Prebuilds need to be valid. Before this change, you can push a template
version that's preset will fail when making a prebuild. This PR ensures
all presets that are used for prebuilds are valid
Closes https://github.com/coder/internal/issues/1353
Does not solve the issue, but the error is currently opaque. This fails
the test when the init fails, hopefully raising up the error.
`--secure-auth-cookie` now automatically sources it's default value from `--access-url`
If the access url uses HTTPS, secure is set to `true`.
To revert to old behavior, set the value explicitly to `false`
If a deployment has 2 domains, overriding the oidc url allows the oidc
redirect to differ from the access_url
response to https://github.com/coder/coder/discussions/21500
**This config setting is hidden by default**
Archiving modules attempts to save as many modules as it can before it hits the limit. Enabling the template as much as it can, rather than a hard failure.
Only task workspaces have the checks in wsbuilder for violating the
managed agent caps in the license.
Stopped tasks that are resumed with a regular workspace start **still
count as usage**.
Also remove the "restart" button. The language is still there, and the
restart button is already on this page. That button being present had
users clicking it when waiting was the correct solution.
So the "Restart" option is not as pushed.
# What this does
Dynamic parameters caches the `./terraform/modules` directory for parameter usage. What this PR does is send over this archive to the provisioner when building workspaces.
This allow terraform to skip downloading modules from their registries, a step that takes seconds.
<img width="1223" height="429" alt="Screenshot From 2025-12-29 12-57-52" src="https://github.com/user-attachments/assets/16066e0a-ac79-4296-819d-924f4b0418dc" />
# Wire protocol
The wire protocol reuses the same mechanism used to download the modules `provisoner -> coder`. It splits up large archives into multiple protobuf messages so larger archives can be sent under the message size limit.
# 🚨 Behavior Change (Breaking Change) 🚨
**Before this PR** modules were downloaded on every workspace build. This means unpinned modules always fetched the latest version
**After this PR** modules are cached at template import time, and their versions are effectively pinned for all subsequent workspace builds.
**This is just the protobuf changes for the PR https://github.com/coder/coder/pull/21398**
Moved `UploadFileRequest` from `provisionerd.proto` -> `provisioner.proto`.
Renamed to `FileUpload` because it is now bi-directional.
This **is backwards compatible**. I tested it to confirm the payloads are identical. Types were just renamed and moved around.
```golang
func TestTypeUpgrade(t *testing.T) {
t.Parallel()
x := &proto2.UploadFileRequest{
Type: &proto2.UploadFileRequest_ChunkPiece{
ChunkPiece: &proto.ChunkPiece{
Data: []byte("Hello World!"),
FullDataHash: []byte("Foobar"),
PieceIndex: 42,
},
},
}
data, err := protobuf.Marshal(x)
require.NoError(t, err)
// Exactly the same output
// EhgKDEhlbGxvIFdvcmxkIRIGRm9vYmFyGCo= on `main`
// EhgKDEhlbGxvIFdvcmxkIRIGRm9vYmFyGCo= on this branch
fmt.Println(base64.StdEncoding.EncodeToString(data))
}
```
# What this does
This allows provisioner daemons to download files from `coderd`'s `files` table. This is used to send over cached module files and prevent the need of downloading these modules on each workspace build.
Delete builds were not deleting resources as the tf state being sent in the apply request was empty.
State removed from apply request and read from the session instead.
**Breaking Change:** Existing oauth apps might now use PKCE. If an
unknown IdP type was being used, and it does not support PKCE, it will
break.
To fix, set the PKCE methods on the external auth to `none`
```
export CODER_EXTERNAL_AUTH_1_PKCE_METHODS=none
```
Provisioner steps broken into smaller granular actions.
Changes:
- `ExtractArchive` moved to `init` request (was in `configure`)
- Writing `tfstate` moved to `plan` (was in `configure`)
- Moved most plan/apply outputs to `GraphComplete`
I noticed we have a defer to make sure all log output is captured by the
async log handling routine.
```golang
defer func() {
_ = outWriter.Close()
_ = errWriter.Close()
<-doneOut
<-doneErr
}()
```
But `e.timings.aggregate()` compiles the **current** parsed logs. I was
getting some race conditions in testing if the logs did not completely
parse before `aggregate` was called.
So now `aggregate` happens outside the cmd exec function, to make sure
it is completed before logs are accumulated.
Refactoring `ConvertState` is something we should eventually do. This PR
adds a golden file unit test for the output of `ConvertState` (even
errors).
That way if a refactor occurs, we can verify the output is unchanged for
our test cases.
Database insert errors will fail the transaction. So this error is
fatal. Properly return it for a better error call stack, and not just
hiding the error in the logs.
## Description
Adds a brief section to the API & Session Tokens documentation
explaining API key scopes.
## Changes
- Added "API Key Scopes" section to
`docs/admin/users/sessions-tokens.md`
- Includes overview of scope functionality and security benefits
- Documents scope format (`resource:action`) and wildcard usage
- Provides CLI examples for creating scoped tokens
- Lists common scope examples with descriptions
## Motivation
Users need documentation on how to create and use scoped API tokens for
improved security by limiting token permissions to only necessary
operations.
## Testing
- Reviewed documentation formatting
- Verified markdown structure
- Confirmed examples are accurate
For experimental and dogfood purposes, this adds the ability to opt in a single template.
Leaving the rest of the templates as is.
For GA, this setting might be removed or changed.
Prior to this, every workspace build ran `terraform init` in a fresh
directory. This would mean the `modules` are downloaded fresh. If the
module is not pinned, subsequent workspace builds would have different
modules.
Experiments passed to provisioners to determine behavior. This adds
`--experiments` flag to provisioner daemons. Prior to this, provisioners
had no method to turn on/off experiments.
Adds some extra meta data sent to provisioners. Also adds a field
`reuse_terraform_workspace` to tell the provisioner whether or not to
use the caching experiment.
Refactors all Terraform file path logic into a centralized tfpath package. This consolidates all path construction into a single, testable Layout type.
Instead of passing around `string` for directories, pass around the `Layout` which has the file location methods on it.
This uses the `terraform init -json` logs to add more visibility into
the `init` phase of `plan`. The `init` logs omit `ms` precision, so we
have to use `time.Now()` 😢
Open TF PR: https://github.com/hashicorp/terraform/pull/37818
This PR uses the same sha256 hashing technique as we use for APIKeys. So
now all randomly generated secrets will be hashed with sha256 for
consistency.
This is a breaking change for the oauth tokens. Since oauth is only
allowed for dev builds and experimental, this is ok.
The authz recorder is causing a lot of memory to be allocated, and is a
memory leak for websocket connections.
This change makes it opt-in on a per request basis (ontop of `isDev`).
To get the authz headers, use `Copy as cURL` on chrome and append the
header `x-authz-checks=true`.
Some comments are likely still omitted as guts does not cover all cases
right now. And some cases are omitted because the output format is
ambiguous on how best to format it.
Closes https://github.com/coder/internal/issues/716
This prevents a scan over the entire `workspace_build` table by removing
a `join`. This is still imperfect as we are still scanning over the
number of builds for the workspaces in the arguments. Ideally we would
have some index or something precomputed. Then we could skip scanning
over the builds for the correct workspaces that are not the latest.
External auth refresh errors lose the original error thrown on the first
refresh. This PR saves that error to the database to be raised on
subsequent refresh attempts
PProf labels segment the code into groups for determing the source of
cpu/memory profiles. Since the web server and background jobs share a
lot of the same code (eg wsbuilder), it helps to know if the load is
user induced, or background job based.
Can do `author:username` to filter templates created by a certain
author. Adding to help clean out some templates that I created on our
dev instance. This makes sorting a bit easier.
Steven Masley