The existing "type" argument allows specifying a type constraint that
allows for some basic validation, but often there are more constraints on
a variable value than just its type.
This new feature (requiring an experiment opt-in for now, while we refine
it) allows specifying arbitrary validation rules for any variable which
can then cause custom error messages to be returned when a caller provides
an inappropriate value.
variable "example" {
validation {
condition = var.example != "nope"
error_message = "Example value must not be \"nope\"."
}
}
The core parts of this are designed to do as little new work as possible
when no validations are specified, and thus the main new checking codepath
here can therefore only run when the experiment is enabled in order to
permit having validations.
These are intended to make it easier to work with arbitrary data
structures whose shape might not be known statically, such as the result
of jsondecode(...) or yamldecode(...) of data from a separate system.
For example, in an object value which has attributes that may or may not
be set we can concisely provide a fallback value to use when the attribute
isn't set:
try(local.example.foo, "fallback-foo")
Using a "try to evaluate" model rather than explicit testing fits better
with the usual programming model of the Terraform language where values
are normally automatically converted to the necessary type where possible:
the given expression is subject to all of the same normal type conversions,
which avoids inadvertently creating a more restrictive evaluation model
as might happen if this were handled using checks like a hypothetical
isobject(...) function, etc.
In earlier versions of Terraform the result of terraform state show was
in the pre-0.12 "flatmap" structure that was unable to reflect nested
data structures. That was fixed in Terraform 0.12, but as a consequence
this statement about the output being machine-parseable (which was
debateable even in older versions) is incorrect.
Fortunately, we now have "terraform show -json" to get output that is
intentionally machine-parseable, so we'll recommend to use that instead
here. The JSON output of that command is a superset of what's produced by
"terraform state show", so should be usable to meet any use-case that
might previously have been met by parsing the "terraform state show"
output.
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.
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.
I've seen folks ask about how to express this in resource address syntax
a number of times now, so adding this example here to illustrate how it
looks when there are multiple levels of module to traverse through.
This is redundant with other information further up the page, but having
it as an entirely separate example gives an opportunity to include more
introductory text to explain what the example is showing.
There are some differences between the Terraform CLI and Terraform Cloud ideas of workspaces.
This documentation aims to explain those differences and show different patterns for configuring the remote backend and the implications of different approaches.
As mentioned in #17871 the current example can hide the fact that the module
path plays an important role. The example's explanation is expanded.
Moreover, the verb "attach" is replaced with "map" to make the vocabulary
consistent with the wording in the documentation of the terraform state.
A very common question since we launched the two repetition constructs
is how to deal with situations where the input data structure doesn't
match one-to-one with the desired configuration.
This adds some full worked examples of two common situations that have
come up in questions. To avoid adding a lot of extra content to the
already-large "expressions" and "resources" pages, the main bulk of this
new content lives with the relevant functions themselves as a full example
of one thing they are good for, and then we'll link to them from the two
general documentation sections where folks are likely to be reading when
they encounter the problem.
* upstream/master: (66 commits)
lang/eval: more evalContext fixups
Update CHANGELOG.md
Cleanup after v0.12.10 release
v0.12.10
website / help: reconcile 'validate' command docs
website: Document behavior of `self` object for provisioners
vendor: switch to HCL 2.0 in the HCL repository
Update communicator/ssh/communicator.go
copy client pointer for keep-alive loop
Update CHANGELOG.md
slow down tfce polling to 1s
typos. some code, some text.
Remove -check-variables flag from the docs
Merge cleanup, remove `license` parameter in favor of bool `accept_license`, adjust how license acceptance is done, update hab provisioner doc.
vendor latest go-tfe
clean up go mod for go-tfe
tfce test additions
update to go-tfe 0.3.23
cost estimation status polling
go-tfe dep update to 0.3.22
...
Previously we were using the experimental HCL 2 repository, but now we'll
shift over to the v2 import path within the main HCL repository as part of
actually releasing HCL 2.0 as stable.
This is a mechanical search/replace to the new import paths. It also
switches to the v2.0.0 release of HCL, which includes some new code that
Terraform didn't previously have but should not change any behavior that
matters for Terraform's purposes.
For the moment the experimental HCL2 repository is still an indirect
dependency via terraform-config-inspect, so it remains in our go.sum and
vendor directories for the moment. Because terraform-config-inspect uses
a much smaller subset of the HCL2 functionality, this does still manage
to prune the vendor directory a little. A subsequent release of
terraform-config-inspect should allow us to completely remove that old
repository in a future commit.
The cidrsubnets function signature is intentionally very low-level and
focused on the core requirement of generating addresses. This registry
module then wraps it with some additional functionality to make it more
convenient to generate and use subnet address ranges.
This is a companion to cidrsubnet that allows bulk-allocation of multiple
subnet addresses at once, with automatic numbering.
Unlike cidrsubnet, cidrsubnets allows each of the allocations to have a
different prefix length, and will pack the networks consecutively into the
given address space. cidrsubnets can potentially create more complicated
addressing schemes than cidrsubnet alone can, because it's able to take
into account the full set of requested prefix lengths rather than just
one at a time.
* command/import: properly use `-provider` supplied on the command line
The import command now attaches the provider configuration in the resource
instance, if set. That config is attached to the NodeAbstractResource
during the import graph building. This prevents errors when the implied
provider is not actually in the configuration at all, which may happen
when a configuration is using the `-beta` version of a provider (and
only that `-beta` version).
* command/import: fix variable reassignment and update docs
Fixes#22564
For a long time now we've been advising against the use of provisioners,
but our documentation for them is pretty prominent on the website in
comparision to the better alternatives, and so it's little surprise that
many users end up making significant use of them.
Although in the longer term a change to our information architecture would
probably address this even better, this is an attempt to be explicit about
the downsides of using provisioners and to prominently describe the
alternatives that are available for common use-cases, along with some
reasons why we consider them to be better.
I took the unusual step here of directly linking to specific provider
documentation pages about the alternatives, even though we normally try
to keep the core documentation provider-agnostic, because otherwise that
information tends to be rather buried in the provider documentation and
thus the reader would be reasonable to use provisioners just because we're
not giving specific enough alternative recommendations.
* website/formatdate: update example
The given example was showing HOUR:MONTH instead of HOUR:MINUTE
Fixes#22598
* website/import: remove reference to no-longer-working option
Users can no longer supply `-config=""` to tell Terraform not to load
configuration for import.
Fixes#22294
* website/provisioners: `host` is required in connection blocks
Fixes#21877
* website/variables: clarify variable definition precedence
It was not entirely obvious that a variable could not be assigned
multiples times in a single source.
Fixes#21682
* website/backend/local: add `workspace_dir` attribute
Fixes#21391
* website/output: `sensitive` outputs are redacted in output
Fixes#21502
* website/backends: sidebar order tweak
It makes sense for backend 'configuration' to appear before 'init'.
Fixes#13796
* Revert "website/formatdate: update example"
This reverts commit ccd93c86ddd15a21625c0767702ee1cc62e77254.
Reference: https://github.com/hashicorp/terraform/issues/16697
Enumerates a set of regular file names from a given glob pattern. Implemented via the Go stdlib `path/filepath.Glob()` functionality. Notably, stdlib does not support `**` or `{}` extended patterns. See also: https://github.com/golang/go/issues/11862
To support the extended glob patterns, it will require adding a dependency on a third party library or adding our own matching code.
The Terraform Enterprise brand has now been split into two parts:
- Terraform Cloud is the application that helps teams use Terraform together,
with remote state storage, a shared run environment, etc.
- Terraform Enterprise is the on-premise distribution that lets enterprises run
a private instance of the Terraform Cloud application.
The former TFE docs have been split accordingly.
- Make these descriptions more similar, since they do basically the same thing.
- Add some subheaders to break up the wall of text and make it more skimmable.
- Nudge people more firmly toward `for_each` if they need to actually
incorporate data from a variable into their instances.
- Add version note so you know whether you can use this yet.
These existing upstream cty functions allow matching strings against
regular expression patterns, which can be useful if you need to consume
a non-standard string format that Terraform doesn't (and can't) have a
built-in function for.
We added the csvdecode function originally with the intent of it being
used with for_each, but because csvdecode was released first we had a
section in its documentation warning about the downsides of using it with
"count", since that seemed like something people would be likely to try.
With resource "for_each" now merged, we can replace that scary section
with a more positive example of using these two features together.
We still include a paragraph noting that "count" _could_ be used here, but
with a caution against doing so. This is in the hope of helping users
understand the difference between these two patterns and why for_each is
the superior choice for most situations.
Ran into an issue requiring `-var='foo=bar'` over `-var 'foo=bar'`.
Without the `=` the generic response came back:
```
Usage: terraform plan [options] [DIR]
Generates an execution plan for Terraform.
This execution plan can be reviewed...
```
Terraform `0.12.2`
Team tokens never worked with the `atlas` backend, but the `remote` backend
uses them as intended; they can perform plans and applies on workspaces where
the associated team has at least plan or write permissions, respectively.
The search "terraform leading zero" does not find the `format()`
function, which is perfectly capable of adding leading zeros.
Thus I have added this one word to help people find `format()`.
The correct environment variable corresponding to the `ca_file` variable is `CONSUL_CACERT` and not `CONSUL_CAFILE`.
See `backend/remote-state/consul/backend.go` line 77.
This also includes a previously-missing test that verifies the behavior
described here, implemented as a planning context test for consistency
with how the other ignore_changes tests are handled.
* Correct fmt -check
With `-check=false` the exit status is always zero.
With `-check=true` the exit status is zero when all files are properly formatted and non-zero otherwise.
* update fmt documentation to use short form for -diff and -check
We previously had some notes about handling configuration variants just
tacked on to the "dependency inversion" section as an afterthought, but
this idea is a major use-case for dependency inversion so it deserves its
own section and a specific example.
There have been a few questions about this so far which indicated that the
previous docs for this feature were very lacking. This is an attempt to
describe more completely what "any" means, and in particular that it isn't
actually a type at all but rather a placeholder for a type to be selected
dynamically.
Based on some common questions and feedback since the v0.12.0 release,
here we add some small additional content to the documentation for
"dynamic" blocks, covering how to access the keys of the collection being
iterated over and how to fold multiple collections into a single one to
achieve the effect of a nested iteration.
These follow the same principle as jsondecode and jsonencode, but use
YAML instead of JSON.
YAML has a much more complex information model than JSON, so we can only
support a subset of it during decoding, but hopefully the subset supported
here is a useful one.
Because there are many different ways to _generate_ YAML, the yamlencode
function is forced to make some decisions, and those decisions are likely
to affect compatibility with other real-world YAML parsers. Although the
format here is intended to be generic and compatible, we may find that
there are problems with it that'll we'll want to adjust for in a future
release, so yamlencode is therefore marked as experimental for now until
the underlying library is ready to commit to ongoing byte-for-byte
compatibility in serialization.
The main use-case here is met by yamldecode, which will allow reading in
files written in YAML format by humans for use in Terraform modules, in
situations where a higher-level input format than direct Terraform
language declarations is helpful.
This is similar to the function of the same name in Python, generating a
sequence of numbers as a list that can then be used in other
sequence-oriented operations.
The primary use-case for it is to turn a count expressed as a number into
a list of that length, which can then be iterated over or passed to a
collection function to produce that number of something else, as shown
in the example at the end of its documentation page.
Using az login and then terraform init from the command line I got `Error: Either an Access Key / SAS Token or the Resource Group for the Storage Account must be specified`
We've seen in the past that some users try to use this form with the
ssh:// URL prefix, so we'll mention explicitly that this is invalid and
show a working example of how to use it without the URL scheme prefix.
- Note that we intentionally omitted it from the sidebar, to reduce confusion.
- Write a summary up top so you can stop reading sooner if you don't actually need this.
* lang/funcs: testing of functions through the lang package API
The function-specific unit tests do not cover the HCL conversion that happens when the functions are called in a terraform configuration. For e.g., HCL converts sets to lists before passing it to the function. This means that we could not test passing a set in the function _unit_ tests.
This adds a higher-level acceptance test, plus a check that every (pure) function has a test.
* website/docs: update function documentation
There was some leftover v0.11-style interpolation syntax here.
We prefer to use a "naked" expression in situations like this where the result
isn't a string, because interpolations returning non-strings is a common source
of confusion for new users.
* funcs/coalesce: return the first non-null, non-empty element from a
sequence.
The go-cty coalesce function, which was originally used here, returns the
first non-null element from a sequence. Terraform 0.11's coalesce,
however, returns the first non-empty string from a list of strings.
This new coalesce function aims to preserve terraform's documented
functionality while adding support for additional argument types. The
tests include those in go-cty and adapted tests from the 0.11 version of
coalesce.
* website/docs: update coalesce function document
The re-introduction of some of the ambiguity between argument and nested
block syntax (for compatibility with existing provider patterns)
unfortunately leads to some interesting consequences for attributes using
this mode.
While the behavior is generally as before in straightforward cases, this
page aims to spell out some of the different usage patterns explicitly
for the benefit of those writing more complex configurations, such as
generic re-usable modules where using argument vs. block syntax leads to
some real differences.
This page is intentionally not linked from anywhere in the part of the
website maintained in the Terraform repository. Instead, it can be linked
from the provider documentation for any argument where this pattern is
used, to help users understand the ways in which that argument might
deviate from the usual behaviors of arguments vs. nested blocks.
Some users are not accustomed to thinking of IP addresses in a bitwise
fashion, so the hope here is to give enough of an introduction to that way
of thinking for the reader to understand what the "newbits" and "netnum"
arguments represent.
The definition of split was referring the built-in function join. However, join is just one of the ways a string might have been created, and this could cause confusion.
In the terraform state documentation the verb "map" is widely used to
describe the relationship between an item in the state and in the real world
whereas the verb "attach" is not used anywhere.
For 0.11 I just specified the naming rules; for 0.12, I added some info about
referencing values and tightened up the layout of the optional arguments.
This commit also syncs up descriptions of `depends_on`.
'legacy' doesn't seem to be a thing anymore, and we were missing some
of the other values for -type. Also -no-color doesn't seem to be
relevant to this command.
1. The double "$" in the template confuses readers
2. As far as I can tell the variable "count" is not used either in this example
(it is in the next example though)
* docs: elaborate on supported remote backend versions
This PR adds a few lines to the docs to indicate which commands are
supported by what version of the remote backend and it makes a
recommendation about which version to use.
* Clarify remote state storage w/ TFE [skip ci]
Specifically, that this is the backend to use with remote state (all
tiers) and Free-Tier vs. Enterprise tiers differ in remote operations
* website: Arrange remote backend info differently
We have released the v0.12-oriented content to the website early in order
to support the beta process, but in some places we neglected to explicitly
mark features or content as being v0.12-only.
Here we add explicit markers to the main cases we've seen where readers
have reported confusion, along with some other tweaks in similar vein.
Terraform is way bigger than the core CLI tools and the language
now, and the docs have grown accordingly. So we're adding a
global index page to help users get around the many sections of the
docs site, and bumping the CLI/core docs down so they're no longer at the
top of the hierarchy.
The "right" (as in, conceptually pure) way to do this would be to actually
create a new level of directory hierarchy in between. But that would be real
expensive and annoying — the amount of 301s and links to edit would be
monumental, and it wouldn't gain us much beyond a certain picture-straightening
satisfaction, so I'm resisting the temptation.
As part of this, I'm copying the entire text of the 0.12
docs/configuration/modules.html page into docs/configuration-0-11/modules.html —
some of the 0.11 pages needed to be able to link to the moved content, I
didn't want to jump versions jarringly, and a close reading didn't reveal
anything in there that's inaccurate for 0.11.
The "terraform fmt" command produces a different canonical form than we
were showing in our examples here. Our examples should always reflect the
conventions applied by "terraform fmt" to avoid confusion.
(This particular decision is a pragmatic one because the formatter design
needs to use the same rules for the colon in the ? : conditional operator
as for the colon in "for" expressions.)
Since references to attributes of resources are by far the most common
reference type, and the mapping of resource type config to the attributes
is not always obvious, here we give some real examples of patterns for
accessing different configuration constructs within resource blocks along
with the resource type's exported attributes.
Since we don't have any real examples of labelled nested blocks yet (the
current SDK doesn't support them) I've included a hypothetical example for
now just to establish the patterns around them in preparation for
beginning to introduce them as we roll out this feature in the SDK.
As well as some general consolidation and reorganizing, this also includes
some updated advice for making the best use of new Terraform v0.12
features to create infrastructure building-blocks.
In particular, the "Module Usage" documentation is now consolidated into
the configuration section in order to bring all of our general language
documentation together, and the top-level "Modules" section is now
primarily focused on module _authors_ as an audience, covering topics such
as publishing modules and designing them for reuse.
* docs: update plan command documentation. Fixes#19235
* docs: added a missing reserved variable name. Fixes#19159.
* website: add note that resource names cannot start with a number
* website: add some notes to the 0.12 upgrade guide
The go-getter library that is used by the module loader validates S3 URLs in the parseURL function. That function assumes path-style URLs and fails on virtual-hosted-style URLs.
In 0.12, the outputs for a data source of terraform_remote_state are
nested under the 'outputs' attribute [1]. This updates the docs
to make this change clearer.
Worked with @radeksimko at Terraform hackday, who has submitted a
related upgrade guide [2]
[1] 1f4d2f4c50/builtin/providers/terraform/data_source_state.go (L16-L43)
[2] d8e00191b7
Because of the different possibilities for arranging the nav sidebars, we want
to make sure:
- IDs for the 0.11 and 0.12 language docs have a common prefix.
- That prefix is not the exact string `docs-config`.
Have I mentioned before that I really dislike this prefix matching behavior.
This is a non-working commit, because a bunch of links (including the sidebar
nav) are broken. Using a transition commit like this makes it easier to see the
changes necessary to get this content woven into the site.
In prior versions, we recommended using hash functions in conjunction with
the file function as an idiom for detecting changes to upstream blobs
without fetching and comparing the whole blob.
That approach relied on us being able to return raw binary data from
file(...). Since Terraform strings pass through intermediate
representations that are not binary-safe (e.g. the JSON state), there was
a risk of string corruption in prior versions which we have avoided for
0.12 by requiring that file(...) be used only with UTF-8 text files.
The specific case of returning a string and immediately passing it into
another function was not actually subject to that corruption risk, since
the HIL interpreter would just pass the string through verbatim, but this
is still now forbidden as a result of the stricter handling of file(...).
To avoid breaking these use-cases, here we introduce variants of the hash
functions a with "file" prefix that take a filename for a disk file to
hash rather than hashing the given string directly. The configuration
upgrade tool also now includes a rule to detect the documented idiom and
rewrite it into a single function call for one of these new functions.
This does cause a bit of function sprawl, but that seems preferable to
introducing more complex rules for when file(...) can and cannot read
binary files, making the behavior of these various functions easier to
understand in isolation.
It's not normally necessary to make explicit type conversions in Terraform
because the language implicitly converts as necessary, but explicit
conversions are useful in a few specialized cases:
- When defining output values for a reusable module, it may be desirable
to force a "cleaner" output type than would naturally arise from a
computation, such as forcing a string containing digits into a number.
- Our 0.12upgrade mechanism will use some of these to replace use of the
undocumented, hidden type conversion functions in HIL, and force
particular type interpretations in some tricky cases.
- We've found that type conversion functions can be useful as _temporary_
workarounds for bugs in Terraform and in providers where implicit type
conversion isn't working correctly or a type constraint isn't specified
precisely enough for the automatic conversion behavior.
These all follow the same convention of being named "to" followed by a
short type name. Since we've had a long-standing convention of running all
the words together in lowercase in function names, we stick to that here
even though some of these names are quite strange, because these should
be rarely-used functions anyway.
The sethaselement, setintersection, and setunion functions are defined in
the cty stdlib. Making them available in Terraform will make it easier to
work with sets, and complement the currently-Terraform-specific setproduct
function.
In the long run setproduct should probably move into the cty stdlib too,
but since it was submitted as a Terraform function originally we'll leave
it here now for simplicity's sake and reorganize later.
In our new world it produces either a set of a tuple type or a list of a
tuple type, depending on the given argument types.
The resulting collection's element tuple type is decided by the element
types of the given collections, allowing type information to propagate
even if unknown values are present.
This document was previously copied to the "Extending Terraform" section (in the
terraform-website repo), and the old URL was redirected so that the copy in
/guides can no longer be reached on the website. But the old copy of the file
remained, and now it runs the risk of confusing contributors, since the copy in
terraform-website/.../docs/extend is the more up-to-date version.
The AWS Go SDK automatically provides a default request retryer with exponential backoff that is invoked via setting `MaxRetries` or leaving it `nil` will default to 3. The terraform-aws-provider `config.Client()` sets `MaxRetries` to 0 unless explicitly configured above 0. Previously, we were not overriding this behavior by setting the configuration and therefore not invoking the default request retryer.
The default retryer already handles HTTP error codes above 500, including S3's InternalError response, so the extraneous handling can be removed. This will also start automatically retrying many additional cases, such as temporary networking issues or other retryable AWS service responses.
Changes:
* s3/backend: Add `max_retries` argument
* s3/backend: Enhance S3 NoSuchBucket error to include additional information
We missed this one on a previous pass of bringing in most of the cty
stdlib functions.
This will resolve#17625 by allowing conversion from Terraform's
conventional RFC 3339 timestamps into various other formats.
This function is similar to the template_file data source offered by the
template provider, but having it built in to the language makes it more
convenient to use, allowing templates to be rendered from files anywhere
an inline template would normally be allowed:
user_data = templatefile("${path.module}/userdata.tmpl", {
hostname = format("petserver%02d", count.index)
})
Unlike the template_file data source, this function allows values of any
type in its variables map, passing them through verbatim to the template.
Its tighter integration with Terraform also allows it to return better
error messages with source location information from the template itself.
The template_file data source was originally created to work around the
fact that HIL didn't have any support for map values at the time, and
even once map support was added it wasn't very usable. With HCL2
expressions, there's little reason left to use a data source to render
a template; the only remaining reason left to use template_file is to
render a template that is constructed dynamically during the Terraform
run, which is a very rare need.
This commit is a wide-ranging set of edits to the pages under
/docs/configuration. Among other things, it
- Separates style conventions out into their own page.
- Separates type constraints and conversion info into their own page.
- Conflates similar complex types a little more freely, since the distinction is
only relevant when restricting inputs for a reusable module or resource.
- Clarifies several concepts that confused me during edits.
James Bardin