Command reference

The triage trio — scan, inspect, bottleneck — plus doctor. Every command takes a job by its name or id and connects via your configured profile.

Flags shown here are the common ones. Run flinkit <command> --help for the complete, version-accurate list for your build.

`flinkit scan`

Lists every running job on the cluster at a glance — state, parallelism, restart count, and checkpoint health. Start here to find the job you care about.

shell

$ flinkit scan
JOB                 STATE        PARALLELISM   RESTARTS   CHECKPOINT
orders-rollup       RESTARTING   2             21         failing
sessionize-clicks   RUNNING      8             0          ok
fraud-features      RUNNING      4             0          backpressured

The ok / warn / critical colors are flinkit's severity cues — the same ones you'll see in inspect and findings.

`flinkit inspect`

The main event: a single-screen live triage dashboard for one job. Pass a job name or id.

shell

$ flinkit inspect orders-rollup --watch

flinkit inspect dashboard for orders-rollup: RESTARTING headline, checkpoint and restart stats, per-operator backpressure, I/O, and ranked findings. — `flinkit inspect orders-rollup --watch`

The dashboard is organized top to bottom:

Headline — job name, state, Flink version, and uptime.
checkpoints — succeeded/failed counts, last size and duration, average duration, and the most recent failure reason with a correlated log line.
restarts — the restart count and the last failover exception.
operators — per-operator busy and backpressure, with a bar per operator; arrow keys select one to drill in.
i/o — source and sink records-in / records-out and rates.
findings — ranked, inline (see Findings).
bottleneck — the choke point, or none when the job isn't throughput-bound.
config and logs — collapsible panels.

Flags

Flag	Effect
`--watch`	Refresh the dashboard live in place instead of printing once.

Keybindings

While the dashboard is open:

Key	Action
`↑` `↓`	Move between operators
`enter`	Drill into the selected operator
`c`	Expand / collapse the config panel
`l`	Expand / collapse logs
`w`	Filter logs to warnings
`p`	Pause / resume the live refresh
`q`	Quit

`flinkit bottleneck`

For a throughput-bound job, bottleneck walks the operator graph to the true choke point — not just the first operator that shows backpressure — classifies the cause, and suggests a fix.

shell

$ flinkit bottleneck orders-rollup

If a job isn't backpressured, flinkit says bottleneck: none rather than inventing one — a job can be unhealthy (e.g. failing checkpoints) without being throughput-bound.

`flinkit doctor`

doctor runs flinkit's health checks and prints a ranked, one-shot verdict — the findings, each with evidence, a fix, and a docs link. It's primarily used to verify your flinkit setup and get a quick overall read; for live, per-job triage reach for inspect.

shell

$ flinkit doctor
HIGH    DOC-CKPT-001   Checkpoints are failing; the job cannot recover cleanly from failure.
MEDIUM  DOC-STAB-001   The job has restarted; repeated restarts point to an unstable pipeline.

See Findings for what each code means and how to act on it.