Only used in targets for now. The plan is to use this for interpolation
as well.
This allows us to target:
* individual resources expanded by `count` using bracket / index notation.
* deposed / tainted resources with an `InstanceType` field after name
Docs to follow.
Add `-target=resource` flag to core operations, allowing users to
target specific resources in their infrastructure. When `-target` is
used, the operation will only apply to that resource and its
dependencies.
The calculated dependencies are different depending on whether we're
running a normal operation or a `terraform destroy`.
Generally, "dependencies" refers to ancestors: resources falling
_before_ the target in the graph, because their changes are required to
accurately act on the target.
For destroys, "dependencies" are descendents: those resources which fall
_after_ the target. These resources depend on our target, which is going
to be destroyed, so they should also be destroyed.
We were previously only recording the schema version on refresh. This
caused the state to be incorrectly written after a `terraform apply`
causing subsequent commands to run the state through an unnecessary
migration.
Providers get a per-resource SchemaVersion integer that they can bump
when a resource's schema changes format. Each InstanceState with an
older recorded SchemaVersion than the cureent one is yielded to a
`MigrateSchema` function to be transformed such that it can be addressed
by the current version of the resource's Schema.
Deposed instances need to be stored as a list for certain pathological
cases where destroys fail for some reason (e.g. upstream API failure,
Terraform interrupted mid-run). Terraform needs to be able to remember
all Deposed nodes so that it can clean them up properly in subsequent
runs.
Deposed instances will now never touch the Tainted list - they're fully
managed from within their own list.
Added a "multiDepose" test case that walks through a scenario to
exercise this.
After a lot of fun debugging with @mitchellh we finally have a diagnosis
for #1056.
I'm going to attempt to reproduce the diagnosis in prose to test out my
understanding.
----
The `DestroyTransformer` runs twice:
* `DestroyPrimary` mode creates destroy nodes for normal resources
* `DestroyTainted` mode creates destroy nodes for tainted resources
These destroy nodes are specializations of `GraphConfigNode`, which has
a `DynamicExpand` step.
In `DynamicExpand` we have a race condition between the normal destroy
node and the tainted destroy node for a given resource when
`CreateBeforeDestroy` is off.
The `DestroyTainted` `GraphConfigNode` must run the `TaintedTransform`
to draw out tainted nodes, since it is reponsible for destroying them
for replacement.
The `DestroyPrimary` `GraphConfigNode` _also_ runs `TaintedTransform` -
this is to catch `Deposed` nodes from CBD, which are piggy backing on
the end of the `Tainted` list.
Here's the bug: when CBD is off and you start an apply with a tainted
resource in your state, both `DestroyPrimary` and `DestroyTainted` catch
it and create their own `graphNodeTaintedResource` in their respective
subgraphs.
If parallelism is disabled, this doesn't happen because the
`DestroyPrimary` subgraph resolves completely before the
`DestroyTainted` node does its `DynamicExpand`, so the `Tainted` list
has been cleared by the time `DestroyTainted` is expanding. With
parallelism, each of these two subgraphs plays out in its own goroutine
simultaneously, and each picks up the tainted resource(s) that the apply
starts with. So you get two `graphNodeTaintedResource` nodes, and two
destroys.
This band-aid fixes that by skipping the TaintedTransform alltogether in
the `DestroyPrimary` node if CBD is off.
A better fix will follow, which involves reworking the `Deposed` concept
so it no longer piggybacks on `Tainted`.
fixes#1056
Paul Hinze