Dagster
Released in late May, 1.13.7 patched a SQL injection vulnerability in the ClickHouse libraries and fixed a race condition in run-failure polling.1
Recent patch releases tightened reliability across the stack. 1.13.5 raised the SQLite busy_timeout from 5s to 30s, and job backfills now retry on transient daemon errors — the retry env var was renamed DAGSTER_MAX_BACKFILL_RETRIES.2 1.13.6 added round-robin dispatch for schedule/sensor/asset daemon ticks, so a single busy code location no longer delays others.3 1.13.7 added owners on build_schedule_from_partitioned_job and gave Fivetran components a fetch_column_metadata option.4
The dg CLI expanded its operational surface. dg api run launch landed in 1.13.6 for launching runs against the Dagster+ API.5 Earlier patches added dg api issue create/update/list, dg api job list/get, dg api asset-check list, and dg api asset get-partition-status.
Azure sovereign cloud and Databricks auth broadened integration reach. 1.13.4 added a configurable endpoint_suffix to ADLS2 and Blob Storage resources, covering Azure Government Cloud and Azure China.6 DatabricksClientResource gained a credentials_strategy argument for federated auth flows.7
Version 1.13 introduced the headline features now running in production. Virtual assets (is_virtual), partitioned asset checks, state-backed integration components defaulting to LOCAL_FILESYSTEM, and 20+ new components all shipped in 1.13.0.8 BI assets across Tableau, Looker, Power BI, and Sigma now carry dagster/table_name automatically.
Dagster+ added a Terraform provider and an MCP server. Platform teams can manage deployments, access controls, and alerting declaratively via the provider.9 The MCP integration gives AI assistants programmatic access to Dagster Cloud operations.
