var Driver = cloud.DriverFunc(newAzureInstanceSet)
type azureInstanceSetConfig struct {
- SubscriptionID string
- ClientID string
- ClientSecret string
- TenantID string
- CloudEnvironment string
- ResourceGroup string
- ImageResourceGroup string
- Location string
- Network string
- NetworkResourceGroup string
- Subnet string
- StorageAccount string
- BlobContainer string
- DeleteDanglingResourcesAfter arvados.Duration
- AdminUsername string
+ SubscriptionID string
+ ClientID string
+ ClientSecret string
+ TenantID string
+ CloudEnvironment string
+ ResourceGroup string
+ ImageResourceGroup string
+ Location string
+ Network string
+ NetworkResourceGroup string
+ Subnet string
+ StorageAccount string
+ BlobContainer string
+ SharedImageGalleryName string
+ SharedImageGalleryImageVersion string
+ DeleteDanglingResourcesAfter arvados.Duration
+ AdminUsername string
}
type containerWrapper interface {
return nil
}
+func (az *azureInstanceSet) cleanupNic(nic network.Interface) {
+ _, delerr := az.netClient.delete(context.Background(), az.azconfig.ResourceGroup, *nic.Name)
+ if delerr != nil {
+ az.logger.WithError(delerr).Warnf("Error cleaning up NIC after failed create")
+ }
+}
+
func (az *azureInstanceSet) Create(
instanceType arvados.InstanceType,
imageID cloud.ImageID,
Tags: tags,
InterfacePropertiesFormat: &network.InterfacePropertiesFormat{
IPConfigurations: &[]network.InterfaceIPConfiguration{
- network.InterfaceIPConfiguration{
+ {
Name: to.StringPtr("ip1"),
InterfaceIPConfigurationPropertiesFormat: &network.InterfaceIPConfigurationPropertiesFormat{
Subnet: &network.Subnet{
var storageProfile *compute.StorageProfile
re := regexp.MustCompile(`^http(s?)://`)
- if re.MatchString(string(imageID)) && az.blobcont != nil {
+ if re.MatchString(string(imageID)) {
+ if az.blobcont == nil {
+ az.cleanupNic(nic)
+ return nil, wrapAzureError(errors.New("Invalid configuration: can't configure unmanaged image URL without StorageAccount and BlobContainer"))
+ }
blobname = fmt.Sprintf("%s-os.vhd", name)
instanceVhd := fmt.Sprintf("https://%s.blob.%s/%s/%s",
az.azconfig.StorageAccount,
},
},
}
- } else if az.blobcont == nil {
+ } else {
+ id := to.StringPtr("/subscriptions/" + az.azconfig.SubscriptionID + "/resourceGroups/" + az.imageResourceGroup + "/providers/Microsoft.Compute/images/" + string(imageID))
+ if az.azconfig.SharedImageGalleryName != "" && az.azconfig.SharedImageGalleryImageVersion != "" {
+ id = to.StringPtr("/subscriptions/" + az.azconfig.SubscriptionID + "/resourceGroups/" + az.imageResourceGroup + "/providers/Microsoft.Compute/galleries/" + az.azconfig.SharedImageGalleryName + "/images/" + string(imageID) + "/versions/" + az.azconfig.SharedImageGalleryImageVersion)
+ } else if az.azconfig.SharedImageGalleryName != "" || az.azconfig.SharedImageGalleryImageVersion != "" {
+ az.cleanupNic(nic)
+ return nil, wrapAzureError(errors.New("Invalid configuration: SharedImageGalleryName and SharedImageGalleryImageVersion must both be set or both be empty"))
+ }
storageProfile = &compute.StorageProfile{
ImageReference: &compute.ImageReference{
- ID: to.StringPtr("/subscriptions/" + az.azconfig.SubscriptionID + "/resourceGroups/" + az.imageResourceGroup + "/providers/Microsoft.Compute/images/" + string(imageID)),
+ ID: id,
},
OsDisk: &compute.OSDisk{
OsType: compute.Linux,
CreateOption: compute.DiskCreateOptionTypesFromImage,
},
}
- } else {
- return nil, wrapAzureError(errors.New("Invalid configuration: can't configure unmanaged image URL without StorageAccount and BlobContainer"))
}
vmParameters := compute.VirtualMachine{
StorageProfile: storageProfile,
NetworkProfile: &compute.NetworkProfile{
NetworkInterfaces: &[]compute.NetworkInterfaceReference{
- compute.NetworkInterfaceReference{
+ {
ID: nic.ID,
NetworkInterfaceReferenceProperties: &compute.NetworkInterfaceReferenceProperties{
Primary: to.BoolPtr(true),
},
}
+ var maxPrice float64
+ if instanceType.Preemptible {
+ // Setting maxPrice to -1 is the equivalent of paying spot price, up to the
+ // normal price. This means the node will not be pre-empted for price
+ // reasons. It may still be pre-empted for capacity reasons though. And
+ // Azure offers *no* SLA on spot instances.
+ maxPrice = -1
+ vmParameters.VirtualMachineProperties.Priority = compute.Spot
+ vmParameters.VirtualMachineProperties.EvictionPolicy = compute.Delete
+ vmParameters.VirtualMachineProperties.BillingProfile = &compute.BillingProfile{MaxPrice: &maxPrice}
+ }
+
vm, err := az.vmClient.createOrUpdate(az.ctx, az.azconfig.ResourceGroup, name, vmParameters)
if err != nil {
- if az.blobcont != nil {
+ // Do some cleanup. Otherwise, an unbounded number of new unused nics and
+ // blobs can pile up during times when VMs can't be created and the
+ // dispatcher keeps retrying, because the garbage collection in manageBlobs
+ // and manageNics is only triggered periodically. This is most important
+ // for nics, because those are subject to a quota.
+ az.cleanupNic(nic)
+
+ if blobname != "" {
_, delerr := az.blobcont.GetBlobReference(blobname).DeleteIfExists(nil)
if delerr != nil {
az.logger.WithError(delerr).Warnf("Error cleaning up vhd blob after failed create")
}
}
- // Leave cleaning up of managed disks to the garbage collection in manageDisks()
- _, delerr := az.netClient.delete(context.Background(), az.azconfig.ResourceGroup, *nic.Name)
- if delerr != nil {
- az.logger.WithError(delerr).Warnf("Error cleaning up NIC after failed create")
- }
+ // Leave cleaning up of managed disks to the garbage collection in manageDisks()
return nil, wrapAzureError(err)
}
}
for ; response.NotDone(); err = response.Next() {
+ if err != nil {
+ az.logger.WithError(err).Warn("Error getting next page of disks")
+ return
+ }
for _, d := range response.Values() {
if d.DiskProperties.DiskState == compute.Unattached &&
d.Name != nil && re.MatchString(*d.Name) &&
projects / arvados.git / blobdiff