Browse code

Merge pull request #31055 from Microsoft/jjh/fix30278

Windows: Retry workaround for RS1/RS2 compute system enumeration
(cherry picked from commit 7761c69e239171c6b388d53d03619eacc49338a0)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>

Victor Vieux authored on 2017/02/18 09:53:16
Showing 4 changed files
... ...
@@ -31,6 +31,7 @@ import (
31 31
 	"github.com/docker/docker/pkg/ioutils"
32 32
 	"github.com/docker/docker/pkg/longpath"
33 33
 	"github.com/docker/docker/pkg/reexec"
34
+	"github.com/docker/docker/pkg/system"
34 35
 	units "github.com/docker/go-units"
35 36
 	"golang.org/x/sys/windows"
36 37
 )
... ...
@@ -265,19 +266,35 @@ func (d *Driver) Remove(id string) error {
265 265
 	// it is a transient error. Retry until it succeeds.
266 266
 	var computeSystems []hcsshim.ContainerProperties
267 267
 	retryCount := 0
268
+	osv := system.GetOSVersion()
268 269
 	for {
269
-		// Get and terminate any template VMs that are currently using the layer
270
+		// Get and terminate any template VMs that are currently using the layer.
271
+		// Note: It is unfortunate that we end up in the graphdrivers Remove() call
272
+		// for both containers and images, but the logic for template VMs is only
273
+		// needed for images - specifically we are looking to see if a base layer
274
+		// is in use by a template VM as a result of having started a Hyper-V
275
+		// container at some point.
276
+		//
277
+		// We have a retry loop for ErrVmcomputeOperationInvalidState and
278
+		// ErrVmcomputeOperationAccessIsDenied as there is a race condition
279
+		// in RS1 and RS2 building during enumeration when a silo is going away
280
+		// for example under it, in HCS. AccessIsDenied added to fix 30278.
281
+		//
282
+		// TODO @jhowardmsft - For RS3, we can remove the retries. Also consider
283
+		// using platform APIs (if available) to get this more succinctly. Also
284
+		// consider enlighting the Remove() interface to have context of why
285
+		// the remove is being called - that could improve efficiency by not
286
+		// enumerating compute systems during a remove of a container as it's
287
+		// not required.
270 288
 		computeSystems, err = hcsshim.GetContainers(hcsshim.ComputeSystemQuery{})
271 289
 		if err != nil {
272
-			if err == hcsshim.ErrVmcomputeOperationInvalidState {
273
-				if retryCount >= 5 {
274
-					// If we are unable to get the list of containers
275
-					// go ahead and attempt to delete the layer anyway
276
-					// as it will most likely work.
290
+			if (osv.Build < 15139) &&
291
+				((err == hcsshim.ErrVmcomputeOperationInvalidState) || (err == hcsshim.ErrVmcomputeOperationAccessIsDenied)) {
292
+				if retryCount >= 500 {
277 293
 					break
278 294
 				}
279 295
 				retryCount++
280
-				time.Sleep(2 * time.Second)
296
+				time.Sleep(10 * time.Millisecond)
281 297
 				continue
282 298
 			}
283 299
 			return err
... ...
@@ -1,6 +1,6 @@
1 1
 # the following lines are in sorted order, FYI
2 2
 github.com/Azure/go-ansiterm 388960b655244e76e24c75f48631564eaefade62
3
-github.com/Microsoft/hcsshim v0.5.9
3
+github.com/Microsoft/hcsshim v0.5.12
4 4
 github.com/Microsoft/go-winio v0.3.8
5 5
 github.com/Sirupsen/logrus v0.11.0
6 6
 github.com/davecgh/go-spew 6d212800a42e8ab5c146b8ace3490ee17e5225f9
... ...
@@ -50,6 +50,10 @@ var (
50 50
 
51 51
 	// ErrProcNotFound is an error encountered when the the process cannot be found
52 52
 	ErrProcNotFound = syscall.Errno(0x7f)
53
+
54
+	// ErrVmcomputeOperationAccessIsDenied is an error which can be encountered when enumerating compute systems in RS1/RS2
55
+	// builds when the underlying silo might be in the process of terminating. HCS was fixed in RS3.
56
+	ErrVmcomputeOperationAccessIsDenied = syscall.Errno(0x5)
53 57
 )
54 58
 
55 59
 // ProcessError is an error encountered in HCS during an operation on a Process object
... ...
@@ -41,30 +41,32 @@ type HvRuntime struct {
41 41
 // ContainerConfig is used as both the input of CreateContainer
42 42
 // and to convert the parameters to JSON for passing onto the HCS
43 43
 type ContainerConfig struct {
44
-	SystemType               string      // HCS requires this to be hard-coded to "Container"
45
-	Name                     string      // Name of the container. We use the docker ID.
46
-	Owner                    string      // The management platform that created this container
47
-	IsDummy                  bool        // Used for development purposes.
48
-	VolumePath               string      `json:",omitempty"` // Windows volume path for scratch space. Used by Windows Server Containers only. Format \\?\\Volume{GUID}
49
-	IgnoreFlushesDuringBoot  bool        // Optimization hint for container startup in Windows
50
-	LayerFolderPath          string      `json:",omitempty"` // Where the layer folders are located. Used by Windows Server Containers only. Format  %root%\windowsfilter\containerID
51
-	Layers                   []Layer     // List of storage layers. Required for Windows Server and Hyper-V Containers. Format ID=GUID;Path=%root%\windowsfilter\layerID
52
-	Credentials              string      `json:",omitempty"` // Credentials information
53
-	ProcessorCount           uint32      `json:",omitempty"` // Number of processors to assign to the container.
54
-	ProcessorWeight          uint64      `json:",omitempty"` // CPU Shares 0..10000 on Windows; where 0 will be omitted and HCS will default.
55
-	ProcessorMaximum         int64       `json:",omitempty"` // CPU maximum usage percent 1..100
56
-	StorageIOPSMaximum       uint64      `json:",omitempty"` // Maximum Storage IOPS
57
-	StorageBandwidthMaximum  uint64      `json:",omitempty"` // Maximum Storage Bandwidth in bytes per second
58
-	StorageSandboxSize       uint64      `json:",omitempty"` // Size in bytes that the container system drive should be expanded to if smaller
59
-	MemoryMaximumInMB        int64       `json:",omitempty"` // Maximum memory available to the container in Megabytes
60
-	HostName                 string      // Hostname
61
-	MappedDirectories        []MappedDir // List of mapped directories (volumes/mounts)
62
-	SandboxPath              string      `json:",omitempty"` // Location of unmounted sandbox. Used by Hyper-V containers only. Format %root%\windowsfilter
63
-	HvPartition              bool        // True if it a Hyper-V Container
64
-	EndpointList             []string    // List of networking endpoints to be attached to container
65
-	HvRuntime                *HvRuntime  `json:",omitempty"` // Hyper-V container settings. Used by Hyper-V containers only. Format ImagePath=%root%\BaseLayerID\UtilityVM
66
-	Servicing                bool        // True if this container is for servicing
67
-	AllowUnqualifiedDNSQuery bool        // True to allow unqualified DNS name resolution
44
+	SystemType                 string      // HCS requires this to be hard-coded to "Container"
45
+	Name                       string      // Name of the container. We use the docker ID.
46
+	Owner                      string      // The management platform that created this container
47
+	IsDummy                    bool        // Used for development purposes.
48
+	VolumePath                 string      `json:",omitempty"` // Windows volume path for scratch space. Used by Windows Server Containers only. Format \\?\\Volume{GUID}
49
+	IgnoreFlushesDuringBoot    bool        // Optimization hint for container startup in Windows
50
+	LayerFolderPath            string      `json:",omitempty"` // Where the layer folders are located. Used by Windows Server Containers only. Format  %root%\windowsfilter\containerID
51
+	Layers                     []Layer     // List of storage layers. Required for Windows Server and Hyper-V Containers. Format ID=GUID;Path=%root%\windowsfilter\layerID
52
+	Credentials                string      `json:",omitempty"` // Credentials information
53
+	ProcessorCount             uint32      `json:",omitempty"` // Number of processors to assign to the container.
54
+	ProcessorWeight            uint64      `json:",omitempty"` // CPU Shares 0..10000 on Windows; where 0 will be omitted and HCS will default.
55
+	ProcessorMaximum           int64       `json:",omitempty"` // CPU maximum usage percent 1..100
56
+	StorageIOPSMaximum         uint64      `json:",omitempty"` // Maximum Storage IOPS
57
+	StorageBandwidthMaximum    uint64      `json:",omitempty"` // Maximum Storage Bandwidth in bytes per second
58
+	StorageSandboxSize         uint64      `json:",omitempty"` // Size in bytes that the container system drive should be expanded to if smaller
59
+	MemoryMaximumInMB          int64       `json:",omitempty"` // Maximum memory available to the container in Megabytes
60
+	HostName                   string      // Hostname
61
+	MappedDirectories          []MappedDir // List of mapped directories (volumes/mounts)
62
+	SandboxPath                string      `json:",omitempty"` // Location of unmounted sandbox. Used by Hyper-V containers only. Format %root%\windowsfilter
63
+	HvPartition                bool        // True if it a Hyper-V Container
64
+	EndpointList               []string    // List of networking endpoints to be attached to container
65
+	NetworkSharedContainerName string      `json:",omitempty"` // Name (ID) of the container that we will share the network stack with.
66
+	HvRuntime                  *HvRuntime  `json:",omitempty"` // Hyper-V container settings. Used by Hyper-V containers only. Format ImagePath=%root%\BaseLayerID\UtilityVM
67
+	Servicing                  bool        // True if this container is for servicing
68
+	AllowUnqualifiedDNSQuery   bool        // True to allow unqualified DNS name resolution
69
+	DNSSearchList              string      `json:",omitempty"` // Comma seperated list of DNS suffixes to use for name resolution
68 70
 }
69 71
 
70 72
 type ComputeSystemQuery struct {