Documentation ¶
Index ¶
- Constants
- Variables
- func AddStateInstance(stor storage.Storage, id instance.Id) error
- func Bootstrap(ctx environs.BootstrapContext, env environs.Environ, ...) (*environs.BootstrapResult, error)
- func BootstrapInstance(bootstrapContext environs.BootstrapContext, env environs.Environ, ...) (_ *environs.StartInstanceResult, resultBase *corebase.Base, ...)
- func ConfigureMachine(ctx environs.BootstrapContext, client ssh.Client, host string, ...) error
- func CreateStateFile(stor storage.Storage) (string, error)
- func CredentialNotValidError(err error) error
- func DefaultHostSSHOptions(string) (*ssh.Options, func(), error)
- func DeleteStateFile(stor storage.Storage) error
- func Destroy(env environs.Environ, ctx envcontext.ProviderCallContext) error
- func EnvFullName(modelUUID string) string
- func GetCheckNonceCommand(instanceConfig *instancecfg.InstanceConfig) string
- func HandleCredentialError(isAuthError func(error) bool, err error, ctx envcontext.ProviderCallContext)
- func MaybeHandleCredentialError(isAuthError func(error) bool, err error, ctx envcontext.ProviderCallContext) bool
- func MiBToGiB(m uint64) uint64
- func MinRootDiskSizeGiB(_ ostype.OSType) uint64
- func ProviderStateInstances(stor storage.StorageReader) ([]instance.Id, error)
- func RemoveStateInstances(stor storage.Storage, ids ...instance.Id) error
- func SaveState(storage storage.StorageWriter, state *BootstrapState) error
- func WaitSSH(ctx context.Context, stdErr io.Writer, client ssh.Client, ...) (addr string, err error)
- type AvailabilityZoneInstances
- type BootstrapState
- type DefaultProvider
- type HostSSHOptionsFunc
- type InstanceConfigurator
- type InstanceRefresher
- type RefreshableInstance
- type ZonedEnviron
Constants ¶
const ( // ErrorCredentialNotValid represents an error when a provider credential is // not valid. Realistically, this is not a transient error. Without a valid // credential we cannot do much on the provider. This is fatal. ErrorCredentialNotValid = errors.ConstError("credential not valid") )
const StateFile = "provider-state"
StateFile is the name of the file where the provider's state is stored.
Variables ¶
var AuthorisationFailureStatusCodes = set.NewInts( http.StatusUnauthorized, http.StatusPaymentRequired, http.StatusForbidden, http.StatusProxyAuthRequired, )
AuthorisationFailureStatusCodes contains http status code that signify authorisation difficulties.
var FinishBootstrap = func( ctx environs.BootstrapContext, client ssh.Client, env environs.Environ, callCtx envcontext.ProviderCallContext, inst instances.Instance, instanceConfig *instancecfg.InstanceConfig, opts environs.BootstrapDialOpts, ) error { interrupted := make(chan os.Signal, 1) ctx.InterruptNotify(interrupted) defer ctx.StopInterruptNotify(interrupted) hostSSHOptions := bootstrapSSHOptionsFunc(instanceConfig) addr, err := WaitSSH( ctx, ctx.GetStderr(), client, GetCheckNonceCommand(instanceConfig), &RefreshableInstance{inst, env}, callCtx, opts, hostSSHOptions, ) if err != nil { return err } ctx.Infof("Connected to %v", addr) sshOptions, cleanup, err := hostSSHOptions(addr) if err != nil { return err } defer cleanup() return ConfigureMachine(ctx, client, addr, instanceConfig, sshOptions) }
FinishBootstrap completes the bootstrap process by connecting to the instance via SSH and carrying out the cloud-config.
Note: FinishBootstrap is exposed so it can be replaced for testing.
var LongAttempt = utils.AttemptStrategy{ Total: 3 * time.Minute, Delay: 1 * time.Second, }
A request may fail to due "eventual consistency" semantics, which should resolve fairly quickly. These delays are specific to the provider and best tuned there. Other requests fail due to a slow state transition (e.g. an instance taking a while to release a security group after termination). If you need to poll for the latter kind, use LongAttempt.
TODO(katco): 2016-08-09: lp:1611427
var ShortAttempt = utils.AttemptStrategy{ Total: 5 * time.Second, Delay: 200 * time.Millisecond, }
Use ShortAttempt to poll for short-term events. TODO: This may need tuning for different providers (or even environments).
TODO(katco): 2016-08-09: lp:1611427
Functions ¶
func AddStateInstance ¶
AddStateInstance adds a controller instance ID to the provider-state file in storage.
func Bootstrap ¶
func Bootstrap( ctx environs.BootstrapContext, env environs.Environ, callCtx envcontext.ProviderCallContext, args environs.BootstrapParams, ) (*environs.BootstrapResult, error)
Bootstrap is a common implementation of the Bootstrap method defined on environs.Environ; we strongly recommend that this implementation be used when writing a new provider.
func BootstrapInstance ¶
func BootstrapInstance( bootstrapContext environs.BootstrapContext, env environs.Environ, callCtx envcontext.ProviderCallContext, args environs.BootstrapParams, ) (_ *environs.StartInstanceResult, resultBase *corebase.Base, _ environs.CloudBootstrapFinalizer, err error)
BootstrapInstance creates a new instance with the series of its choice, constrained to those of the available tools, and returns the instance result, series, and a function that must be called to finalize the bootstrap process by transferring the tools and installing the initial Juju controller. This method is called by Bootstrap above, which implements environs.Bootstrap, but is also exported so that providers can manipulate the started instance.
func ConfigureMachine ¶
func ConfigureMachine( ctx environs.BootstrapContext, client ssh.Client, host string, instanceConfig *instancecfg.InstanceConfig, sshOptions *ssh.Options, ) error
func CreateStateFile ¶
CreateStateFile creates an empty state file on the given storage, and returns its URL.
func CredentialNotValidError ¶
CredentialNotValidError returns an error that satisfy both Is(err, ErrorCredentialNotValid) and the errors.Locationer interface.
func DefaultHostSSHOptions ¶
DefaultHostSSHOptions returns a nil *ssh.Options, which means to use the defaults; and a no-op cleanup function.
func DeleteStateFile ¶
DeleteStateFile deletes the state file on the given storage.
func Destroy ¶
func Destroy(env environs.Environ, ctx envcontext.ProviderCallContext) error
Destroy is a common implementation of the Destroy method defined on environs.Environ; we strongly recommend that this implementation be used when writing a new provider.
func EnvFullName ¶
EnvFullName returns a string based on the provided model UUID that is suitable for identifying the env on a provider.
The resulting string clearly associates the value with juju, whereas the model's UUID alone isn't very distinctive for humans. This benefits users by helping them quickly identify in their hosting management tools which instances are juju related.
func GetCheckNonceCommand ¶
func GetCheckNonceCommand(instanceConfig *instancecfg.InstanceConfig) string
func HandleCredentialError ¶
func HandleCredentialError(isAuthError func(error) bool, err error, ctx envcontext.ProviderCallContext)
HandleCredentialError determines if a given error relates to an invalid credential.
func MaybeHandleCredentialError ¶
func MaybeHandleCredentialError(isAuthError func(error) bool, err error, ctx envcontext.ProviderCallContext) bool
MaybeHandleCredentialError determines if a given error relates to an invalid credential. If it is, the credential is invalidated and the return bool is true.
func MiBToGiB ¶
MiBToGiB converts the provided megabytes (base-2) into the nearest gigabytes (base-2), rounding up. This is useful for providers that deal in gigabytes (while juju deals in megabytes).
func MinRootDiskSizeGiB ¶
MinRootDiskSizeGiB is the minimum size for the root disk of an instance, in Gigabytes. This value accommodates the anticipated size of the initial image, any updates, and future application data.
func ProviderStateInstances ¶
func ProviderStateInstances(stor storage.StorageReader) ([]instance.Id, error)
ProviderStateInstances extracts the instance IDs from provider-state.
func RemoveStateInstances ¶
RemoveStateInstances removes controller instance IDs from the provider-state file in storage. Instance IDs that are not found in the file are ignored.
func SaveState ¶
func SaveState(storage storage.StorageWriter, state *BootstrapState) error
SaveState writes the given state to the given storage.
func WaitSSH ¶
func WaitSSH( ctx context.Context, stdErr io.Writer, client ssh.Client, checkHostScript string, inst InstanceRefresher, callCtx envcontext.ProviderCallContext, opts environs.BootstrapDialOpts, hostSSHOptions HostSSHOptionsFunc, ) (addr string, err error)
WaitSSH waits for the instance to be assigned a routable address, then waits until we can connect to it via SSH.
waitSSH attempts on all addresses returned by the instance in parallel; the first succeeding one wins. We ensure that private addresses are for the correct machine by checking the presence of a file on the machine that contains the machine's nonce. The "checkHostScript" is a bash script that performs this file check.
Types ¶
type AvailabilityZoneInstances ¶
type AvailabilityZoneInstances struct { // ZoneName is the name of the availability zone. ZoneName string // Instances is a set of instances within the availability zone. Instances []instance.Id }
AvailabilityZoneInstances describes an availability zone and a set of instances in that zone.
func AvailabilityZoneAllocations ¶
func AvailabilityZoneAllocations( env ZonedEnviron, ctx envcontext.ProviderCallContext, group []instance.Id, ) ([]AvailabilityZoneInstances, error)
AvailabilityZoneAllocations returns the availability zones and their instance allocations from the specified group, in ascending order of population. Availability zones with the same population size are ordered by name.
If the specified group is empty, then it will behave as if the result of AllRunningInstances were provided.
type BootstrapState ¶
type BootstrapState struct { // StateInstances are the controllers. StateInstances []instance.Id `yaml:"state-instances"` }
BootstrapState is the state information that is stored in StateFile.
Individual providers may define their own state structures instead of this one, and use their own code for loading and saving those, but this is the definition that most practically useful providers share unchanged.
func LoadState ¶
func LoadState(stor storage.StorageReader) (*BootstrapState, error)
LoadState reads state from the given storage.
type DefaultProvider ¶
type DefaultProvider struct { // Env is the Juju environment that methods target. Env environs.Environ }
DefaultProvider exposes the various common implementations found in this package as methods of a single type. This facilitates treating the implementations as a bundle, e.g. satisfying interfaces.
func (DefaultProvider) BootstrapEnv ¶
func (dp DefaultProvider) BootstrapEnv(ctx environs.BootstrapContext, callCtx envcontext.ProviderCallContext, args environs.BootstrapParams) (*environs.BootstrapResult, error)
BootstrapEnv bootstraps the Juju environment.
func (DefaultProvider) DestroyEnv ¶
func (dp DefaultProvider) DestroyEnv(ctx envcontext.ProviderCallContext) error
DestroyEnv destroys the Juju environment.
type HostSSHOptionsFunc ¶
HostSSHOptionsFunc is a function that, given a hostname, returns an ssh.Options and a cleanup function, or an error.
type InstanceConfigurator ¶
type InstanceConfigurator interface { // DropAllPorts denies access to all ports. DropAllPorts(exceptPorts []int, addr string) error // ChangeIngressRules opens and/or closes ports. ChangeIngressRules(ipAddress string, insert bool, rules firewall.IngressRules) error // FindIngressRules returns all firewall rules. FindIngressRules() (firewall.IngressRules, error) }
func NewSshInstanceConfigurator ¶
func NewSshInstanceConfigurator(host string) InstanceConfigurator
NewSshInstanceConfigurator creates new sshInstanceConfigurator.
type InstanceRefresher ¶
type InstanceRefresher interface { // Refresh refreshes the addresses for the instance. Refresh(ctx envcontext.ProviderCallContext) error // Addresses returns the addresses for the instance. // To ensure that the results are up to date, call // Refresh first. Addresses(ctx envcontext.ProviderCallContext) (network.ProviderAddresses, error) // Status returns the provider-specific status for the // instance. Status(ctx envcontext.ProviderCallContext) instance.Status }
InstanceRefresher is the subet of the Instance interface required for waiting for SSH access to become available.
type RefreshableInstance ¶
func (*RefreshableInstance) Refresh ¶
func (i *RefreshableInstance) Refresh(ctx envcontext.ProviderCallContext) error
Refresh refreshes the addresses for the instance.
type ZonedEnviron ¶
type ZonedEnviron interface { environs.Environ // AvailabilityZones returns all availability zones in the environment. AvailabilityZones(ctx envcontext.ProviderCallContext) (network.AvailabilityZones, error) // InstanceAvailabilityZoneNames returns the names of the availability // zones for the specified instances. The error returned follows the same // rules as Environ.Instances. InstanceAvailabilityZoneNames(ctx envcontext.ProviderCallContext, ids []instance.Id) (map[instance.Id]string, error) // DeriveAvailabilityZones attempts to derive availability zones from // the specified StartInstanceParams. // // The parameters for starting an instance may imply (or explicitly // specify) availability zones, e.g. due to placement, or due to the // attachment of existing volumes, or due to subnet placement. If // there is no such restriction, then DeriveAvailabilityZones should // return an empty string slice to indicate that the caller should // choose an availability zone. DeriveAvailabilityZones(ctx envcontext.ProviderCallContext, args environs.StartInstanceParams) ([]string, error) }
ZonedEnviron is an environs.Environ that has support for availability zones.