## Changes
This PR now allows you to define variables in the bundle config and set
them in three ways
1. command line args
2. process environment variable
3. in the bundle config itself
## Tests
manually, unit, and black box tests
---------
Co-authored-by: Miles Yucht <miles@databricks.com>
## Changes
Improved error message when 'bricks bundle run' is executed before
'bricks bundle deploy'
The error happens when we attempt to load terraform state when it does
not exist.
The best way to check if terraform state actually exists is to call
`terraform show -json` and that's what already happens here
https://github.com/databricks/bricks/compare/main...error-before-deploy#diff-8c50f8c04e568397bc865b7e02d1f4ec5b18379d8d32daddfeb041035d804f5fL28
Absence of `state.Values` indicates that there is no state and likely
bundle was just never deployed.
## Tests
Ran `bricks bundle run test_job` on a new non-deployed bundle.
**Output:**
`Error: terraform show: No state. Did you forget to run 'bricks bundle
deploy'?`
Running `bricks bundle deploy && bricks bundle run test_job` succeeds.
---------
Co-authored-by: Pieter Noordhuis <pieter.noordhuis@databricks.com>
## Changes
Add omit empty tag to git details. Otherwise this field becomes a
required field in the config json schema
## Tests
Tested by regenerating the json schema and checking that the git field
is now optional
## Changes
This config block contains commit, branch and remote_url which will be
automatically loaded if specified in the repo, and can also be specified
by the user
## Tests
Unit and black-box tests
## Changes
Traverses the variables referred in a depth first manner to resolve
string fields.
Errors out if a cycle is detected
## Tests
Manually and unit/blackbox tests
## Changes
Adds a IsInplaceSupported() function to the event interface. Any event
that now uses the progress logger has to declare whether they support in
place logging
## Tests
Manually
## Changes
This PR adds checks during bundle config load and merge to error out if
there are duplicate keys for resource definitions
## Tests
Using unit tests and manually
## Changes
1. Events are now printed in chronological order
2. Simplify events rendering by removing update/flow name. This makes it
more consistent with the web UI too
3. Switch to server side filtering on update_id
## Tests
Manually
Happy run:
```
shreyas.goenka@THW32HFW6T pipeline-progress % bricks bundle run foo
2023-04-12T20:00:22.879Z update_progress INFO "Update e1becc is INITIALIZING."
2023-04-12T20:00:22.906Z update_progress INFO "Update e1becc is SETTING_UP_TABLES."
2023-04-12T20:00:24.496Z update_progress INFO "Update e1becc is RUNNING."
2023-04-12T20:00:24.497Z flow_progress INFO "Flow 'sales_orders_raw' is QUEUED."
2023-04-12T20:00:24.586Z flow_progress INFO "Flow 'sales_orders_raw' is STARTING."
2023-04-12T20:00:24.748Z flow_progress INFO "Flow 'sales_orders_raw' is RUNNING."
2023-04-12T20:00:26.672Z flow_progress INFO "Flow 'sales_orders_raw' has COMPLETED."
2023-04-12T20:00:27.753Z update_progress INFO "Update e1becc is COMPLETED."
```
Sad run:
```
shreyas.goenka@THW32HFW6T pipeline-progress % bricks bundle run foo
2023-04-12T20:02:07.764Z update_progress INFO "Update 04b80e is INITIALIZING."
2023-04-12T20:02:07.870Z update_progress ERROR "Update 04b80e is FAILED."
Error: update failed
```
## Changes
<!-- Summary of your changes that are easy to understand -->
Output now:
```
shreyas.goenka@THW32HFW6T pipeline-progress % bricks bundle run foo
The update can be found at https://e2-dogfood.staging.cloud.databricks.com/#joblist/pipelines/1cc605db-daab-4218-b38a-a63030e3eb03/updates/f92f2159-1141-47de-b1e2-1ca854b7238f
2023-04-12T20:41:19.813Z update_progress INFO "Update f92f21 is INITIALIZING."
2023-04-12T20:41:19.841Z update_progress INFO "Update f92f21 is SETTING_UP_TABLES."
2023-04-12T20:41:21.270Z update_progress INFO "Update f92f21 is RUNNING."
2023-04-12T20:41:21.271Z flow_progress INFO "Flow 'sales_orders_raw' is QUEUED."
2023-04-12T20:41:21.349Z flow_progress INFO "Flow 'sales_orders_raw' is STARTING."
2023-04-12T20:41:21.480Z flow_progress INFO "Flow 'sales_orders_raw' is RUNNING."
2023-04-12T20:41:23.493Z flow_progress INFO "Flow 'sales_orders_raw' has COMPLETED."
2023-04-12T20:41:25.484Z update_progress INFO "Update f92f21 is COMPLETED."
```
## Tests
<!-- How is this tested? -->
## Changes
These are unlikely to ever be DBFS paths so we can remove this level of indirection to simplify.
**Note:** this is a breaking change. Downstream usage of these fields must be updated.
## Tests
Existing tests pass.
## Changes
If a configuration file is located in a subdirectory of the bundle root,
files referenced from that configuration file should be relative to its
configuration file's directory instead of the bundle root.
## Tests
* New tests in `bundle/config/mutator/translate_paths_test.go`.
* Existing tests under `bundle/tests` pass and are augmented to assert
on paths.
---------
Co-authored-by: shreyas-goenka <88374338+shreyas-goenka@users.noreply.github.com>
## Changes
This PR changes the files.Delete() mutator to delete the sync snapshots
file on destroy. This ensures that files will be uploaded when the
bundle is uploaded again.
## Tests
- [x] Manual test: Ran `bricks bundle destroy`, observed that the sync
snapshots file was deleted.
## Changes
This is useful when developing the Databricks Terraform provider where
you keep a local-only build of the provider and refer to it using $HOME
from `~/.terraformrc`, for example like this:
```
plugin_cache_dir = "$HOME/.terraform.d/plugin-cache"
```
## Tests
That $HOME is passed through cannot be tested as is because the
`tfexec.Terraform` struct doesn't expose it through public fields or
methods. What can be tested is a successful run of the initialize
mutator and this is included in this commit.
## Changes
Consider the following host based configuration:
```
bundle:
name: job_with_file_task
workspace:
host: https://e2-dogfood.staging.cloud.databricks.com/
```
If you have a DEFAULT profile, then this host is ignored. The solution
proposed here is to remove the profile config loader if host is
explicitly specified in the bundle config.
This does come with a cost, namely that if a `DATABRICKS_CONFIG_PROFILE`
env var will be ignored, which maybe goes against unified auth spec
The ideal solution here is probably to make a change to go-SDK to not
select DEFAULT profile if host is not empty
## Tests
<!-- How is this tested? -->
## Changes
This change also swaps the order of mutators such that interpolation
happens before path translation. This means that is is possible to use
variables (e.g. `${bundle.environment}`) in notebook or file paths.
## Tests
New tests pass and verified manually.
This PR adds a bundle: "readonly" struct tag to the json schema
generator. This allows us to skip generating json schema for internal
readonly fields
Tested using unit test
## Changes
Pull state before deploying and push state after deploying.
Note: the run command was missing mutators to initialize Terraform. This
is necessary if the cache directory is removed between running "deploy"
and "run" (which is valid now that we synchronize state).
## Tests
Manually.
## Changes
The databricks_permissions resource may be generated if a bundle
resource includes a `permissions` block. There's no need to incorporate
details from the materialization into the bundle configuration struct.
## Tests
Confirmed that this fixes `bricks bundle run` when dealing with a bundle
with permission configuration.
## Changes
Auth relied on setting a profile. In this change we enumerate all
configuration properties and export all non-empty ones as a map with
environment variables. We then pass this map to the Terraform execution
wrapper.
This results in Terraform using the bundle's authentication
configuration.
This change is needed to make #287 work.
## Tests
Manually.
## Changes
This improves out of the box usability where a user who already
configured a `.databrickscfg` file will be able to reference the
workspace host in their `bundle.yml` and it will automatically pick up
the right profile.
## Tests
* Newly added tests pass.
* Manual testing confirms intended behavior.
---------
Co-authored-by: shreyas-goenka <88374338+shreyas-goenka@users.noreply.github.com>
Add configuration:
```
bundle:
lock:
enabled: true
force: false
```
The force field can be set by passing the `--force` argument to `bricks
bundle deploy`. Doing so means the deployment lock is acquired even if
it is currently held. This should only be used in exceptional cases
(e.g. a previous deployment has failed to release the lock).
Adds check for whether file exists locally
case 1: local (relative) file does not exist
```
foo:
name: "[job-output] test-job by shreyas"
tasks:
- task_key: my_notebook_task
existing_cluster_id: ***
notebook_task:
notebook_path: "./doesnotexist"
```
output:
```
shreyas.goenka@THW32HFW6T job-output % bricks bundle deploy
Error: notebook ./doesnotexist not found. Error: open /Users/shreyas.goenka/projects/job-output/doesnotexist: no such file or directory
```
case 2: remote (absolute) file does not exist
```
foo:
name: "[job-output] test-job by shreyas"
tasks:
- task_key: my_notebook_task
existing_cluster_id: ***
notebook_task:
notebook_path: "/Users/shreyas.goenka@databricks.com/doesnotexist"
```
output:
```
shreyas.goenka@THW32HFW6T job-output % bricks bundle deploy
shreyas.goenka@THW32HFW6T job-output % bricks bundle run foo
Error: failed to reach TERMINATED or SKIPPED, got INTERNAL_ERROR: Task my_notebook_task failed with message: Notebook not found: /Users/shreyas.goenka@databricks.com/doesnotexist. This caused all downstream tasks to get skipped.
```
case 3: remote exists
Successful deploy and run
Tested manually:
Before we did not have get any errors/logs and silently failed in this
case
```
shreyas.goenka@THW32HFW6T job-output % bricks bundle run foo
Error: run skipped: Skipping this run because the limit of 1 maximum concurrent runs has been reached.
```
We incorrectly relied on map key iteration order to print debug trace.
This PR switches over to using the tracker struct to allow more reliable
json schema reference loop detection and logging
This also fixes the failing TestSelfReferenceLoopErrors and
TestCrossReferenceLoopErrors tests
This PR:
1. Adds autogeneration of descriptions for `resources` field
2. Autogenerates empty descriptions for any properties in DABs
3. Defines SOPs for how to refresh these descriptions
4. Adds command to generate this documentation
5. Adds Automatically copy any descriptions over to `environments`
property
Basically it provides a framework for adding descriptions to the
generated JSON schema
Tested manually and using unit tests
PR for how to render errors on console for jobs.
Here is the bundle used for the logs below:
```
bundle:
name: deco-438
workspace:
host: https://adb-309687753508875.15.azuredatabricks.net
resources:
jobs:
foo:
name: "[${bundle.name}][${bundle.environment}] a test notebook"
tasks:
- task_key: alpha
existing_cluster_id: 1109-115254-ox7poobk
notebook_task:
notebook_path: "/Users/shreyas.goenka@databricks.com/[deco-438] invalid notebook"
- task_key: beta
existing_cluster_id: 1109-115254-ox7poobk
notebook_task:
notebook_path: "/does-not-exist"
- task_key: gamma
existing_cluster_id: 1109-115254-ox7poobk
notebook_task:
notebook_path: "/Users/shreyas.goenka@databricks.com/[deco-438] valid notebook"
```
And this is a screenshot of the logs from the console:
<img width="1057" alt="Screenshot 2023-02-17 at 7 12 29 PM"
src="https://user-images.githubusercontent.com/88374338/219744768-ab7f1e79-db8f-466a-ad6d-f2b6f85ed17c.png">
Here are the logs when only tasks gamma is executed (successfully):
<img width="1059" alt="Screenshot 2023-02-17 at 7 13 04 PM"
src="https://user-images.githubusercontent.com/88374338/219744992-011d8b91-ec1d-44f0-a849-83c81816dd9f.png">
TODO: Investigate more possible job errors, and make sure state for them
is handled in a robust way here
1. Perform file synchronization on deploy
2. Update notebook file path translation logic to point to the
synchronization target rather than treating the notebook as an artifact
and uploading it separately.
Invoke with `bricks sync SRC DST`.
In bundle context `SRC` and `DST` arguments are taken from bundle configuration.
This PR adds `bricks bundle sync` to disambiguate between the two.
Once the VS Code extension is bundle aware they can again be consolidated.
Consolidating them today would regress the VS Code experience if a
`bundle.yml` file is present in the file tree.
The workspace root path is a base path for bundle storage. If not
specified, it defaults to `~/.bundle/name/environment`. This default, or
other paths starting with `~` are expanded to the current user's home
directory. The configuration also includes fields for the files path,
artifacts path, and state path. By default, these are nested under the
root path, but can be overridden if needed.
This PR contains a struct to allow you to generate JSON schemas from
Golang types and a struct to allow injecting documentation into the json
schema. This will support autocomplete for DABs
If the environment is not set through command line argument or
environment variable, the bundle loads either 1) the only environment,
2) the only environment with the default flag set.
Users can opt out and use the system-installed version with the
following configuration:
```
bundle:
terraform:
exec_path: terraform
```
This will find the binary in $PATH and replace it with the found value.
If this is not set, the initialize phase will install Terraform in the
bundle's cache directory.
Summary:
* All remote path arguments for deployer and locker are now relative to
root specified at initialization
* The workspace client is now a struct field so it doesn't have to be
passed around
This includes 3 mutators:
* Interpolate resources references to TF compatible format
* Convert resources struct to TF JSON format and write it to disk
* Run TF apply
By specifying a function typed `LookupFunction` the caller can customize
which path expressions to interpolate and which ones to skip. When we
express dependencies between resources their values are known by
Terraform at deploy time. Therefore, we have to skip interpolation for
`${resources.jobs.my_job.id}` and instead rewrite it to
`${databricks_job.my_job.id}` before passing it along to Terraform.
Performs interpolation on string field.
It looks for patterns `${foo.bar}` where `foo.bar` points to a string
field in the configuration data model.
It does not support traversal (e.g. `${foo}` with `foo` equal
to`${bar}`), hence "rudimentary".
This adds:
* Top level "artifacts" configuration key
* Support for notebooks (does language detection and upload)
* Merge of per-environment artifacts (or artifact overrides) into top level
Unit tests are now run in all three big OS.
Some of the changes are to make the tests green for windows while we are
skipping some of the other tests on windows/macOS to make the tests
pass. This is a temporary measure and we will incrementally migrate
these tests over so there is parity in unit testing along all three
environments!
While working on artifact upload and workspace interrogation I realized
this mutator interface needs to:
1. Operate at the whole bundle level so it can apply to both
configuration and internal state
2. Include a `context.Context` parameter for a) long running operations
and b) progress reporting
Previous interface:
```
Apply(*config.Root) ([]Mutator, error)
```
New interface:
```
Apply(context.Context, *Bundle) ([]Mutator, error)
```
Used to inspect the bundle configuration after loading and merging all
files.
Once we add variable interpolation this command could show the result
after interpolation as well.
Each of the mutations to this configuration is observable, so we could
add a mode that writes each of the intermediate versions to disk for
even more fine grained introspection.
Load a tree of configuration files anchored at `bundle.yml` into the
`config.Root` struct.
All mutations (from setting defaults to merging files) are observable
through the `mutator.Mutator` interface.
shreyas-goenka
Andrew Nester
Serge Smertin
Shreyas Goenka
Pieter Noordhuis
Miles Yucht
dependabot[bot]
Fabian Jakobs
