.Done():

go/vt/discovery/topology_watcher_test.go

@@ -18,6 +18,7 @@ package discovery
 
 import (
 	"context"
+	"errors"
 	"math/rand"
 	"testing"
 	"time"
@@ -630,3 +631,145 @@ func TestFilterByKeypsaceSkipsIgnoredTablets(t *testing.T) {
 
 	tw.Stop()
 }
+
+func TestGetTabletErrorDoesNotRemoveFromHealthcheck(t *testing.T) {
+	ctx := utils.LeakCheckContext(t)
+
+	ts, factory := memorytopo.NewServerAndFactory(ctx, "aa")
+	defer ts.Close()
+	fhc := NewFakeHealthCheck(nil)
+	defer fhc.Close()
+	topologyWatcherOperations.ZeroAll()
+	counts := topologyWatcherOperations.Counts()
+	tw := NewCellTabletsWatcher(context.Background(), ts, fhc, nil, "aa", 10*time.Minute, true, 5)
+	defer tw.Stop()
+
+	counts = checkOpCounts(t, counts, map[string]int64{})
+	checkChecksum(t, tw, 0)
+
+	// Add a tablet to the topology.
+	tablet1 := &topodatapb.Tablet{
+		Alias: &topodatapb.TabletAlias{
+			Cell: "aa",
+			Uid:  0,
+		},
+		Hostname: "host1",
+		PortMap: map[string]int32{
+			"vt": 123,
+		},
+		Keyspace: "keyspace",
+		Shard:    "shard",
+	}
+	require.NoError(t, ts.CreateTablet(ctx, tablet1), "CreateTablet failed for %v", tablet1.Alias)
+
+	tw.loadTablets()
+	counts = checkOpCounts(t, counts, map[string]int64{"ListTablets": 1, "GetTablet": 1, "AddTablet": 1})
+	checkChecksum(t, tw, 3238442862)
+
+	// Check the tablet is returned by GetAllTablets().
+	allTablets := fhc.GetAllTablets()
+	key1 := TabletToMapKey(tablet1)
+	assert.Len(t, allTablets, 1)
+	assert.Contains(t, allTablets, key1)
+	assert.True(t, proto.Equal(tablet1, allTablets[key1]))
+
+	// Add a second tablet to the topology.
+	tablet2 := &topodatapb.Tablet{
+		Alias: &topodatapb.TabletAlias{
+			Cell: "aa",
+			Uid:  2,
+		},
+		Hostname: "host2",
+		PortMap: map[string]int32{
+			"vt": 789,
+		},
+		Keyspace: "keyspace",
+		Shard:    "shard",
+	}
+	require.NoError(t, ts.CreateTablet(ctx, tablet2), "CreateTablet failed for %v", tablet2.Alias)
+
+	// Cause the Get for the first tablet to fail.
+	factory.AddOperationError(memorytopo.Get, "tablets/aa-0000000000/Tablet", errors.New("fake error"))
+
+	// Ensure that a topo GetTablet error fails. If not, the rest of this test is invalid.
+	_, err := ts.GetTablet(ctx, tablet1.Alias)
+	require.ErrorContains(t, err, "fake error")
+
+	// Now force the error during loadTablets.
+	tw.loadTablets()
+	checkOpCounts(t, counts, map[string]int64{"ListTablets": 1, "GetTablet": 2, "AddTablet": 1})
+	checkChecksum(t, tw, 2762153755)
+
+	// Ensure the first tablet is still returned by GetAllTablets() and the second tablet has been added.
+	allTablets = fhc.GetAllTablets()
+	key2 := TabletToMapKey(tablet2)
+	assert.Len(t, allTablets, 2)
+	assert.Contains(t, allTablets, key1)
+	assert.Contains(t, allTablets, key2)
+	assert.True(t, proto.Equal(tablet1, allTablets[key1]))
+	assert.True(t, proto.Equal(tablet2, allTablets[key2]))
+}
+
+func TestGetTabletNoNodeErrorRemovesFromHealthcheck(t *testing.T) {
+	ctx := utils.LeakCheckContext(t)
+
+	ts, factory := memorytopo.NewServerAndFactory(ctx, "aa")
+	defer ts.Close()
+	fhc := NewFakeHealthCheck(nil)
+	defer fhc.Close()
+	topologyWatcherOperations.ZeroAll()
+	counts := topologyWatcherOperations.Counts()
+	tw := NewCellTabletsWatcher(context.Background(), ts, fhc, nil, "aa", 10*time.Minute, true, 5)
+	defer tw.Stop()
+
+	counts = checkOpCounts(t, counts, map[string]int64{})
+	checkChecksum(t, tw, 0)
+
+	// Add a tablet to the topology.
+	tablet1 := &topodatapb.Tablet{
+		Alias: &topodatapb.TabletAlias{
+			Cell: "aa",
+			Uid:  0,
+		},
+		Hostname: "host1",
+		PortMap: map[string]int32{
+			"vt": 123,
+		},
+		Keyspace: "keyspace",
+		Shard:    "shard",
+	}
+	require.NoError(t, ts.CreateTablet(ctx, tablet1), "CreateTablet failed for %v", tablet1.Alias)
+
+	tw.loadTablets()
+	counts = checkOpCounts(t, counts, map[string]int64{"ListTablets": 1, "GetTablet": 1, "AddTablet": 1})
+	checkChecksum(t, tw, 3238442862)
+
+	// Check the tablet is returned by GetAllTablets().
+	allTablets := fhc.GetAllTablets()
+	key1 := TabletToMapKey(tablet1)
+	assert.Len(t, allTablets, 1)
+	assert.Contains(t, allTablets, key1)
+	assert.True(t, proto.Equal(tablet1, allTablets[key1]))
+
+	// Cause the Get for the tablet to fail with a NoNode error. This simulates a race condition where
+	// the tablet is removed from the topo after the ListTablets call but before the GetTablet call.
+	factory.AddOperationError(
+		memorytopo.Get,
+		"tablets/aa-0000000000/Tablet",
+		topo.NewError(topo.NoNode, "tablets/aa-0000000000/Tablet"),
+	)
+
+	// Ensure that a topo GetTablet error fails. If not, the rest of this test is invalid.
+	_, err := ts.GetTablet(ctx, tablet1.Alias)
+	require.Error(t, err)
+	require.True(t, topo.IsErrType(err, topo.NoNode))
+
+	// Now force the error during loadTablets.
+	tw.loadTablets()
+	checkOpCounts(t, counts, map[string]int64{"ListTablets": 1, "GetTablet": 1, "RemoveTablet": 1})
+	checkChecksum(t, tw, 0)
+
+	// Ensure the tablet is no longer returned by GetAllTablets()
+	allTablets = fhc.GetAllTablets()
+	assert.Len(t, allTablets, 0)
+}

go/vt/topo/memorytopo/directory.go

@@ -37,6 +37,9 @@ func (c *Conn) ListDir(ctx context.Context, dirPath string, full bool) ([]topo.D
 	if c.factory.err != nil {
 		return nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(ListDir, dirPath); err != nil {
+		return nil, err
+	}
 
 	isRoot := false
 	if dirPath == "" || dirPath == "/" {

go/vt/topo/memorytopo/election.go

@@ -33,6 +33,10 @@ func (c *Conn) NewLeaderParticipation(name, id string) (topo.LeaderParticipation
 	c.factory.mu.Lock()
 	defer c.factory.mu.Unlock()
 
+	if err := c.factory.getOperationError(NewLeaderParticipation, id); err != nil {
+		return nil, err
+	}
+
 	// Make sure the global path exists.
 	electionPath := path.Join(electionsPath, name)
 	if n := c.factory.getOrCreatePath(c.cell, electionPath); n == nil {

go/vt/topo/memorytopo/file.go

@@ -44,6 +44,9 @@ func (c *Conn) Create(ctx context.Context, filePath string, contents []byte) (to
 	if c.factory.err != nil {
 		return nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(Create, filePath); err != nil {
+		return nil, err
+	}
 
 	// Get the parent dir.
 	dir, file := path.Split(filePath)
@@ -88,6 +91,9 @@ func (c *Conn) Update(ctx context.Context, filePath string, contents []byte, ver
 	if c.factory.err != nil {
 		return nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(Update, filePath); err != nil {
+		return nil, err
+	}
 
 	// Get the parent dir, we'll need it in case of creation.
 	dir, file := path.Split(filePath)
@@ -162,6 +168,9 @@ func (c *Conn) Get(ctx context.Context, filePath string) ([]byte, topo.Version,
 	if c.factory.err != nil {
 		return nil, nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(Get, filePath); err != nil {
+		return nil, nil, err
+	}
 
 	// Get the node.
 	n := c.factory.nodeByPath(c.cell, filePath)
@@ -187,6 +196,9 @@ func (c *Conn) List(ctx context.Context, filePathPrefix string) ([]topo.KVInfo,
 	if c.factory.err != nil {
 		return nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(List, filePathPrefix); err != nil {
+		return nil, err
+	}
 
 	dir, file := path.Split(filePathPrefix)
 	// Get the node to list.
@@ -246,6 +258,9 @@ func (c *Conn) Delete(ctx context.Context, filePath string, version topo.Version
 	if c.factory.err != nil {
 		return c.factory.err
 	}
+	if err := c.factory.getOperationError(Delete, filePath); err != nil {
+		return err
+	}
 
 	// Get the parent dir.
 	dir, file := path.Split(filePath)

go/vt/topo/memorytopo/lock.go

@@ -42,11 +42,25 @@ type memoryTopoLockDescriptor struct {
 
 // TryLock is part of the topo.Conn interface. Its implementation is same as Lock
 func (c *Conn) TryLock(ctx context.Context, dirPath, contents string) (topo.LockDescriptor, error) {
+	c.factory.mu.Lock()
+	err := c.factory.getOperationError(TryLock, dirPath)
+	c.factory.mu.Unlock()
+	if err != nil {
+		return nil, err
+	}
+
 	return c.Lock(ctx, dirPath, contents)
 }
 
 // Lock is part of the topo.Conn interface.
 func (c *Conn) Lock(ctx context.Context, dirPath, contents string) (topo.LockDescriptor, error) {
+	c.factory.mu.Lock()
+	err := c.factory.getOperationError(Lock, dirPath)
+	c.factory.mu.Unlock()
+	if err != nil {
+		return nil, err
+	}
+
 	return c.lock(ctx, dirPath, contents)
 }
 

go/vt/topo/memorytopo/memorytopo.go

@@ -23,6 +23,7 @@ import (
 	"context"
 	"errors"
 	"math/rand"
+	"regexp"
 	"strings"
 	"sync"
 	"sync/atomic"
@@ -49,6 +50,25 @@ const (
 	UnreachableServerAddr = "unreachable"
 )
 
+// Operation is one of the operations defined by topo.Conn
+type Operation int
+
+// The following is the list of topo.Conn operations
+const (
+	ListDir = Operation(iota)
+	Create
+	Update
+	Get
+	List
+	Delete
+	Lock
+	TryLock
+	Watch
+	WatchRecursive
+	NewLeaderParticipation
+	Close
+)
+
 // Factory is a memory-based implementation of topo.Factory.  It
 // takes a file-system like approach, with directories at each level
 // being an actual directory node. This is meant to be closer to
@@ -71,6 +91,15 @@ type Factory struct {
 	// err is used for testing purposes to force queries / watches
 	// to return the given error
 	err error
+	// operationErrors is used for testing purposes to fake errors from
+	// operations and paths matching the spec
+	operationErrors map[Operation][]errorSpec
+}
+
+type errorSpec struct {
+	op          Operation
+	pathPattern *regexp.Regexp
+	err         error
 }
 
 // HasGlobalReadOnlyCell is part of the topo.Factory interface.
@@ -236,8 +265,9 @@ func (n *node) PropagateWatchError(err error) {
 // in case of a problem.
 func NewServerAndFactory(ctx context.Context, cells ...string) (*topo.Server, *Factory) {
 	f := &Factory{
-		cells:      make(map[string]*node),
-		generation: uint64(rand.Int63n(1 << 60)),
+		cells:           make(map[string]*node),
+		generation:      uint64(rand.Int63n(1 << 60)),
+		operationErrors: make(map[Operation][]errorSpec),
 	}
 	f.cells[topo.GlobalCell] = f.newDirectory(topo.GlobalCell, nil)
 
@@ -349,3 +379,24 @@ func (f *Factory) recursiveDelete(n *node) {
 		f.recursiveDelete(parent)
 	}
 }
+
+func (f *Factory) AddOperationError(op Operation, pathPattern string, err error) {
+	f.mu.Lock()
+	defer f.mu.Unlock()
+
+	f.operationErrors[op] = append(f.operationErrors[op], errorSpec{
+		op:          op,
+		pathPattern: regexp.MustCompile(pathPattern),
+		err:         err,
+	})
+}
+
+func (f *Factory) getOperationError(op Operation, path string) error {
+	specs := f.operationErrors[op]
+	for _, spec := range specs {
+		if spec.pathPattern.MatchString(path) {
+			return spec.err
+		}
+	}
+	return nil
+}

go/vt/topo/memorytopo/watch.go

@@ -35,6 +35,9 @@ func (c *Conn) Watch(ctx context.Context, filePath string) (*topo.WatchData, <-c
 	if c.factory.err != nil {
 		return nil, nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(Watch, filePath); err != nil {
+		return nil, nil, err
+	}
 
 	n := c.factory.nodeByPath(c.cell, filePath)
 	if n == nil {
@@ -85,6 +88,9 @@ func (c *Conn) WatchRecursive(ctx context.Context, dirpath string) ([]*topo.Watc
 	if c.factory.err != nil {
 		return nil, nil, c.factory.err
 	}
+	if err := c.factory.getOperationError(WatchRecursive, dirpath); err != nil {
+		return nil, nil, err
+	}
 
 	n := c.factory.getOrCreatePath(c.cell, dirpath)
 	if n == nil {