This endpoint is relatively expensive, as it returns a large amount of
data for each stack (essentially the stack's entire checkpoint). We
currently call this endpoint in a few places that really just need
information about a single stack. In these places we should simply
interact with that stack directly.
After these changes, there is only one call to the `stacks` endpoint on
the startup path: `upgradeConfigurationFiles` loops over all of the
stacks in each backend and attempts to upgrade each stack's
configuration. Once we no longer need to do this we will not be hitting
the `stacks` endpoint on each CLI invocation.
This change includes a bunch of refactorings I made in prep for
doing refresh (first, the command, see pulumi/pulumi#1081):
* The primary change is to change the way the engine's core update
functionality works with respect to deploy.Source. This is the
way we can plug in new sources of resource information during
planning (and, soon, diffing). The way I intend to model refresh
is by having a new kind of source, deploy.RefreshSource, which
will let us do virtually everything about an update/diff the same
way with refreshes, which avoid otherwise duplicative effort.
This includes changing the planOptions (nee deployOptions) to
take a new SourceFunc callback, which is responsible for creating
a source specific to the kind of plan being requested.
Preview, Update, and Destroy now are primarily differentiated by
the kind of deploy.Source that they return, rather than sprinkling
things like `if Destroying` throughout. This tidies up some logic
and, more importantly, gives us precisely the refresh hook we need.
* Originally, we used the deploy.NullSource for Destroy operations.
This simply returns nothing, which is how Destroy works. For some
reason, we were no longer doing this, and instead had some
`if Destroying` cases sprinkled throughout the deploy.EvalSource.
I think this is a vestige of some old way we did configuration, at
least judging by a comment, which is apparently no longer relevant.
* Move diff and diff-printing logic within the engine into its own
pkg/engine/diff.go file, to prepare for upcoming work.
* I keep noticing benign diffs anytime I regenerate protobufs. I
suspect this is because we're also on different versions. I changed
generate.sh to also dump the version into grpc_version.txt. At
least we can understand where the diffs are coming from, decide
whether to take them (i.e., a newer version), and ensure that as
a team we are monotonically increasing, and not going backwards.
* I also tidied up some tiny things I noticed while in there, like
comments, incorrect types, lint suppressions, and so on.
We previously locked our dependency on google.golang.org/grpc to 1.7.2 due to issues we had seen on 1.8.x as noted in #701. However, this has prevented us using some other dependencies which require newer grpc. A test in this repo and AWS showed no problems with the latest 1.10.x versions of the library in our tests.
We'll go ahead and remove this constraint and allow grpc to float forward. If we see issues again, we'll use that repro case to investigate an alternative fix in our code.
Resolves#701.
These changes add the API types and cloud backend code necessary to
interact with service-managed stacks (i.e. stacks that do not have
PPC-managed deployments). The bulk of these changes are unremarkable:
the API types are straightforward, as are most of the interactions with
the new APIs. The trickiest bits are token and log management.
During an update to a managed stack, the CLI must continually renew the
token used to authorize the operations on that stack that comprise the
update. Once a token has been renewed, the old token should be
discarded. The CLI supports this by running a goroutine that is
responsible for both periodically renewing the token for an update and
servicing requests for the token itself from the rest of the backend.
In addition to token renewal, log output must be captured and uploaded
to the service during an update to a managed stack. Implementing this in
a reasonable fashion required a bit of refactoring in order to reuse
what already exists for the local backend. Each event-specific `Display`
function was replaced with an equivalent `Render` function that returns
a string rather than writing to a stream. This approach was chosen
primarily to avoid dealing with sheared colorization tags, which would
otherwise require clients to fuse log lines before colorizing. We could
take that approach in the future.
These changes refactor direct interactions with the Pulumi API out of
the cloud backend and into a subpackage, `pkg/backend/cloud/client`.
This package exposes a slightly higher-level API that takes care of
calculating paths, performing HTTP calls, and occasionally wrapping
multiple physical calls into a single logical call (notably the creation
of an update and the upload of its program).
This is primarily intended as preparation for some of the changes
suggested in the feedback for #1067.
We needed to have two types of `UpdateProgramRequest` to serve both newer shapes of Pulumi configuration values as well as the older, untyped version. (The original change was supporting the `ConfigValue` type, which had a flag to indiciate if the configuration value was encrypted or not.)
Now that LM has been migrated to the M10 bits https://github.com/pulumi/home/issues/168 , we can remove this type. (The PR to remove existing references in the `pulumi-service` repo is https://github.com/pulumi/pulumi-service/pull/957.)
Just a nit: It's possible (though, unlikely) that the repo file is
deleted between the call to `os.Stat` and `ioutil.ReadFile`. Instead,
just try to read the file -- if the file doesn't exist,
`ioutil.ReadFile` will return an error that works the same with
`os.IsNotExist(err)` as the error returned from `os.Stat`.
Note: This is a minor issue that I didn't get to for M11 that isn't
required for M11 and would be fine merging for post-M11.
When you specify a template name explicitly (e.g.
`pulumi new typescript`), we'll try to download the template tarball
without first downloading the JSON list of available templates. The JSON
includes a description used when replacing the `${DESCRIPTION}` string
in template files. Since we didn't download the JSON, we won't have a
description, so we fallback to a default value (`"A Pulumi project."`).
This also happens when specifying `--offline` to use an existing
template under `~/.pulumi/templates`; we won't have a description for
the template, so we fallback to a default description. The fallback
value happens to be the same as the description for each of our current
templates, so noone will currently notice an issue.
For M11, I included initial support for a template manifest file where
the description (and any future metadata) could be stored, but didn't go
as far as actually reading the file.
This change makes it so the CLI actually reads the description from the
manifest file (if it exists), otherwise falling back to the default
value as is done currently. Some minor related cleanup is included in
this change.
This change uses the prior checkpoint's deployment manifest to pre-
populate all plugins required to complete the destroy operation. This
allows for subsequent attempts to load a resource's plugin to match the
already-loaded version. This approach obviously doesn't work in a
hypothetical future world where plugins for the same resource provider
are loaded side-by-side, but we already know that.
Also, rename/cleanup a bunch of serialization code.
Also, generate better environment names in the serialized closure code. Thsi code should be much easier to make sense of as hte names will better track to the original names in the user code.
Also, dedupe simple non-capturing functions. This helps ensure we don't spit out N copies of __awaiter (one per file it is declared in).
This takes the existing `apitype.Checkpoint` type and renames it to
`apitype.CheckpointV1` locking in the shape. In addition, we introduce
a `apitype.VersionedCheckpoint` type, which holds a version number and
a json document representing a checkpoint at that version. Now, when
reading a checkpoint, the CLI can determine if it's in a format it
understands, and fail gracefully if it is not.
While the CLI understands the older checkpoint version, it always
writes the newest version format, meaning that if you manage a
fire-and-forget stack with this version of the CLI, it will be
un-readable by previous versions.
Stacks managed by Pulumi.com are not impacted by this change.
Fixes: #887
* Improve error messages output by the CLI
This fixes a couple known issues with the way that we present errors
from the Pulumi CLI:
1. Any errors from RPC endpoints were bubbling up as they were to
the top-level, which was unfortunate because they contained
RPC-specific noise that we don't want to present to the user. This
commit unwraps errors from resource providers.
2. The "catastrophic error" message often got printed twice
3. Fatal errors are often printed twice, because our CLI top-level
prints out the fatal error that it receives before exiting. A lot of
the time this error has already been printed.
4. Errors were prefixed by PU####.
* Feedback: Omit the 'catastrophic' error message and use a less verbose error message as the final error
* Code review feedback: interpretRPCError -> resourceStateAndError
* Code review feedback: deleting some commented-out code, error capitalization
* Cleanup after rebase
This adds a `pulumi new` command which makes it easy to quickly
automatically create the handful of needed files to get started building
an empty Pulumi project.
Usage:
```
$ pulumi new typescript
```
Or you can leave off the template name, and it will ask you to choose
one:
```
$ pulumi new
Please choose a template:
> javascript
python
typescript
```
We were not colorizing a summary in one case. Also, there were
codepaths that would print to the console without first calling
spinner.Reset(), so the spinner would not be cleared from the screen.
When a stack has secrets, we now take the secret values and construct
a regular expression which is just an alternation of all the secret
values. Then, before pushing any string data into an Event, we run the
regular expression and replace all matches with '[secret]'.
Fixes#747
The engine now emits events with richer metadata during the
ResourceOutputs and ResourcePre callbacks. The CLI can then use this
information to decide if it should display the event or not and how
much of the event to display.
Options dealing with what to display and how to display it have moved
into the CLI and the engine now emits all information for each event.
I believe because of the way we have structured the code, it is
impossible to know a resource's parent but not printed it. I've
changed the test which would print the parent resource to an assert
that ensure we have printed it.
The next commit is going to remove the shown array because we no
longer need it, but this commit is here so that if there are display
bugs as part of the larger refactoring in how we display events, we
can bisect back and see this failure.
The `shouldShow` method always marked a step as seen, and having the
side effect there is a little confusing. Because we call shouldShow in
the StepPre, StepPost and Output handlers, its also hard to ensure an
invarant I think we want, which is that in the Post and Output
handlers, we've already seen the event.
So, let's move the marking out of `shouldShow` and into
`OnResourceStepPre` and then assert we've already seen it in
`OnResourceStepPre` and `OnResourceOutputs` handlers.
This means that shouldShow is now a pure function and makes it easier
to move the decision on if we should print information about a step
out of the engine and into the CLI.
As it stands, we allow plugin load requests to race. Not only does this
create a situation in which we may load and then immediately throw away
a plugin (potentially leaking its process), it also creates the
possibility of races when reading from/writing to the various plugin
caches. These changes serialize all plugin loads and cache accesses by
running all accesses for a particular host in a single goroutine.
Fixes#1020.
The engine now unconditionally emits a new type of event, a
PreludeEvent, which contains the configuration for a stack as well as
an indication if the stack is being previewed or updated. The
responsibility for interpreting the --show-config flag on the command
line is now handled by the CLI, which uses this to decide if it should
print the configuration or not, and then writes the "Previewing
changes" or "Deploying chanages" header.
This value was unused across all of our display code. We did thread it
everywhere, but we never actually used the value to make any
decisions. Since we want to move to a model where the engine does not
decide *what* to display, it's helpful to remove this policy stuff
anyway.
This helper method is only really used for testing, but we should not
allow it to create a Key who's namespace has a colon (as ParseKey
would not build something like this).
This API was introduced to aid the refactoring, but it isn't something
we want to support long term. Remove it and for a few places, push
passing config.Key around more, instead of converting to the old type
eagerly.
When serializing config.Key's we now write them as <package>:<name>
instead of <package>:config:<name>. We continue to support reading the
older format for compatability with older files.
config.Key has become a pair of namespace and name. Because the whole
world has not changed yet, there continues to be a way to convert
between a tokens.ModuleMember and config.Key, however now sometime the
conversion from tokens.ModuleMember can fail (when the module member
is not of the form `<package>:config:<name>`).
I'll be changing the structure of the representation of config.Key, so
let's write some tests first to ensure we can continue to treat
everything as JSON and YAML.
Right now, config.Key is a type alias for tokens.ModuleMember. I did a
pass over the codebase such that we use config.Key everywhere it
looked like the value did not leak to some external process (e.g a
resource provider or a langhost).
Doing this makes it a little clearer (hopefully) where code is
depending on a module member structure (e.g. <package>:config:<value>)
instead of just an opaque type.
