Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
32 changes: 32 additions & 0 deletions cmd/api/api/instances.go
Original file line number Diff line number Diff line change
Expand Up @@ -542,6 +542,16 @@ func (s *ApiService) StandbyInstance(ctx context.Context, request oapi.StandbyIn
}
standbyReq.Compression = compression
}
if request.Body != nil && request.Body.CompressionDelay != nil {
compressionDelay, err := parseOptionalDuration(*request.Body.CompressionDelay, "compression_delay")
if err != nil {
return oapi.StandbyInstance400JSONResponse{
Code: "invalid_compression_delay",
Message: err.Error(),
}, nil
}
standbyReq.CompressionDelay = compressionDelay
}

result, err := s.InstanceManager.StandbyInstance(ctx, inst.Id, standbyReq)
if err != nil {
Expand Down Expand Up @@ -1152,6 +1162,13 @@ func toInstanceSnapshotPolicy(policy oapi.SnapshotPolicy) (*instances.SnapshotPo
}
out.Compression = compression
}
if policy.StandbyCompressionDelay != nil {
delay, err := parseOptionalDuration(*policy.StandbyCompressionDelay, "standby_compression_delay")
if err != nil {
return nil, err
}
out.StandbyCompressionDelay = delay
}
return out, nil
}

Expand All @@ -1176,5 +1193,20 @@ func toOAPISnapshotPolicy(policy instances.SnapshotPolicy) oapi.SnapshotPolicy {
compression := toOAPISnapshotCompressionConfig(*policy.Compression)
out.Compression = &compression
}
if policy.StandbyCompressionDelay != nil {
delay := policy.StandbyCompressionDelay.String()
out.StandbyCompressionDelay = &delay
}
return out
}

func parseOptionalDuration(value string, field string) (*time.Duration, error) {
duration, err := time.ParseDuration(value)
if err != nil {
return nil, fmt.Errorf("%s must be a valid duration: %w", field, err)
}
if duration < 0 {
return nil, fmt.Errorf("%s cannot be negative", field)
}
return &duration, nil
}
152 changes: 151 additions & 1 deletion cmd/api/api/instances_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -480,10 +480,85 @@ func TestCreateInstance_MapsNetworkEgressEnforcementMode(t *testing.T) {
assert.Equal(t, instances.EgressEnforcementModeHTTPHTTPSOnly, mockMgr.lastReq.NetworkEgress.EnforcementMode)
}

func TestCreateInstance_MapsAutoStandbyPolicy(t *testing.T) {
func TestCreateInstance_MapsStandbyCompressionDelayInSnapshotPolicy(t *testing.T) {
t.Parallel()

svc := newTestService(t)
origMgr := svc.InstanceManager
mockMgr := &captureCreateManager{Manager: origMgr}
svc.InstanceManager = mockMgr

delay := "2m30s"
resp, err := svc.CreateInstance(ctx(), oapi.CreateInstanceRequestObject{
Body: &oapi.CreateInstanceRequest{
Name: "test-standby-compression-delay",
Image: "docker.io/library/alpine:latest",
SnapshotPolicy: &oapi.SnapshotPolicy{
StandbyCompressionDelay: &delay,
},
},
})
require.NoError(t, err)
_, ok := resp.(oapi.CreateInstance201JSONResponse)
require.True(t, ok, "expected 201 response")

require.NotNil(t, mockMgr.lastReq)
require.NotNil(t, mockMgr.lastReq.SnapshotPolicy)
require.NotNil(t, mockMgr.lastReq.SnapshotPolicy.StandbyCompressionDelay)
assert.Equal(t, 150*time.Second, *mockMgr.lastReq.SnapshotPolicy.StandbyCompressionDelay)
}

func TestCreateInstance_InvalidStandbyCompressionDelayInSnapshotPolicy(t *testing.T) {
t.Parallel()

svc := newTestService(t)
delay := "not-a-duration"

resp, err := svc.CreateInstance(ctx(), oapi.CreateInstanceRequestObject{
Body: &oapi.CreateInstanceRequest{
Name: "test-invalid-standby-delay",
Image: "docker.io/library/alpine:latest",
SnapshotPolicy: &oapi.SnapshotPolicy{
StandbyCompressionDelay: &delay,
},
},
})
require.NoError(t, err)

badReq, ok := resp.(oapi.CreateInstance400JSONResponse)
require.True(t, ok, "expected 400 response")
assert.Equal(t, "invalid_snapshot_policy", badReq.Code)
assert.Contains(t, badReq.Message, "standby_compression_delay")
}

func TestInstanceToOAPI_EmitsStandbyCompressionDelayInSnapshotPolicy(t *testing.T) {
t.Parallel()

delay := 90 * time.Second
inst := instances.Instance{
StoredMetadata: instances.StoredMetadata{
Id: "inst-standby-delay",
Name: "inst-standby-delay",
Image: "docker.io/library/alpine:latest",
CreatedAt: time.Now(),
HypervisorType: hypervisor.TypeCloudHypervisor,
SnapshotPolicy: &instances.SnapshotPolicy{
StandbyCompressionDelay: &delay,
},
},
State: instances.StateStandby,
}

oapiInst := instanceToOAPI(inst)
require.NotNil(t, oapiInst.SnapshotPolicy)
require.NotNil(t, oapiInst.SnapshotPolicy.StandbyCompressionDelay)
assert.Equal(t, "1m30s", *oapiInst.SnapshotPolicy.StandbyCompressionDelay)
}

func TestCreateInstance_MapsAutoStandbyPolicy(t *testing.T) {
t.Parallel()

svc := newTestService(t)
origMgr := svc.InstanceManager
mockMgr := &captureCreateManager{Manager: origMgr}
svc.InstanceManager = mockMgr
Expand Down Expand Up @@ -907,6 +982,81 @@ func TestStandbyInstance_InvalidRequest(t *testing.T) {
assert.Contains(t, badReq.Message, "invalid snapshot compression level")
}

func TestStandbyInstance_MapsCompressionDelay(t *testing.T) {
t.Parallel()

svc := newTestService(t)
now := time.Now()
source := instances.Instance{
StoredMetadata: instances.StoredMetadata{
Id: "standby-delay-src",
Name: "standby-delay-src",
Image: "docker.io/library/alpine:latest",
CreatedAt: now,
HypervisorType: hypervisor.TypeCloudHypervisor,
},
State: instances.StateRunning,
}

mockMgr := &captureStandbyManager{
Manager: svc.InstanceManager,
result: &source,
}
svc.InstanceManager = mockMgr

delay := "45s"
resp, err := svc.StandbyInstance(
mw.WithResolvedInstance(ctx(), source.Id, source),
oapi.StandbyInstanceRequestObject{
Id: source.Id,
Body: &oapi.StandbyInstanceRequest{
CompressionDelay: &delay,
},
},
)
require.NoError(t, err)
_, ok := resp.(oapi.StandbyInstance200JSONResponse)
require.True(t, ok, "expected 200 response")

require.NotNil(t, mockMgr.lastReq)
require.NotNil(t, mockMgr.lastReq.CompressionDelay)
assert.Equal(t, 45*time.Second, *mockMgr.lastReq.CompressionDelay)
}

func TestStandbyInstance_InvalidCompressionDelay(t *testing.T) {
t.Parallel()

svc := newTestService(t)
now := time.Now()
source := instances.Instance{
StoredMetadata: instances.StoredMetadata{
Id: "standby-invalid-delay-src",
Name: "standby-invalid-delay-src",
Image: "docker.io/library/alpine:latest",
CreatedAt: now,
HypervisorType: hypervisor.TypeCloudHypervisor,
},
State: instances.StateRunning,
}

delay := "-5s"
resp, err := svc.StandbyInstance(
mw.WithResolvedInstance(ctx(), source.Id, source),
oapi.StandbyInstanceRequestObject{
Id: source.Id,
Body: &oapi.StandbyInstanceRequest{
CompressionDelay: &delay,
},
},
)
require.NoError(t, err)

badReq, ok := resp.(oapi.StandbyInstance400JSONResponse)
require.True(t, ok, "expected 400 response")
assert.Equal(t, "invalid_compression_delay", badReq.Code)
assert.Contains(t, badReq.Message, "compression_delay")
}

func TestForkInstance_FromRunningFlagForwarded(t *testing.T) {
t.Parallel()
svc := newTestService(t)
Expand Down
37 changes: 29 additions & 8 deletions lib/instances/compression_integration_linux_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -264,25 +264,46 @@ func waitForRunningAndExecReady(t *testing.T, ctx context.Context, mgr *manager,

func writeGuestMarker(t *testing.T, ctx context.Context, inst *Instance, path string, value string) {
t.Helper()
execCtx, cancel := context.WithTimeout(ctx, integrationTestTimeout(compressionGuestExecTimeout))
defer cancel()

output, exitCode, err := execCommand(execCtx, inst, "sh", "-c", fmt.Sprintf("printf %q > %s && sync", value, path))
output, exitCode, err := execCommandWithRetry(ctx, inst, compressionGuestExecTimeout, "sh", "-c", fmt.Sprintf("printf %q > %s && sync", value, path))
require.NoError(t, err)
require.Equal(t, 0, exitCode, output)
}

func assertGuestMarker(t *testing.T, ctx context.Context, inst *Instance, path string, expected string) {
t.Helper()
execCtx, cancel := context.WithTimeout(ctx, integrationTestTimeout(compressionGuestExecTimeout))
defer cancel()

output, exitCode, err := execCommand(execCtx, inst, "cat", path)
output, exitCode, err := execCommandWithRetry(ctx, inst, compressionGuestExecTimeout, "cat", path)
require.NoError(t, err)
require.Equal(t, 0, exitCode, output)
assert.Equal(t, expected, output)
}

func execCommandWithRetry(ctx context.Context, inst *Instance, timeout time.Duration, command ...string) (string, int, error) {
Comment thread
sjmiller609 marked this conversation as resolved.
deadline := time.Now().Add(integrationTestTimeout(timeout))
var lastOutput string
var lastExitCode int
var lastErr error

for {
execCtx, cancel := context.WithTimeout(ctx, integrationTestTimeout(5*time.Second))
output, exitCode, err := execCommand(execCtx, inst, command...)
cancel()

if err == nil {
return output, exitCode, nil
}

lastOutput = output
lastExitCode = exitCode
lastErr = err

if time.Now().After(deadline) {
return lastOutput, lastExitCode, lastErr
}

time.Sleep(500 * time.Millisecond)
}
}

func waitForCompressionJobStart(t *testing.T, mgr *manager, key string, timeout time.Duration) {
t.Helper()
deadline := time.Now().Add(timeout)
Expand Down
5 changes: 5 additions & 0 deletions lib/instances/create.go
Original file line number Diff line number Diff line change
Expand Up @@ -599,6 +599,11 @@ func validateCreateRequest(req *CreateInstanceRequest) error {
return err
}
}
if req.SnapshotPolicy != nil && req.SnapshotPolicy.StandbyCompressionDelay != nil {
if _, err := normalizeStandbyCompressionDelay(req.SnapshotPolicy.StandbyCompressionDelay); err != nil {
return err
}
}
normalizedAutoStandby, err := normalizeAutoStandbyPolicy(req.AutoStandby)
if err != nil {
return err
Expand Down
4 changes: 2 additions & 2 deletions lib/instances/delete.go
Original file line number Diff line number Diff line change
Expand Up @@ -39,8 +39,8 @@ func (m *manager) deleteInstance(
if err != nil {
return fmt.Errorf("wait for instance compression to stop: %w", err)
}
if target != nil {
m.recordSnapshotCompressionPreemption(ctx, snapshotCompressionPreemptionDeleteInstance, *target)
if target != nil && target.State == compressionJobStateRunning {
m.recordSnapshotCompressionPreemption(ctx, snapshotCompressionPreemptionDeleteInstance, target.Target)
}

// 2. Get network allocation BEFORE killing VMM (while we can still query it)
Expand Down
9 changes: 8 additions & 1 deletion lib/instances/fork.go
Original file line number Diff line number Diff line change
Expand Up @@ -269,7 +269,7 @@ func (m *manager) forkInstanceFromStoppedOrStandby(ctx context.Context, id strin
}

now := time.Now()
forkMeta := cloneStoredMetadata(meta.StoredMetadata)
forkMeta := cloneStoredMetadataWithoutPendingStandbyCompression(meta.StoredMetadata)
forkMeta.Id = forkID
forkMeta.Name = req.Name
forkMeta.CreatedAt = now
Expand Down Expand Up @@ -472,6 +472,7 @@ func (m *manager) cleanupForkInstanceOnError(ctx context.Context, forkID string)

func cloneStoredMetadata(src StoredMetadata) StoredMetadata {
dst := src
dst.PendingStandbyCompression = clonePendingStandbyCompression(src.PendingStandbyCompression)

if src.Env != nil {
dst.Env = make(map[string]string, len(src.Env))
Expand Down Expand Up @@ -536,3 +537,9 @@ func cloneStoredMetadata(src StoredMetadata) StoredMetadata {

return dst
}

func cloneStoredMetadataWithoutPendingStandbyCompression(src StoredMetadata) StoredMetadata {
dst := cloneStoredMetadata(src)
dst.PendingStandbyCompression = nil
return dst
}
36 changes: 36 additions & 0 deletions lib/instances/fork_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -282,9 +282,11 @@ func TestCloneStoredMetadataForFork_DeepCopiesReferenceFields(t *testing.T) {
t.Parallel()
startedAt := time.Now().Add(-2 * time.Minute)
stoppedAt := time.Now().Add(-1 * time.Minute)
notBefore := time.Now().Add(5 * time.Minute)
pid := 1234
exitCode := 17
compressionLevel := 5
pendingLevel := 3

src := StoredMetadata{
Env: map[string]string{"A": "1"},
Expand All @@ -310,6 +312,14 @@ func TestCloneStoredMetadataForFork_DeepCopiesReferenceFields(t *testing.T) {
Level: &compressionLevel,
},
},
PendingStandbyCompression: &PendingStandbyCompression{
Policy: snapshotstore.SnapshotCompressionConfig{
Enabled: true,
Algorithm: snapshotstore.SnapshotCompressionAlgorithmZstd,
Level: &pendingLevel,
},
NotBefore: notBefore,
},
}

cloned := cloneStoredMetadata(src)
Expand All @@ -327,6 +337,8 @@ func TestCloneStoredMetadataForFork_DeepCopiesReferenceFields(t *testing.T) {
cloned.AutoStandby.IgnoreDestinationPorts[0] = 443
*cloned.SnapshotPolicy.Compression.Level = 9
now := time.Now()
*cloned.PendingStandbyCompression.Policy.Level = 1
cloned.PendingStandbyCompression.NotBefore = now
*cloned.StartedAt = now
*cloned.StoppedAt = now

Expand All @@ -341,10 +353,34 @@ func TestCloneStoredMetadataForFork_DeepCopiesReferenceFields(t *testing.T) {
require.Equal(t, "10.0.0.0/8", src.AutoStandby.IgnoreSourceCIDRs[0])
require.Equal(t, uint16(22), src.AutoStandby.IgnoreDestinationPorts[0])
require.Equal(t, 5, *src.SnapshotPolicy.Compression.Level)
require.NotNil(t, src.PendingStandbyCompression)
require.NotNil(t, src.PendingStandbyCompression.Policy.Level)
require.Equal(t, 3, *src.PendingStandbyCompression.Policy.Level)
require.Equal(t, notBefore, src.PendingStandbyCompression.NotBefore)
require.Equal(t, startedAt, *src.StartedAt)
require.Equal(t, stoppedAt, *src.StoppedAt)
}

func TestCloneStoredMetadataWithoutPendingStandbyCompression_ClearsPendingPlan(t *testing.T) {
t.Parallel()

level := 4
src := StoredMetadata{
PendingStandbyCompression: &PendingStandbyCompression{
Policy: snapshotstore.SnapshotCompressionConfig{
Enabled: true,
Algorithm: snapshotstore.SnapshotCompressionAlgorithmZstd,
Level: &level,
},
NotBefore: time.Now().Add(2 * time.Minute),
},
}

cloned := cloneStoredMetadataWithoutPendingStandbyCompression(src)
assert.Nil(t, cloned.PendingStandbyCompression)
require.NotNil(t, src.PendingStandbyCompression)
}

func TestRotateSourceVsockForRestore_CloudHypervisorDoesNotPersistCIDRewrite(t *testing.T) {
t.Parallel()
manager, _ := setupTestManager(t)
Expand Down
Loading
Loading