This is a "should never happen" case, but we have reports of it actually
happening. In order to try to collect a bit more data about what's going
on here, we're changing what was previously a hard panic into a normal
error message that can include the address of the instance we were working
on and the action we were trying to do to it at the time.
The hope is to narrow down what situations can trigger this in order to
find a reliable reproduction case in order to debug further. This also
means that for those who _do_ encounter this problem in the meantime
Terraform will have a chance to shut down cleanly and therefore be more
likely to be able to recover on a subsequent plan/apply cycle.
Further investigation of this will follow once we see a report or two of
this updated error message.
Since a planned destroy can no longer indicate it is a full destroy,
unused values were being left in the apply graph for evaluation. If
these values contains interpolations that can fail, (for example, a
zipmap with mismatched list sizes), it will cause the apply to abort.
The PrunUnusedValuesTransformer was only previously run during destroy,
more out of conservatism than for any other particular reason. Adapt it
to always remove unused values from the graph, with the exception being
the root module outputs, which must be retained when we don't have a
clear indication that a full destroy is being executed.
The resource cleanup node does not need a provider. We can't directly
remove the ProvidedBy method, but this node only needs to be eval-able
so we can remove all the NodeAbstractResource methods at once.
This also includes an upgrade to cty v1.1.1 because HCL calls for it.
The changes in these two libraries are mainly to codepaths that don't
directly affect Terraform, but including this upgrade will cause some
small improvements to Terraform's error messages for type conversion
problems.
References from a resource-level connection blocks were not returned
from NodeAbstractResource.References, causing the provisioner connection
attributes to sometimes be evaluated too early.
If there is an error when opening the session for agent forwarding in
the process ssh connention, there is a deadlock when recursively
calling Connect on an internal reattempt. Avoid that, and let the
connection be reattempted externally.
Right now, the only environment variable available is the same
environment variable that will be picked up by the GCP provider. Users
would like to be able to store state in separate projects or accounts or
otherwise authenticate to the provider with a service account that
doesn't have access to the state. This seems like a reasonable enough
practice to me, and the solution seems straightforward--offer an
environment variable that doesn't mean anything to the provider to
configure the backend credentials. I've added GOOGLE_BACKEND_CREDENTIALS
to manage just the backend credentials, and documented it appropriately.
It's a common source of errors to try to produce JSON or YAML syntax
using string concatenation via our template language but to miss some
details like correct string escaping, quoting, required commas, etc.
The jsonencode and yamlencode functions are a better way to generate JSON
and YAML, but it's not immediately obvious that both of these functions
are available for use in external templates (via templatefile) too.
Given that questions related to this come up a lot in our community forum
and elsewhere, it seems worth having a documentation section to show the
pattern of having a template that consists only of a single function call.
* configs: move ProviderConfigCompact[Str] from addrs to configs
The configs package is aware of provider name and type (which are the
same thing today, but expected to be two different things in a future
release), and should be the source of truth for a provider config
address. This is an intermediate step; the next step will change the returned types to something based in the configs package.
* command: rename choosePlugins to chooseProviders to clarify scope of function
* use `Provider.LegacyString()` (instead of `Provider.Type`) consistently
* explicitly create legacy-style provider (continuing from above change)
When warnings appear in isolation (not accompanied by an error) it's
reasonable to want to defer resolving them for a while because they are
not actually blocking immediate work.
However, our warning messages tend to be long by default in order to
include all of the necessary context to understand the implications of
the warning, and that can make them overwhelming when combined with other
output.
As a compromise, this adds a new CLI option -compact-warnings which is
supported for all the main operation commands and which uses a more
compact format to print out warnings as long as they aren't also
accompanied by errors.
The default remains unchanged except that the threshold for consolidating
warning messages is reduced to one so that we'll now only show one of
each distinct warning summary.
Full warning messages are always shown if there's at least one error
included in the diagnostic set too, because in that case the warning
message could contain additional context to help understand the error.
Traditionally we've preferred to release new language features in major
releases only, because we can then use the beta cycle to gather feedback
on the feature and learn about any usability challenges or other
situations we didn't consider during our design in time to make those
changes before inclusion in a stable release.
This "experiments" feature is intended to decouple the feedback cycle for
new features from the major release rhythm, and thus allow us to release
new features in minor releases by first releasing them as experimental for
a minor release or two, adjust for any feedback gathered during that
period, and then finally remove the experiment gate and enable the feature
for everyone.
The intended model here is that anything behind an experiment gate is
subject to breaking changes even in patch releases, and so any module
using these experimental features will be broken by a future Terraform
upgrade.
The behavior implemented here is:
- Recognize a new "experiments" setting in the "terraform" block which
allows module authors to explicitly opt in to experimental features.
terraform {
experiments = [resource_for_each]
}
- Generate a warning whenever loading a module that has experiments
enabled, to avoid accidentally depending on experimental features and
thus risking unexpected breakage on next Terraform upgrade.
- We check the enabled experiments against the configuration at module
load time, which means that experiments are scoped to a particular
module. Enabling an experiment in one module does not automatically
enable it in any other module.
This experiments mechanism is itself an experiment, and so I'd like to
use the resource for_each feature to trial it. Because any configuration
using experiments is subject to breaking changes, we are free to adjust
this experiments feature in future releases as we see fit, but once
for_each is shipped without an experiment gate we'll be blocked from
making significant changes to it until the next major release at least.
Martin Atkins