opentofu/terraform/eval_state.go
Martin Atkins 12572e97bc core: Automatically upgrade resource instance states on read
If an instance object in state has an earlier schema version number then
it is likely that the schema we're holding won't be able to decode the
raw data that is stored. Instead, we must ask the provider to upgrade it
for us first, which might also include translating it from flatmap form
if it was last updated with a Terraform version earlier than v0.12.

This ends up being a "seam" between our use of int64 for schema versions
in the providers package and uint64 everywhere else. We intend to
standardize on int64 everywhere eventually, but for now this remains
consistent with existing usage in each layer to keep the type conversion
noise contained here and avoid mass-updates to other Terraform components
at this time.

This also includes a minor change to the test helpers for the
backend/local package, which were inexplicably setting a SchemaVersion of
1 on the basic test state but setting the mock schema version to zero,
creating an invalid situation where the state would need to be downgraded.
2018-11-30 11:22:39 -08:00

466 lines
16 KiB
Go

package terraform
import (
"fmt"
"log"
"github.com/hashicorp/terraform/addrs"
"github.com/hashicorp/terraform/configs"
"github.com/hashicorp/terraform/providers"
"github.com/hashicorp/terraform/states"
"github.com/hashicorp/terraform/tfdiags"
)
// EvalReadState is an EvalNode implementation that reads the
// current object for a specific instance in the state.
type EvalReadState struct {
// Addr is the address of the instance to read state for.
Addr addrs.ResourceInstance
// ProviderSchema is the schema for the provider given in Provider.
ProviderSchema **ProviderSchema
// Provider is the provider that will subsequently perform actions on
// the the state object. This is used to perform any schema upgrades
// that might be required to prepare the stored data for use.
Provider *providers.Interface
// Output will be written with a pointer to the retrieved object.
Output **states.ResourceInstanceObject
}
func (n *EvalReadState) Eval(ctx EvalContext) (interface{}, error) {
if n.Provider == nil || *n.Provider == nil {
panic("EvalReadState used with no Provider object")
}
if n.ProviderSchema == nil || *n.ProviderSchema == nil {
panic("EvalReadState used with no ProviderSchema object")
}
absAddr := n.Addr.Absolute(ctx.Path())
log.Printf("[TRACE] EvalReadState: reading state for %s", absAddr)
src := ctx.State().ResourceInstanceObject(absAddr, states.CurrentGen)
if src == nil {
// Presumably we only have deposed objects, then.
log.Printf("[TRACE] EvalReadState: no state present for %s", absAddr)
return nil, nil
}
schema, currentVersion := (*n.ProviderSchema).SchemaForResourceAddr(n.Addr.ContainingResource())
if schema == nil {
// Shouldn't happen since we should've failed long ago if no schema is present
return nil, fmt.Errorf("no schema available for %s while reading state; this is a bug in Terraform and should be reported", absAddr)
}
var diags tfdiags.Diagnostics
src, diags = UpgradeResourceState(absAddr, *n.Provider, src, schema, currentVersion)
if diags.HasErrors() {
// Note that we don't have any channel to return warnings here. We'll
// accept that for now since warnings during a schema upgrade would
// be pretty weird anyway, since this operation is supposed to seem
// invisible to the user.
return nil, diags.Err()
}
obj, err := src.Decode(schema.ImpliedType())
if err != nil {
return nil, err
}
if n.Output != nil {
*n.Output = obj
}
return obj, nil
}
// EvalReadStateDeposed is an EvalNode implementation that reads the
// deposed InstanceState for a specific resource out of the state
type EvalReadStateDeposed struct {
// Addr is the address of the instance to read state for.
Addr addrs.ResourceInstance
// Key identifies which deposed object we will read.
Key states.DeposedKey
// ProviderSchema is the schema for the provider given in Provider.
ProviderSchema **ProviderSchema
// Provider is the provider that will subsequently perform actions on
// the the state object. This is used to perform any schema upgrades
// that might be required to prepare the stored data for use.
Provider *providers.Interface
// Output will be written with a pointer to the retrieved object.
Output **states.ResourceInstanceObject
}
func (n *EvalReadStateDeposed) Eval(ctx EvalContext) (interface{}, error) {
if n.Provider == nil || *n.Provider == nil {
panic("EvalReadStateDeposed used with no Provider object")
}
if n.ProviderSchema == nil || *n.ProviderSchema == nil {
panic("EvalReadStateDeposed used with no ProviderSchema object")
}
key := n.Key
if key == states.NotDeposed {
return nil, fmt.Errorf("EvalReadStateDeposed used with no instance key; this is a bug in Terraform and should be reported")
}
absAddr := n.Addr.Absolute(ctx.Path())
log.Printf("[TRACE] EvalReadStateDeposed: reading state for %s deposed object %s", absAddr, n.Key)
src := ctx.State().ResourceInstanceObject(absAddr, key)
if src == nil {
// Presumably we only have deposed objects, then.
log.Printf("[TRACE] EvalReadStateDeposed: no state present for %s deposed object %s", absAddr, n.Key)
return nil, nil
}
schema, currentVersion := (*n.ProviderSchema).SchemaForResourceAddr(n.Addr.ContainingResource())
if schema == nil {
// Shouldn't happen since we should've failed long ago if no schema is present
return nil, fmt.Errorf("no schema available for %s while reading state; this is a bug in Terraform and should be reported", absAddr)
}
var diags tfdiags.Diagnostics
src, diags = UpgradeResourceState(absAddr, *n.Provider, src, schema, currentVersion)
if diags.HasErrors() {
// Note that we don't have any channel to return warnings here. We'll
// accept that for now since warnings during a schema upgrade would
// be pretty weird anyway, since this operation is supposed to seem
// invisible to the user.
return nil, diags.Err()
}
obj, err := src.Decode(schema.ImpliedType())
if err != nil {
return nil, err
}
if n.Output != nil {
*n.Output = obj
}
return obj, nil
}
// EvalRequireState is an EvalNode implementation that exits early if the given
// object is null.
type EvalRequireState struct {
State **states.ResourceInstanceObject
}
func (n *EvalRequireState) Eval(ctx EvalContext) (interface{}, error) {
if n.State == nil {
return nil, EvalEarlyExitError{}
}
state := *n.State
if state == nil || state.Value.IsNull() {
return nil, EvalEarlyExitError{}
}
return nil, nil
}
// EvalUpdateStateHook is an EvalNode implementation that calls the
// PostStateUpdate hook with the current state.
type EvalUpdateStateHook struct{}
func (n *EvalUpdateStateHook) Eval(ctx EvalContext) (interface{}, error) {
// In principle we could grab the lock here just long enough to take a
// deep copy and then pass that to our hooks below, but we'll instead
// hold the hook for the duration to avoid the potential confusing
// situation of us racing to call PostStateUpdate concurrently with
// different state snapshots.
stateSync := ctx.State()
state := stateSync.Lock().DeepCopy()
defer stateSync.Unlock()
// Call the hook
err := ctx.Hook(func(h Hook) (HookAction, error) {
return h.PostStateUpdate(state)
})
if err != nil {
return nil, err
}
return nil, nil
}
// EvalWriteState is an EvalNode implementation that saves the given object
// as the current object for the selected resource instance.
type EvalWriteState struct {
// Addr is the address of the instance to read state for.
Addr addrs.ResourceInstance
// State is the object state to save.
State **states.ResourceInstanceObject
// ProviderSchema is the schema for the provider given in ProviderAddr.
ProviderSchema **ProviderSchema
// ProviderAddr is the address of the provider configuration that
// produced the given object.
ProviderAddr addrs.AbsProviderConfig
}
func (n *EvalWriteState) Eval(ctx EvalContext) (interface{}, error) {
if n.State == nil {
// Note that a pointer _to_ nil is valid here, indicating the total
// absense of an object as we'd see during destroy.
panic("EvalWriteState used with no ResourceInstanceObject")
}
absAddr := n.Addr.Absolute(ctx.Path())
state := ctx.State()
obj := *n.State
if obj == nil || obj.Value.IsNull() {
// No need to encode anything: we'll just write it directly.
state.SetResourceInstanceCurrent(absAddr, nil, n.ProviderAddr)
log.Printf("[TRACE] EvalWriteState: removing state object for %s", absAddr)
return nil, nil
}
if n.ProviderSchema == nil || *n.ProviderSchema == nil {
// Should never happen, unless our state object is nil
panic("EvalWriteState used with pointer to nil ProviderSchema object")
}
if obj != nil {
log.Printf("[TRACE] EvalWriteState: writing current state object for %s", absAddr)
} else {
log.Printf("[TRACE] EvalWriteState: removing current state object for %s", absAddr)
}
schema, currentVersion := (*n.ProviderSchema).SchemaForResourceAddr(n.Addr.ContainingResource())
if schema == nil {
// It shouldn't be possible to get this far in any real scenario
// without a schema, but we might end up here in contrived tests that
// fail to set up their world properly.
return nil, fmt.Errorf("failed to encode %s in state: no resource type schema available", absAddr)
}
src, err := obj.Encode(schema.ImpliedType(), currentVersion)
if err != nil {
return nil, fmt.Errorf("failed to encode %s in state: %s", absAddr, err)
}
state.SetResourceInstanceCurrent(absAddr, src, n.ProviderAddr)
return nil, nil
}
// EvalWriteStateDeposed is an EvalNode implementation that writes
// an InstanceState out to the Deposed list of a resource in the state.
type EvalWriteStateDeposed struct {
// Addr is the address of the instance to read state for.
Addr addrs.ResourceInstance
// Key indicates which deposed object to write to.
Key states.DeposedKey
// State is the object state to save.
State **states.ResourceInstanceObject
// ProviderSchema is the schema for the provider given in ProviderAddr.
ProviderSchema **ProviderSchema
// ProviderAddr is the address of the provider configuration that
// produced the given object.
ProviderAddr addrs.AbsProviderConfig
}
func (n *EvalWriteStateDeposed) Eval(ctx EvalContext) (interface{}, error) {
if n.State == nil {
// Note that a pointer _to_ nil is valid here, indicating the total
// absense of an object as we'd see during destroy.
panic("EvalWriteStateDeposed used with no ResourceInstanceObject")
}
absAddr := n.Addr.Absolute(ctx.Path())
key := n.Key
state := ctx.State()
if key == states.NotDeposed {
// should never happen
return nil, fmt.Errorf("can't save deposed object for %s without a deposed key; this is a bug in Terraform that should be reported", absAddr)
}
obj := *n.State
if obj == nil {
// No need to encode anything: we'll just write it directly.
state.SetResourceInstanceDeposed(absAddr, key, nil, n.ProviderAddr)
log.Printf("[TRACE] EvalWriteStateDeposed: removing state object for %s deposed %s", absAddr, key)
return nil, nil
}
if n.ProviderSchema == nil || *n.ProviderSchema == nil {
// Should never happen, unless our state object is nil
panic("EvalWriteStateDeposed used with no ProviderSchema object")
}
schema, currentVersion := (*n.ProviderSchema).SchemaForResourceAddr(n.Addr.ContainingResource())
if schema == nil {
// It shouldn't be possible to get this far in any real scenario
// without a schema, but we might end up here in contrived tests that
// fail to set up their world properly.
return nil, fmt.Errorf("failed to encode %s in state: no resource type schema available", absAddr)
}
src, err := obj.Encode(schema.ImpliedType(), currentVersion)
if err != nil {
return nil, fmt.Errorf("failed to encode %s in state: %s", absAddr, err)
}
log.Printf("[TRACE] EvalWriteStateDeposed: writing state object for %s deposed %s", absAddr, key)
state.SetResourceInstanceDeposed(absAddr, key, src, n.ProviderAddr)
return nil, nil
}
// EvalDeposeState is an EvalNode implementation that moves the current object
// for the given instance to instead be a deposed object, leaving the instance
// with no current object.
// This is used at the beginning of a create-before-destroy replace action so
// that the create can create while preserving the old state of the
// to-be-destroyed object.
type EvalDeposeState struct {
Addr addrs.ResourceInstance
// ForceKey, if a value other than states.NotDeposed, will be used as the
// key for the newly-created deposed object that results from this action.
// If set to states.NotDeposed (the zero value), a new unique key will be
// allocated.
ForceKey states.DeposedKey
// OutputKey, if non-nil, will be written with the deposed object key that
// was generated for the object. This can then be passed to
// EvalUndeposeState.Key so it knows which deposed instance to forget.
OutputKey *states.DeposedKey
}
// TODO: test
func (n *EvalDeposeState) Eval(ctx EvalContext) (interface{}, error) {
absAddr := n.Addr.Absolute(ctx.Path())
state := ctx.State()
var key states.DeposedKey
if n.ForceKey == states.NotDeposed {
key = state.DeposeResourceInstanceObject(absAddr)
} else {
key = n.ForceKey
state.DeposeResourceInstanceObjectForceKey(absAddr, key)
}
log.Printf("[TRACE] EvalDeposeState: prior object for %s now deposed with key %s", absAddr, key)
if n.OutputKey != nil {
*n.OutputKey = key
}
return nil, nil
}
// EvalMaybeRestoreDeposedObject is an EvalNode implementation that will
// restore a particular deposed object of the specified resource instance
// to be the "current" object if and only if the instance doesn't currently
// have a current object.
//
// This is intended for use when the create leg of a create before destroy
// fails with no partial new object: if we didn't take any action, the user
// would be left in the unfortunate situation of having no current object
// and the previously-workign object now deposed. This EvalNode causes a
// better outcome by restoring things to how they were before the replace
// operation began.
//
// The create operation may have produced a partial result even though it
// failed and it's important that we don't "forget" that state, so in that
// situation the prior object remains deposed and the partial new object
// remains the current object, allowing the situation to hopefully be
// improved in a subsequent run.
type EvalMaybeRestoreDeposedObject struct {
Addr addrs.ResourceInstance
// Key is a pointer to the deposed object key that should be forgotten
// from the state, which must be non-nil.
Key *states.DeposedKey
}
// TODO: test
func (n *EvalMaybeRestoreDeposedObject) Eval(ctx EvalContext) (interface{}, error) {
absAddr := n.Addr.Absolute(ctx.Path())
dk := *n.Key
state := ctx.State()
restored := state.MaybeRestoreResourceInstanceDeposed(absAddr, dk)
if restored {
log.Printf("[TRACE] EvalMaybeRestoreDeposedObject: %s deposed object %s was restored as the current object", absAddr, dk)
} else {
log.Printf("[TRACE] EvalMaybeRestoreDeposedObject: %s deposed object %s remains deposed", absAddr, dk)
}
return nil, nil
}
// EvalWriteResourceState is an EvalNode implementation that ensures that
// a suitable resource-level state record is present in the state, if that's
// required for the "each mode" of that resource.
//
// This is important primarily for the situation where count = 0, since this
// eval is the only change we get to set the resource "each mode" to list
// in that case, allowing expression evaluation to see it as a zero-element
// list rather than as not set at all.
type EvalWriteResourceState struct {
Addr addrs.Resource
Config *configs.Resource
ProviderAddr addrs.AbsProviderConfig
}
// TODO: test
func (n *EvalWriteResourceState) Eval(ctx EvalContext) (interface{}, error) {
var diags tfdiags.Diagnostics
absAddr := n.Addr.Absolute(ctx.Path())
state := ctx.State()
count, countDiags := evaluateResourceCountExpression(n.Config.Count, ctx)
diags = diags.Append(countDiags)
if countDiags.HasErrors() {
return nil, diags.Err()
}
// Currently we ony support NoEach and EachList, because for_each support
// is not fully wired up across Terraform. Once for_each support is added,
// we'll need to handle that here too, setting states.EachMap if the
// assigned expression is a map.
eachMode := states.NoEach
if count >= 0 { // -1 signals "count not set"
eachMode = states.EachList
}
// This method takes care of all of the business logic of updating this
// while ensuring that any existing instances are preserved, etc.
state.SetResourceMeta(absAddr, eachMode, n.ProviderAddr)
return nil, nil
}
// EvalForgetResourceState is an EvalNode implementation that prunes out an
// empty resource-level state for a given resource address, or produces an
// error if it isn't empty after all.
//
// This should be the last action taken for a resource that has been removed
// from the configuration altogether, to clean up the leftover husk of the
// resource in the state after other EvalNodes have destroyed and removed
// all of the instances and instance objects beneath it.
type EvalForgetResourceState struct {
Addr addrs.Resource
}
func (n *EvalForgetResourceState) Eval(ctx EvalContext) (interface{}, error) {
absAddr := n.Addr.Absolute(ctx.Path())
state := ctx.State()
pruned := state.RemoveResourceIfEmpty(absAddr)
if !pruned {
// If this produces an error, it indicates a bug elsewhere in Terraform
// -- probably missing graph nodes, graph edges, or
// incorrectly-implemented evaluation steps.
return nil, fmt.Errorf("orphan resource %s still has a non-empty state after apply; this is a bug in Terraform", absAddr)
}
log.Printf("[TRACE] EvalForgetResourceState: Pruned husk of %s from state", absAddr)
return nil, nil
}