These changes add support for injecting client tracing spans into HTTP
requests to the Pulumi API. The server can then rematerialize these span
references and attach its own spans for distributed tracing.
This change moves Git failures from glog.Warnings, which we don't
really pay attention, to true CLI warnings.
This will ensure we at least get bug reports in the event that this
fails on some user machine out in the wild. They are still non-fatal,
of course, since such a failure needn't prevent an update from happening.
This change captures the Git committer and author's login and email
addresses, so that we can display them prominently in the service. At
the moment, we only attribute updates to the identity that performed the
update which, in CI scenarios, is often always the same person for an
organization. This makes Pulumi look like a needlessly lonely place.
These changes enable tracing of Pulumi API calls.
The span with which to associate an API call is passed via a
`context.Context` parameter. This required plumbing a
`context.Context` parameter through a rather large number of APIs,
especially in the backend.
In general, all API calls are associated with a new root span that
exists for essentially the entire lifetime of an invocation of the
Pulumi CLI. There were a few places where the plumbing got a bit hairier
than I was willing to address with these changes; I've used
`context.Background()` in these instances. API calls that receive this
context will create new root spans, but will still be traced.
As it stands, we recompute the workspace for the current directory
potentially many times during some CLI operations, most notably
`stack ls`. These changes add a simple cache based on the complete path
passed to `NewFrom`, and eliminate some lagginess in `stack ls` when
there are multiple stacks.
Another option is to calculate the current workspace once in the CLI and
then fetch it as necessary.
* Initialize a new stack as part of `pulumi new`
* Prompt for values with defaults preselected
* Install dependencies
* Prompt for default config values
Our logic for converting npm style versions to PEP-440 style versions
was not correct in some cases. This change fixes this.
As part of this change we no longer produce a NPM version that would
be just X.Y.Z-dev, instead for development versions we always include
both the timestamp of the commit and the commit hash.
Instead of trying to use a bunch of sed logic to do our conversions,
we now have a small go program that uses a newly added library in
pkg/util. A side effect of this is that we can more easily write tests
to ensure the conversion works as expected.
Fixes#1243
If the file contains a zero byte within the first 8000 bytes (or the entire length of the file if shorter), treat the file as binary and do not do any textual transformations. This is the same approach git uses to determine if a file is binary.
This changes the CLI interface in a few ways:
* `pulumi preview` is back! The alternative of saying
`pulumi update --preview` just felt awkward, and it's a common
operation to want to perform. Let's just make it work.
* There are two flags consistent across all update commands,
`update`, `refresh`, and `destroy`:
- `--skip-preview` will skip the preview step. Note that this
does *not* skip the prompt to confirm that you'd like to proceed.
Indeed, it will still prompt, with a little warning text about
the fact that the preview has been skipped.
* `--yes` will auto-approve the updates.
This lands us in a simpler and more intuitive spot for common scenarios.
I found the flag --force to be a strange name for skipping a preview,
since that name is usually reserved for operations that might be harmful
and yet you're coercing a tool to do it anyway, knowing there's a chance
you're going to shoot yourself in the foot.
I also found that what I almost always want in the situation where
--force was being used is to actually just run a preview and have the
confirmation auto-accepted. Going straight to --force isn't the right
thing in a CI scenario, where you actually want to run a preview first,
just to ensure there aren't any issues, before doing the update.
In a sense, there are four options here:
1. Run a preview, ask for confirmation, then do an update (the default).
2. Run a preview, auto-accept, and then do an update (the CI scenario).
3. Just run a preview with neither a confirmation nor an update (dry run).
4. Just do an update, without performing a preview beforehand (rare).
This change enables all four workflows in our CLI.
Rather than have an explosion of flags, we have a single flag,
--preview, which can specify the mode that we're operating in. The
following are the values which correlate to the above four modes:
1. "": default (no --preview specified)
2. "auto": auto-accept preview confirmation
3. "only": only run a preview, don't confirm or update
4. "skip": skip the preview altogether
As part of this change, I redid a bit of how the preview modes
were specified. Rather than booleans, which had some illegal
combinations, this change introduces a new enum type. Furthermore,
because the engine is wholly ignorant of these flags -- and only the
backend understands them -- it was confusing to me that
engine.UpdateOptions stored this flag, especially given that all
interesting engine options _also_ accepted a dryRun boolean. As of
this change, the backend.PreviewBehavior controls the preview options.
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.
The `go-git` implementation of `git status` is outrageously expensive,
as it performs a hash-based comparision of the working tree against the
committed state with no caching. In some example runs, this takes
upwards of 15 seconds. Because this is on the startup path for updates,
this results in a rather poor user experience.
These changes replace the `go-git` implementation with a call to `git
status --porcelain -z`, which only writes data to stdout if the working
tree is dirty.
Note that these changes also make all git-related update metadata
best-effort.
These changes add a new flag to the CLI, `--profiling`, that enables
CPU and heap profiling as well as execution tracing of the CLI itself.
The argument to this flag serves as a prefix for the profile outputs;
the CPU and heap profiles and execution trace are written to
`[filename].[pid].{cpu,mem,trace}`, respectively.
These changes also fix an issue with `cmdutil.RunFunc` wherein any error
would prevent the command's post-run hooks from executing.
The service has no good way to preview a destroy operation for a stack
managed by a PPC. Until it does, just behave as if --force was passed
to the CLI in this case (i.e. skip the preview).
Fixes#1301
Instead of using a shell script to jump from the language host into
node, just invoke node directly. This makes our start-up path a little
simpler to understand and indirectly fixespulumi/home#156, where we
would fail on Windows if the `-exec` script was in a folder that had
spaces in it (due to a subtle interaction between how go launches cmd
files and how cmd.exe parses arguments).
When linking back to the service, use the newer, simpler, URLs for a
stack: `https://pulumi.com/<owner>/<stack-name>` instead of
`https://pulumi.com/<owner>/-/-/<stack-name>`
Rather than filtering out the `id` and `urn` properties when serializing
the inputs to an invoke, pass these properties along. This enables the
use of invoke endpoints that accepts these as inputs (e.g. the endpoint
that backs `aws.ec2.getSubnet`).
Refactoring to support "preview" and "update" as part of the same
operation interacted poorly with deploying into a PPC via the service.
Per Pat, this is the quickest fix that gets us off the floor.
Part of #1301