State & Checkpoints

Operational commands for inspecting and managing long-running jobs. Most can run against either the embedded local runtime or a remote coordinator.

jobs

krishiv jobs                        # local jobs
krishiv --distributed jobs          # queries the coordinator for its view

Returns a tabular dump of jobs with ID, name, state, and row counts.

state inspect

# Local
krishiv state inspect --job my-pipeline --operator my-operator

# Remote
krishiv -c http://coord.internal:50051 state inspect --job my-pipeline --operator my-operator

# Read state from a custom storage path (e.g. for forensic inspection)
krishiv state inspect --job my-pipeline --operator my-operator --storage-path /var/krishiv/inspect

Dumps the per-key values for a single operator, with type, key (hashed), and value bytes. Sensitive values are redacted unless you opt in.

savepoint

krishiv savepoint --job my-pipeline --label before-v2

Triggers a savepoint with the given label. The savepoint is stored under the job's checkpoint base directory.

restore

# From a specific epoch
krishiv restore --job my-pipeline --epoch 42

# From a savepoint
krishiv restore --job my-pipeline --epoch 17 --storage-path s3://bucket/savepoints/17

Restarts the job from the given epoch. The runtime replays from the chosen checkpoint; the original (running) instance is cancelled.

checkpoints list

krishiv checkpoints list --job my-pipeline
krishiv -c http://coord.internal:50051 checkpoints list --job my-pipeline --storage-path s3://bucket/krishiv/

Lists valid (complete, integrity-checked) checkpoint epochs for a job.

See also

• Savepoints and Migration
• Checkpointing
• Health & Status