In pulumi/pulumi#1356, we observed that we can fail during a destroy
because we attempt to load the language plugin, which now eagerly looks
for the @pulumi/pulumi package.
This is also blocking ingestion of the latest engine bits into the PPC.
It turns out that for destroy (and refresh), we have no need for the
language plugin. So, let's skip loading it when appropriate.
This change suppresses the warning
warning: resource plugin aws is expected to have version >=0.11.3,
but has 0.11.1-dev-1523506162-g06ec765; the wrong version may
be on your path, or this may be a bug in the plugin
when the PULUMI_DEV envvar is set to a truthy value.
This warning keeps popping up in demos since I'm always using dev builds
and I'd like a way to shut it off, even though this can legitimately
point out a problem. Eventually I'll switch to official buildsa but,
until then, it seems worth having a simple way to suppress.
As of this change, the engine will run all Configure calls in parallel.
This improves startup performance, since otherwise, we would block
waiting for all plugins to be configured before proceeding to run a
program. Emperically, this is about 1.5-2s for AWS programs, and
manifests as a delay between the purple "Previewing update of stack"
being printed, and the corresponding grey "Previewing update" message.
This is done simply by using a Goroutine for Configure, and making sure
to synchronize on all actual CRUD operations. I toyed with using double
checked locking to eliminate lock acquisitions -- something we may want
to consider as we add more fine-grained parallelism -- however, I've
kept it simple to avoid all the otherwise implied memory model woes.
I made the judgment call that GetPluginInfo may proceed before
Configure has settled. (Otherwise, we'd immediately call it and block
after loading the plugin, obviating the parallelism benefits.) I also
made the judgment call to do this in the engine, after flip flopping
several times about possibly making it a provider's own decision.
* Refactor the SnapshotManager interface
Lift snapshot management out of the engine by delegating it to the
SnapshotManager implementation in pkg/backend.
* Add a event interface for plugin loads and use that interface to record plugins in the snapshot
* Remove dead code
* Add comments to Events
* Add a number of tests for SnapshotManager
* CR feedback: use a successful bit on 'End' instead of having a separate 'Abort' API
* CR feedback
* CR feedback: register plugins one-at-a-time instead of the entire state at once
The RPC provider interface needs a way to convey back to the engine
that a resource being read no longer exists. To do this, we'll return
the ID property that was read back. If it is empty, it means the
resource is gone. If it is non-empty, we expect it to match the input.
This change lets plugin versions to float in two ways:
1) If a `pulumi plugin install` detects a newer version is available
already, there's no need to download and install the older version.
2) If the engine attempts to load a plugin at a particular version,
if a newer version is available, it will be accepted without error.
As part of this, we permit $PATH to have the final say when determining
which version to accept. That is, it can always override the choice.
Note that I highly suspect, in the limit, that we'll want to stop doing
this for major version incompatibilities. For now, since we don't
envision any such version changes imminently, this will suffice.
This commit changes two things about our resource model:
* Stop performing Pulumi Engine-side diffing of resource state.
Instead, we defer to the resource plugins themselves to determine
whether a change was made and, if so, the extent of it. This
manifests as a simple change to the Diff function; it is done in
a backwards compatible way so that we continue with legacy diffing
for existing resource provider plugins.
* Add a Read RPC method for resource providers. It simply takes a
resource's ID and URN, plus an optional bag of further qualifying
state, and it returns the current property state as read back from
the actual live environment. Note that the optional bag of state
must at least include enough additional properties for resources
wherein the ID is insufficient for the provider to perform a lookup.
It may, however, include the full bag of prior state, for instance
in the case of a refresh operation.
This is part of pulumi/pulumi#1108.
* Improve the error message arising from missing required configs for
resource providers
If the resource provider that we are speaking to is new enough, it will send
across a list of keys and their descriptions alongside an error
indicating that the provider we are configuring is missing required
config. This commit packages up the list of missing keys into an error
that can be presented nicely to the user.
* Code review feedback: renaming simplification and correcting errors in comments
* Send structured errors across RPC boundaries
This brings us closer to gRPC best practices where we send structured
errors with error codes across RPC endpoints. The new "rpcerrors"
package can wrap errors from RPC endpoints, so RPC servers can attach
some additional context as to why a request failed.
* Code review feedback:
1. Rename rpcerrors -> rpcerror, better package name
2. Rename RPCError -> Error, RPCErrorCause -> ErrorCause, names
suggested by gometalinter to improve their package-qualified names
3. Fix import organization in rpcerror.go
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.
* 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
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.
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.
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>`).
As it stands, we only configure those providers for which configuration
is present. This can lead to surprising failure modes if those providers
are then used to create resources. These changes ensure that all
resource providers that are not configured during plan initialization
are configured upon first load.
Fixes#758.
Previously, we would prefer a plugin on the $PATH which is more or
less always the case for people hacking on `pulumi`. Later, when we
went to check the loaded plugin version matched the one we requested,
we fail.
Now, if we have a version, we'll first consult the local plugin
cache. If that fails, we'll fall back to the $PATH as we used to.
When we are loading a plugin without a version, we continue to use the
one on the $PATH (without testing the cache) on the assumption it is
newer.
In addition, we've turned the "plugin versions are mis-matched" from
an error into a warning. We expect that we'll only ever see this
warning when something strange is going on (since in the normal case,
we'll have found the exact version in the cache) but having it not
hard fail does help in development cases.
Fixes#977
This change makes the engine backwards compatible with older
language host binaries, by simply ignoring GetRequiredPlugins
calls when the RPC server has not yet implemented it. This
is benign, since we will eventually fault plugins in on demand,
although it does mean that commands like `pulumi plugin install`
will become no-ops (which, thankfully, is what we want).
Previously, the checkpoint manifest contained the full path to a plugin
binary, in places of its friendly name. Now that we must move to a model
where we install plugins in the PPC based on the manifest contents, we
actually need to store the name, in addition to the version (which is
already there). We still also capture the path for debugging purposes.
We have had a long-standing bug in here where we waiting on a
stdout channel that never got populated, when the language plugin
fails to load entirely. This would lead to hung processes. The
fix is simple: only wait for stdout/stderr channels to drain that
have actually been wired up to enjoy the requisite signaling.
This adds support for two things:
* Installing all plugins that a project requires with a single command:
$ pulumi plugin install
* Listing the plugins that this project requires:
$ pulumi plugin ls --project
$ pulumi plugin ls -p
This brings back the Node.js language plugin's GetRequiredPlugins
function, reimplemented in Go now that the language host has been
rewritten from JavaScript. Fairly rote translation, along with
some random fixes required to get tests passing again.
This change adds a GetRequiredPlugins RPC method to the language
host, enabling us to query it for its list of plugin requirements.
This is language-specific because it requires looking at the set
of dependencies (e.g., package.json files).
It also adds a call up front during any update/preview operation
to compute the set of plugins and require that they are present.
These plugins are populated in the cache and will be used for all
subsequent plugin-related operations during the engine's activity.
We now cache the language plugins, so that we may load them
eagerly too, which we never did previously due to the fact that
we needed to pass the monitor address at load time. This was a
bit bizarre anyhow, since it's really the Run RPC function that
needs this information. So, to enable caching and eager loading
-- which we need in order to invoke GetRequiredPlugins -- the
"phone home" monitor RPC address is passed at Run time.
In a subsequent change, we will switch to faulting in the plugins
that are missing -- rather than erroring -- in addition to
supporting the `pulumi plugin install` CLI command.
This change introduces a workspace.GetPluginPath function that probes
the central workspace cache of plugins for a matching plugin binary that
matches the desired kind, name, and, optionally, version. It also permits
overriding this with $PATH for developer scenarios.
The analyzer, language, and resource plugin logic now uses this function
for deciding which binary path to load at runtime.
My previous change to stop supplying unknown properties to providers
broke `pulumi preview` in the case of unknown inputs. This change
restores the previous behavior for previews only; the new unknown-free
behavior remains for applies.
Fixes#790.
Before these changes, we were inconsistent in our treatment of unknown
property values across the resource provider RPC interface. `Check` and
`Diff` were retaining unknown properties in inputs and outputs;
`Create`, `Update`, and `Delete` were not. This interacted badly with
recent changes to `Check` to return all provider inputs--i.e. not just
defaults--from that method: if an unknown input was provided, it would
be present in the returned inputs, which would eventually confuse the
differ by giving the appearance of changes where none were present.
These changes remove unknowns from the provider interface entirely:
unknown property values are never passed to a provider, and a provider
must never return an unknown property value.
This is the primary piece of the fix for pulumi/pulumi-terraform#93.
This change adds a bit more tracing context to RPC marshaling
logging so that it's easier to attribute certain marshaling calls.
Prior to this, we'd just have a flat list of "marshaled property X"
without any information about what the marshaling pertained to.
This change adds rudimentary delete-before-create support (see
pulumi/pulumi#450). This cannot possibly be complete until we also
implement pulumi/pulumi#624, becuase we may try to delete a resource
while it still has dependent resources (which almost certainly will
fail). But until then, we can use this to manually unwedge ourselves
for leaf-node resources that do not support old and new resources
living side-by-side.
This change just flows the project's "main" directory all the way
through to the plugins, fixing #667. In that work item, we discussed
alternative approaches, such as rewriting the asset paths, but this
is tricky because it's very tough to do without those absolute paths
somehow ending up in the checkpoint files. Just launching the
processes with the right pwd is far easier and safer, and it turns
out that, conveniently, we set up the plugin context in exactly the
same place that we read the project information.
As documented in issue #616, the inputs/defaults/outputs model we have
today has fundamental problems. The crux of the issue is that our
current design requires that defaults present in the old state of a
resource are applied to the new inputs for that resource.
Unfortunately, it is not possible for the engine to decide which
defaults remain applicable and which do not; only the provider has that
knowledge.
These changes take a more tactical approach to resolving this issue than
that originally proposed in #616 that avoids breaking compatibility with
existing checkpoints. Rather than treating the Pulumi inputs as the
provider input properties for a resource, these inputs are first
translated by `Check`. In order to accommodate provider defaults that
were chosen for the old resource but should not change for the new,
`Check` now takes the old provider inputs as well as the new Pulumi
inputs. Rather than the Pulumi inputs and provider defaults, the
provider inputs returned by `Check` are recorded in the checkpoint file.
Put simply, these changes remove defaults as a first-class concept
(except inasmuch as is required to retain the ability to read old
checkpoint files) and move the responsibilty for manging and
merging defaults into the provider that supplies them.
Fixes#616.
This change adds a new manifest section to the checkpoint files.
The existing time moves into it, and we add to it the version of
the Pulumi CLI that created it, along with the names, types, and
versions of all plugins used to generate the file. There is a
magic cookie that we also use during verification.
This is to help keep us sane when debugging problems "in the wild,"
and I'm sure we will add more to it over time (checksum, etc).
For example, after an up, you can now see this in `pulumi stack`:
```
Current stack is demo:
Last updated at 2017-12-01 13:48:49.815740523 -0800 PST
Pulumi version v0.8.3-79-g1ab99ad
Plugin pulumi-provider-aws [resource] version v0.8.3-22-g4363e77
Plugin pulumi-langhost-nodejs [language] version v0.8.3-79-g77bb6b6
Checkpoint file is /Users/joeduffy/dev/code/src/github.com/pulumi/pulumi-aws/.pulumi/stacks/webserver/demo.json
```
This addresses pulumi/pulumi#628.
The first exception relates to how we launch plugins. Plugin paths are
calculated using a well-known set of rules; this makes `gas` suspicious
due to the need to use a variable to store the path of the plugin.
The second and third are in test code and aren't terribly concerning.
The latter exception asks `gas` to ignore the access key we hard-code
into the integration tests for our Pulumi test account.
The fourth exception allows use to use more permissive permissions for
the `.pulumi` directory than `gas` would prefer. We use `755`; `gas`
wants `700` or stricter. `755` is the default for `mkdir` and `.git` and
so seems like a reasonable choice for us.
This change adds back component output properties. Doing so
requires splitting the RPC interface for creating resources in
half, with an initial RegisterResource which contains all of the
input properties, and a final CompleteResource which optionally
contains any output properties synthesized by the component.
Because the Pulumi.yaml file demarcates the boundary used when
uploading a program to the Pulumi.com service at the moment, we
have trouble when a Pulumi program uses "up and over" references.
For instance, our customer wants to build a Dockerfile located
in some relative path, such as `../../elsewhere/`.
To support this, we will allow the Pulumi.yaml file to live
somewhere other than the main Pulumi entrypoint. For example,
it can live at the root of the repo, while the Pulumi program
lives in, say, `infra/`:
Pulumi.yaml:
name: as-before
main: infra/
This fixespulumi/pulumi#575. Further work can be done here to
provide even more flexibility; see pulumi/pulumi#574.
These changes introduce a new field, `Raw`, to `diag.Message`. This
field indicates that the contents of the message are not a format string
and should not be rendered via `Sprintf` during stringification.
The plugin std{out,err} readers have been updated to use raw messages,
and the event reader in `pulumi` has been fixed s.t. it does not format
event payloads before display.
Fixes#551.
Note: for the purposes of this discussion, archives will be treated as
assets, as their differences are not particularly meaningful.
Currently, the identity of an asset is derived from the hash and the
location of its contents (i.e. two assets are equal iff their contents
have the same hash and the same path/URI/inline value). This means that
changing the source of an asset will cause the engine to detect a
difference in the asset even if the source's contents are identical. At
best, this leads to inefficiencies such as unnecessary updates. This
commit changes asset identity so that it is derived solely from an
asset's hash. The source of an asset's contents is no longer part of
the asset's identity, and need only be provided if the contents
themselves may need to be available (e.g. if a hash does not yet exist
for the asset or if the asset's contents might be needed for an update).
This commit also changes the way old assets are exposed to providers.
Currently, an old asset is exposed as both its hash and its contents.
This allows providers to take a dependency on the contents of an old
asset being available, even though this is not an invariant of the
system. These changes remove the contents of old assets from their
serialized form when they are passed to providers, eliminating the
ability of a provider to take such a dependency. In combination with the
changes to asset identity, this allows a provider to detect changes to
an asset simply by comparing its old and new hashes.
This is half of the fix for [pulumi/pulumi-cloud#158]. The other half
involves changes in [pulumi/pulumi-terraform].
Adds OpenTracing in the Pulumi engine and plugin + langhost subprocesses.
We currently create a single root span for any `Enging.plan` operation - which is a single `preview`, `update`, `destroy`, etc.
The only sub-spans we currently create are at gRPC boundaries, both on the client and server sides and on both the langhost and provider plugin interfaces.
We could extend this to include spans for any other semantically meaningful sections of compute inside the engine, though initial examples show we get pretty good granularity of coverage by focusing on the gRPC boundaries.
In the future, this should be easily extensible to HTTP boundaries and to track other bulky I/O like datastore read/writes once we hook up to the PPC and Pulumi Cloud.
We expose a `--trace <endpoint>` option to enable tracing on the CLI, which we will aim to thread through to subprocesses.
We currently support sending tracing data to a Zipkin-compatible endpoint. This has been validated with both Zipkin and Jaeger UIs.
We do not yet have any tracing inside the TypeScript side of the JS langhost RPC interface. There is not yet automatic gRPC OpenTracing instrumentation (though it looks like it's in progress now) - so we would need to manually create meaningful spans on that side of the interface.
It's legal and possible for undefined properties to show up in
objects, since that's an idiomatic JavaScript way of initializing
missing properties. Instead of failing for these during deployment,
we should simply skip marshaling them to Terraform and let it do
its thing as usual. This came up during our customer workload.
