CLI Overview

The krishiv binary is the canonical user interface for everything that does not need an SDK. It handles one-shot SQL, continuous streaming jobs, pipeline projects, state inspection, savepoints, and the lifecycle of local and bare-metal clusters.

Discovering commands

krishiv --help
krishiv help sql            # topic-scoped help
krishiv help stream
krishiv help cluster

Global flags

Flag / env	Effect
-c, --coordinator <URL> / KRISHIV_COORDINATOR	Route state, savepoint, restore, checkpoints commands to a remote coordinator.
-V, --version	Print version and exit.
-h, --help	Print help and exit.
KRISHIV_PRODUCTION	Fail-closed on unsafe overrides. See Auth & Security.

Routing a command to a remote coordinator

Several commands (state, savepoint, restore, checkpoints) can run against either the embedded local runtime or a remote coordinator. Pick the target with -c or KRISHIV_COORDINATOR:

# Local (embedded)
krishiv savepoint --job my-pipeline --label before-v2

# Remote
krishiv -c http://coord.internal:50051 savepoint --job my-pipeline --label before-v2
KRISHIV_COORDINATOR=http://coord.internal:50051 krishiv state inspect --job my-pipeline

All subcommands at a glance

Command	What it does
sql	Run a SQL statement. details
explain	Run a SQL statement and print logical / physical plan.
stream submit / push / poll	Run a continuous window job in-process. details
table read	Read a Parquet, Delta, or Hudi file. details
pipeline init / dry-run / run	Scaffold and execute a pipeline project. details
submit	Submit a job to the in-process R2 scheduler (used by benchmarks and tests).
jobs [--distributed]	List running and recent jobs.
state inspect	Dump per-operator state for a job.
savepoint	Trigger a named savepoint.
restore	Restart a job from a checkpoint or savepoint.
checkpoints list	List valid checkpoint epochs for a job.
local start|stop|restart|status	Manage a local Spark-like cluster. details
cluster start|stop|restart|status|verify-network	Manage a bare-metal cluster. details
coordinator	Run the active coordinator daemon.
clusterd	Run the cluster control plane (co-locates coordinator gRPC + HTTP + optional UI + optional Flight SQL).
job-coordinator	Run a per-job coordinator for very-large-job sharding.
executor	Run an executor.
flight-server	Run a standalone Flight SQL server. (Requires flight-sql feature.)
shuffle-svc	Run a standalone external shuffle service. (Requires shuffle feature.)

See also

• SQL & Explain
• Stream
• Table
• Pipeline
• Local & Cluster
• State & Checkpoints
• Deployment