A lot of commands used `c.Meta.flagSet()` to create the initial flagset for the command, while quite a few of them didn’t actually use or support the flags that are then added.
So I updated a few commands to use `flag.NewFlagSet()` instead to only add the flags that are actually needed/supported.
Additionally this prevents a few commands from using locking while they actually don’t need locking (as locking is enabled as a default in `c.Meta.flagSet()`.
This connects a missing link left by earlier refactoring: the command
package is responsible for gathering up variable values provided by the
user and passing them through to the backend to use in operations.
In previous work we didn't quite connect these dots. The connection here
is sub-awesome since the existing interfaces here had some unfortunate
assumptions that we'd like to move away from (like the idea of a "nil
backend" implying the local backend) but we're accepting this for now to
avoid another big round of refactoring.
The main implication of this is that we will now always include a backend
configuration in the plan, though it might just be a placeholder config
for the local backend in the remaining cases where that's still implicitly
set. Later we will change this so that there is no implicit local backend
at all (terraform init is always required, _it_ will deal with setting
implicitly setting the local backend when appropriate), which will allow
us to rework this to be more straightforward and less "spooky".
Due to how often the state and plan types are referenced throughout
Terraform, there isn't a great way to switch them out gradually. As a
consequence, this huge commit gets us from the old world to a _compilable_
new world, but still has a large number of known test failures due to
key functionality being stubbed out.
The stubs here are for anything that interacts with providers, since we
now need to do the follow-up work to similarly replace the old
terraform.ResourceProvider interface with its replacement in the new
"providers" package. That work, along with work to fix the remaining
failing tests, will follow in subsequent commits.
The aim here was to replace all references to terraform.State and its
downstream types with states.State, terraform.Plan with plans.Plan,
state.State with statemgr.State, and switch to the new implementations of
the state and plan file formats. However, due to the number of times those
types are used, this also ended up affecting numerous other parts of core
such as terraform.Hook, the backend.Backend interface, and most of the CLI
commands.
Just as with 5861dbf3fc49b19587a31816eb06f511ab861bb4 before, I apologize
in advance to the person who inevitably just found this huge commit while
spelunking through the commit history.
This is a rather-messy, complex change to get the "command" package
building again against the new backend API that was updated for
the new configuration loader.
A lot of this is mechanical rewriting to the new API, but
meta_config.go and meta_backend.go in particular saw some major
changes to interface with the new loader APIs and to deal with
the change in order of steps in the backend API.
In some cases this is needed to keep the UX clean and to make sure any remote exit codes are passed through to the local process.
The most obvious example for this is when using the "remote" backend. This backend runs Terraform remotely and stream the output back to the local terminal.
When an error occurs during the remote execution, all the needed error information will already be in the streamed output. So if we then return an error ourselves, users will get the same errors twice.
By allowing the backend to specify the correct exit code, the UX remains the same while preserving the correct exit codes.
If the user wishes to interrupt the running operation, only the first
interrupt was communicated to the operation by canceling the provided
context. A second interrupt would start the shutdown process, but not
communicate this to the running operation. This order of event could
cause partial writes of state.
What would happen is that once the command returns, the plugin system
would stop the provider processes. Once the provider processes dies, all
pending Eval operations would return return with an error, and quickly
cause the operation to complete. Since the backend code didn't know that
the process was shutting down imminently, it would continue by
attempting to write out the last known state. Under the right
conditions, the process would exit part way through the writing of the
state file.
Add Stop and Cancel CancelFuncs to the RunningOperation, to allow it to
easily differentiate between the two signals. The backend will then be
able to detect a shutdown and abort more gracefully.
In order to ensure that the backend is not in the process of writing the
state out, the command will always attempt to wait for the process to
complete after cancellation.
Previously we required callers to separately call .Validate on the root
module to determine if there were any value errors, but we did that
inconsistently and would thus see crashes in some cases where later code
would try to use invalid configuration as if it were valid.
Now we run .Validate automatically after config loading, returning the
resulting diagnostics. Since we return a diagnostics here, it's possible
to return both warnings and errors.
We return the loaded module even if it's invalid, so callers are free to
ignore returned errors and try to work with the config anyway, though they
will need to be defensive against invalid configuration themselves in
that case.
As a result of this, all of the commands that load configuration now need
to use diagnostic printing to signal errors. For the moment this just
allows us to return potentially-multiple config errors/warnings in full
fidelity, but also sets us up for later when more subsystems are able
to produce rich diagnostics so we can show them all together.
Finally, this commit also removes some stale, commented-out code for the
"legacy" (pre-0.8) graph implementation, which has not been available
for some time.
Now that the local backend can be cancelled during plan and refresh, we
don't really need the testShutdownHook. Simplify the tests by just
checking for Stop being called on the provider.
Add a shutdown hook to verify that a context has been correctly
cancelled, so we can remove the sleep and stop guessing.
Add a plan version of the shutdown test as well.
There was no cancellation context for a plan, so it would always have to
run to completion as SIGINT was being swallowed.
Move the shutdown channel to the command Meta since it's used in
multiple commands.
This uses the new diagnostics printer for config-related errors in the
main five commands that deal with config.
The immediate motivation for this is to allow HCL2-produced diagnostics
to be printed out in their full fidelity, though it also slightly changes
the presentation of other errors so that they are not presented in all
red text, which can be hard to read on some terminals.
Instead of providing the a path in BackendOpts, provide a loaded
*config.Config instead. This reduces the number of places where
configuration is loaded.
Add the -lock-timeout flag to the appropriate commands.
Add the -lock flag to `init` and `import` which were missing it.
Set both stateLock and stateLockTimeout in Meta.flagsSet, and remove the
extra references for clarity.
Implement debugInfo and the DebugGraph
DebugInfo will be a global variable through which graph debug
information can we written to a compressed archive. The DebugInfo
methods are all safe for concurrent use, and noop with a nil receiver.
The API outside of the terraform package will be to call SetDebugInfo
to create the archive, and CloseDebugInfo() to properly close the file.
Each write to the archive will be flushed and sync'ed individually, so
in the event of a crash or a missing call to Close, the archive can
still be recovered.
The DebugGraph is a representation of a terraform Graph to be written to
the debug archive, currently in dot format. The DebugGraph also contains
an internal buffer with Printf and Write methods to add to this buffer.
The buffer will be written to an accompanying file in the debug archive
along with the graph.
This also adds a GraphNodeDebugger interface. Any node implementing
`NodeDebug() string` can output information to annotate the debug graph
node, and add the data to the log. This interface may change or be
removed to provide richer options for debugging graph nodes.
The new graph builders all delegate the build to the BasicGraphBuilder.
Having a Name field lets us differentiate the actual builder
implementation in the debug graphs.
Internally a data source read is represented as a creation diff for the
resource, but in the UI we'll show it as a distinct icon and color so that
the user can more easily understand that these operations won't affect
any real infrastructure.
Unfortunately by the time we get to formatting the plan in the UI we
only have the resource names to work with, and can't get at the original
resource mode. Thus we're forced to infer the resource mode by exploiting
knowledge of the naming scheme.
This means that terraform commands like `plan`, `apply`, `show`, and
`graph` will expand all modules by default.
While modules-as-black-boxes is still very true in the conceptual design
of modules, feedback on this behavior has consistently suggested that
users would prefer to see more verbose output by default.
The `-module-depth` flag and env var are retained to allow output to be
optionally limited / summarized by these commands.
Previously the plan summary output would consider -/+ diffs as changes
even though they actually destroy and create instances. This was
misleadning and inconsistent with the accounting that gets done for the
similar summary written out after "apply".
Instead we now count the -/+ diffs as both adds and removes, which should
mean that the counts output in the plan summary should match those in
the apply summary, as long as no errors occur during apply.
This fixes#3163.
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.
Sander van Harmelen