Dagster

**Bugfix**

- Enabled `NoOpComputeLogManager` to be configured as the `compute_logs` implementation in
`dagster.yaml`
- Suppressed noisy error messages in logs from skipped steps

**New**

- Improve dagster scheduler state reconciliation.

**New**

- Dagit now allows re-executing arbitrary step subset via step selector syntax, regardless of
whether the previous pipeline failed or not.
- Added a search filter for the root Assets page
- Adds tooltip explanations for disabled run actions
- The last output of the cron job command created by the scheduler is now stored in a file. A new
`dagster schedule logs {schedule_name}` command will show the log file for a given schedule. This
helps uncover errors like missing environment variables and import errors.
- The Dagit schedule page will now show inconsistency errors between schedule state and the cron
tab that were previously only displayed by the `dagster schedule debug` command. As before, these
errors can be resolve using `dagster schedule up`

**Bugfix**

- Fixes an issue with config schema validation on Arrays
- Fixes an issue with initializing K8sRunLauncher when configured via `dagster.yaml`
- Fixes a race condition in Airflow injection logic that happens when multiple Operators try to
create PipelineRun entries simultaneously.
- Fixed an issue with schedules that had invalid config not logging the appropriate error.

**Breaking Changes**

- `dagster pipeline backfill` command no longer takes a `mode` flag. Instead, it uses the mode
specified on the `PartitionSetDefinition`. Similarly, the runs created from the backfill also use
the `solid_subset` specified on the `PartitionSetDefinition`

**BugFix**

- Fixes a bug where using solid subsets when launching pipeline runs would fail config validation.
- (dagster-gcp) allow multiple "bq_solid_for_queries" solids to co-exist in a pipeline
- Improve scheduler state reconciliation with dagster-cron scheduler. `dagster schedule` debug
command will display issues related to missing crob jobs, extraneous cron jobs, and duplicate cron
jobs. Running `dagster schedule up` will fix any issues.

**New**

- The dagster-airflow package now supports loading Airflow dags without depending on initialized
Airflow db
- Improvements to the longitudinal partitioned schedule view, including live updates, run filtering,
and better default states.
- Added user warning for dagster library packages that are out of sync with the core `dagster`
package.

**Bugfix**

- We now only render the subset of an execution plan that has actually executed, and persist that
subset information along with the snapshot.
- pipeline and composite_solid now correctly capture `__doc__` from the function they decorate.
- Fixed a bug with using solid subsets in the Dagit playground

**Bugfix**

- Fixed an issue with strict snapshot ID matching when loading historical snapshots, which caused
errors on the Runs page when viewing historical runs.
- Fixed an issue where `dagster_celery` had introduced a spurious dependency on `dagster_k8s`
(2435)
- Fixed an issue where our Airflow, Celery, and Dask integrations required S3 or GCS storage and
prevented use of filesystem storage. Filesystem storage is now also permitted, to enable use of
these integrations with distributed filesystems like NFS (2436).