pulumi/pkg/resource/stack/deployment.go

387 lines
13 KiB
Go
Raw Normal View History

2018-05-22 21:43:36 +02:00
// Copyright 2016-2018, Pulumi Corporation.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package stack
import (
"encoding/json"
"fmt"
"reflect"
"github.com/blang/semver"
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
"github.com/pkg/errors"
"github.com/pulumi/pulumi/pkg/apitype"
"github.com/pulumi/pulumi/pkg/apitype/migrate"
"github.com/pulumi/pulumi/pkg/resource"
"github.com/pulumi/pulumi/pkg/resource/deploy"
"github.com/pulumi/pulumi/pkg/util/contract"
"github.com/pulumi/pulumi/pkg/workspace"
)
const (
// DeploymentSchemaVersionOldestSupported is the oldest deployment schema that we
// still support, i.e. we can produce a `deploy.Snapshot` from. This will generally
// need to be at least one less than the current schema version so that old deployments can
// be migrated to the current schema.
DeploymentSchemaVersionOldestSupported = 1
)
var (
// ErrDeploymentSchemaVersionTooOld is returned from `DeserializeDeployment` if the
// untyped deployment being deserialized is too old to understand.
ErrDeploymentSchemaVersionTooOld = fmt.Errorf("this stack's deployment is too old")
// ErrDeploymentSchemaVersionTooNew is returned from `DeserializeDeployment` if the
// untyped deployment being deserialized is too new to understand.
ErrDeploymentSchemaVersionTooNew = fmt.Errorf("this stack's deployment version is too new")
)
// SerializeDeployment serializes an entire snapshot as a deploy record.
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
func SerializeDeployment(snap *deploy.Snapshot) *apitype.DeploymentV3 {
contract.Require(snap != nil, "snap")
// Capture the version information into a manifest.
manifest := apitype.ManifestV1{
Time: snap.Manifest.Time,
Magic: snap.Manifest.Magic,
Version: snap.Manifest.Version,
}
for _, plug := range snap.Manifest.Plugins {
var version string
if plug.Version != nil {
version = plug.Version.String()
}
manifest.Plugins = append(manifest.Plugins, apitype.PluginInfoV1{
Name: plug.Name,
Path: plug.Path,
Type: plug.Kind,
Version: version,
})
}
// Serialize all vertices and only include a vertex section if non-empty.
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
var resources []apitype.ResourceV3
for _, res := range snap.Resources {
resources = append(resources, SerializeResource(res))
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
var operations []apitype.OperationV2
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
for _, op := range snap.PendingOperations {
operations = append(operations, SerializeOperation(op))
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
return &apitype.DeploymentV3{
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
Manifest: manifest,
Resources: resources,
PendingOperations: operations,
}
}
// DeserializeUntypedDeployment deserializes an untyped deployment and produces a `deploy.Snapshot`
// from it. DeserializeDeployment will return an error if the untyped deployment's version is
// not within the range `DeploymentSchemaVersionCurrent` and `DeploymentSchemaVersionOldestSupported`.
func DeserializeUntypedDeployment(deployment *apitype.UntypedDeployment) (*deploy.Snapshot, error) {
contract.Require(deployment != nil, "deployment")
switch {
case deployment.Version > apitype.DeploymentSchemaVersionCurrent:
return nil, ErrDeploymentSchemaVersionTooNew
case deployment.Version < DeploymentSchemaVersionOldestSupported:
return nil, ErrDeploymentSchemaVersionTooOld
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
var v3deployment apitype.DeploymentV3
switch deployment.Version {
case 1:
var v1deployment apitype.DeploymentV1
if err := json.Unmarshal([]byte(deployment.Deployment), &v1deployment); err != nil {
return nil, err
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
v2deployment := migrate.UpToDeploymentV2(v1deployment)
v3deployment = migrate.UpToDeploymentV3(v2deployment)
case 2:
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
var v2deployment apitype.DeploymentV2
if err := json.Unmarshal([]byte(deployment.Deployment), &v2deployment); err != nil {
return nil, err
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
v3deployment = migrate.UpToDeploymentV3(v2deployment)
case 3:
if err := json.Unmarshal([]byte(deployment.Deployment), &v3deployment); err != nil {
return nil, err
}
default:
contract.Failf("unrecognized version: %d", deployment.Version)
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
return DeserializeDeploymentV3(v3deployment)
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
// DeserializeDeploymentV3 deserializes a typed DeploymentV3 into a `deploy.Snapshot`.
func DeserializeDeploymentV3(deployment apitype.DeploymentV3) (*deploy.Snapshot, error) {
// Unpack the versions.
manifest := deploy.Manifest{
Time: deployment.Manifest.Time,
Magic: deployment.Manifest.Magic,
Version: deployment.Manifest.Version,
}
for _, plug := range deployment.Manifest.Plugins {
var version *semver.Version
if v := plug.Version; v != "" {
sv, err := semver.ParseTolerant(v)
if err != nil {
return nil, err
}
version = &sv
}
manifest.Plugins = append(manifest.Plugins, workspace.PluginInfo{
Name: plug.Name,
Kind: plug.Type,
Version: version,
})
}
// For every serialized resource vertex, create a ResourceDeployment out of it.
var resources []*resource.State
for _, res := range deployment.Resources {
desres, err := DeserializeResource(res)
if err != nil {
return nil, err
}
resources = append(resources, desres)
}
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
var ops []resource.Operation
for _, op := range deployment.PendingOperations {
desop, err := DeserializeOperation(op)
if err != nil {
return nil, err
}
ops = append(ops, desop)
}
return deploy.NewSnapshot(manifest, resources, ops), nil
}
// SerializeResource turns a resource into a structure suitable for serialization.
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
func SerializeResource(res *resource.State) apitype.ResourceV3 {
contract.Assert(res != nil)
contract.Assertf(string(res.URN) != "", "Unexpected empty resource resource.URN")
// Serialize all input and output properties recursively, and add them if non-empty.
var inputs map[string]interface{}
if inp := res.Inputs; inp != nil {
inputs = SerializeProperties(inp)
}
var outputs map[string]interface{}
if outp := res.Outputs; outp != nil {
outputs = SerializeProperties(outp)
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
return apitype.ResourceV3{
URN: res.URN,
Custom: res.Custom,
Delete: res.Delete,
ID: res.ID,
Type: res.Type,
Parent: res.Parent,
Inputs: inputs,
Outputs: outputs,
Protect: res.Protect,
External: res.External,
Dependencies: res.Dependencies,
InitErrors: res.InitErrors,
Provider: res.Provider,
PropertyDependencies: res.PropertyDependencies,
PendingReplacement: res.PendingReplacement,
}
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
func SerializeOperation(op resource.Operation) apitype.OperationV2 {
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
res := SerializeResource(op.Resource)
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
return apitype.OperationV2{
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
Resource: res,
Type: apitype.OperationType(op.Type),
}
}
// SerializeProperties serializes a resource property bag so that it's suitable for serialization.
func SerializeProperties(props resource.PropertyMap) map[string]interface{} {
dst := make(map[string]interface{})
for _, k := range props.StableKeys() {
if v := SerializePropertyValue(props[k]); v != nil {
dst[string(k)] = v
}
}
return dst
}
// SerializePropertyValue serializes a resource property value so that it's suitable for serialization.
func SerializePropertyValue(prop resource.PropertyValue) interface{} {
// Skip nulls and "outputs"; the former needn't be serialized, and the latter happens if there is an output
// that hasn't materialized (either because we're serializing inputs or the provider didn't give us the value).
if prop.IsComputed() || !prop.HasValue() {
return nil
}
// For arrays, make sure to recurse.
if prop.IsArray() {
srcarr := prop.ArrayValue()
dstarr := make([]interface{}, len(srcarr))
for i, elem := range prop.ArrayValue() {
dstarr[i] = SerializePropertyValue(elem)
}
return dstarr
}
// Also for objects, recurse and use naked properties.
if prop.IsObject() {
return SerializeProperties(prop.ObjectValue())
}
// For assets, we need to serialize them a little carefully, so we can recover them afterwards.
if prop.IsAsset() {
return prop.AssetValue().Serialize()
} else if prop.IsArchive() {
return prop.ArchiveValue().Serialize()
}
if prop.IsSecret() {
//TODO(ellismg): We need to make this error returning (so we can communicate failures in JSON marshalling)
// as well as any errors that occur when encrypting the raw value (which we also need to do!)
value := SerializePropertyValue(prop.SecretValue().Element)
bytes, err := json.Marshal(value)
contract.AssertNoErrorf(err, "marshalling underlying secret value to JSON")
return apitype.SecretV1{
Sig: resource.SecretSig,
Ciphertext: string(bytes),
}
}
// All others are returned as-is.
return prop.V
}
// DeserializeResource turns a serialized resource back into its usual form.
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
func DeserializeResource(res apitype.ResourceV3) (*resource.State, error) {
// Deserialize the resource properties, if they exist.
inputs, err := DeserializeProperties(res.Inputs)
if err != nil {
return nil, err
}
outputs, err := DeserializeProperties(res.Outputs)
if err != nil {
return nil, err
}
return resource.NewState(
res.Type, res.URN, res.Custom, res.Delete, res.ID,
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
inputs, outputs, res.Parent, res.Protect, res.External, res.Dependencies, res.InitErrors, res.Provider,
res.PropertyDependencies, res.PendingReplacement), nil
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
func DeserializeOperation(op apitype.OperationV2) (resource.Operation, error) {
Add a list of in-flight operations to the deployment (#1759) * Add a list of in-flight operations to the deployment This commit augments 'DeploymentV2' with a list of operations that are currently in flight. This information is used by the engine to keep track of whether or not a particular deployment is in a valid state. The SnapshotManager is responsible for inserting and removing operations from the in-flight operation list. When the engine registers an intent to perform an operation, SnapshotManager inserts an Operation into this list and saves it to the snapshot. When an operation completes, the SnapshotManager removes it from the snapshot. From this, the engine can infer that if it ever sees a deployment with pending operations, the Pulumi CLI must have crashed or otherwise abnormally terminated before seeing whether or not an operation completed successfully. To remedy this state, this commit also adds code to 'pulumi stack import' that clears all pending operations from a deployment, as well as code to plan generation that will reject any deployments that have pending operations present. At the CLI level, if we see that we are in a state where pending operations were in-flight when the engine died, we'll issue a human-friendly error message that indicates which resources are in a bad state and how to recover their stack. * CR: Multi-line string literals, renaming in-flight -> pending * CR: Add enum to apitype for operation type, also name status -> type for clarity * Fix the yaml type * Fix missed renames * Add implementation for lifecycle_test.go * Rebase against master
2018-08-11 06:39:59 +02:00
res, err := DeserializeResource(op.Resource)
if err != nil {
return resource.Operation{}, err
}
return resource.NewOperation(res, resource.OperationType(op.Type)), nil
}
// DeserializeProperties deserializes an entire map of deploy properties into a resource property map.
func DeserializeProperties(props map[string]interface{}) (resource.PropertyMap, error) {
result := make(resource.PropertyMap)
for k, prop := range props {
desprop, err := DeserializePropertyValue(prop)
if err != nil {
return nil, err
}
result[resource.PropertyKey(k)] = desprop
}
return result, nil
}
// DeserializePropertyValue deserializes a single deploy property into a resource property value.
func DeserializePropertyValue(v interface{}) (resource.PropertyValue, error) {
if v != nil {
switch w := v.(type) {
case bool:
return resource.NewBoolProperty(w), nil
case float64:
return resource.NewNumberProperty(w), nil
case string:
return resource.NewStringProperty(w), nil
case []interface{}:
var arr []resource.PropertyValue
for _, elem := range w {
ev, err := DeserializePropertyValue(elem)
if err != nil {
return resource.PropertyValue{}, err
}
arr = append(arr, ev)
}
return resource.NewArrayProperty(arr), nil
case map[string]interface{}:
obj, err := DeserializeProperties(w)
if err != nil {
return resource.PropertyValue{}, err
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
// This could be an asset or archive; if so, recover its type.
objmap := obj.Mappable()
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
if sig, hasSig := objmap[resource.SigKey]; hasSig {
switch sig {
case resource.AssetSig:
asset, isasset, err := resource.DeserializeAsset(objmap)
if err != nil {
return resource.PropertyValue{}, err
}
contract.Assert(isasset)
return resource.NewAssetProperty(asset), nil
case resource.ArchiveSig:
archive, isarchive, err := resource.DeserializeArchive(objmap)
if err != nil {
return resource.PropertyValue{}, err
}
contract.Assert(isarchive)
return resource.NewArchiveProperty(archive), nil
case resource.SecretSig:
ciphertext, ok := objmap["ciphertext"].(string)
if !ok {
return resource.PropertyValue{}, errors.New("malformed secret value: missing ciphertext")
}
// TODO(ellismg): The "ciphertext" is actually plaintext right now, so once that changes we'll
// need to decrypt things here.
var elem interface{}
if err := json.Unmarshal([]byte(ciphertext), &elem); err != nil {
return resource.PropertyValue{}, err
}
ev, err := DeserializePropertyValue(elem)
if err != nil {
return resource.PropertyValue{}, err
}
return resource.MakeSecret(ev), nil
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
default:
return resource.PropertyValue{}, errors.Errorf("unrecognized signature '%v' in property map", sig)
}
}
Implement more precise delete-before-replace semantics. (#2369) This implements the new algorithm for deciding which resources must be deleted due to a delete-before-replace operation. We need to compute the set of resources that may be replaced by a change to the resource under consideration. We do this by taking the complete set of transitive dependents on the resource under consideration and removing any resources that would not be replaced by changes to their dependencies. We determine whether or not a resource may be replaced by substituting unknowns for input properties that may change due to deletion of the resources their value depends on and calling the resource provider's Diff method. This is perhaps clearer when described by example. Consider the following dependency graph: A __|__ B C | _|_ D E F In this graph, all of B, C, D, E, and F transitively depend on A. It may be the case, however, that changes to the specific properties of any of those resources R that would occur if a resource on the path to A were deleted and recreated may not cause R to be replaced. For example, the edge from B to A may be a simple dependsOn edge such that a change to B does not actually influence any of B's input properties. In that case, neither B nor D would need to be deleted before A could be deleted. In order to make the above algorithm a reality, the resource monitor interface has been updated to include a map that associates an input property key with the list of resources that input property depends on. Older clients of the resource monitor will leave this map empty, in which case all input properties will be treated as depending on all dependencies of the resource. This is probably overly conservative, but it is less conservative than what we currently implement, and is certainly correct.
2019-01-28 18:46:30 +01:00
// Otherwise, it's just a weakly typed object map.
return resource.NewObjectProperty(obj), nil
default:
contract.Failf("Unrecognized property type: %v", reflect.ValueOf(v))
}
}
return resource.NewNullProperty(), nil
}