flinkit documentation

flinkit is a vendor-neutral CLI that diagnoses running Apache Flink jobs. Point it at a cluster and it tells you what's wrong and why — in the terminal, in seconds.

How it works

flinkit reads a live job over the JobManager REST API — the same API the Flink web UI uses — and turns raw metrics into ranked, human-readable findings. There is nothing to deploy into the cluster: no agent, no sidecar, no dashboard to stand up. It's one statically-linked binary that talks to an endpoint you already have.

It works against:

Open-source Apache Flink
Kubernetes — the Flink Operator, via the API-server proxy
Confluent Platform & Confluent Cloud

The triage trio

Three commands take you from the whole cluster down to a single fix:

Command	Use it to
`scan`	List every running job at a glance — state, parallelism, restarts, checkpoint health.
`inspect`	Open a live triage dashboard for one job: diagnosis, checkpoints, operators, findings, bottleneck.
`bottleneck`	Pinpoint the choke point, classify the cause, and suggest a fix.

There's also doctor, a one-shot health verdict used mainly to verify your flinkit setup.

60-second quickstart

shell

# install (private beta — see Getting started)
$ brew install smanjakuppam-at-s3i/tap/flinkit

# point flinkit at a cluster and list its jobs
$ flinkit scan

# open the live triage dashboard for one job
$ flinkit inspect orders-rollup --watch

Want the whole thing on your laptop in five minutes? The try-flinkit demo spins up a local Flink cluster with a deliberately misbehaving job so you can watch flinkit diagnose it.

Where to go next

Getting started — install, license, and your first run.
Configuration & connecting — config file, profiles, and each platform.
Command reference — every command, flag by flag.
Findings — the catalog of what flinkit reports, and what to do about each.

flinkit is in private beta with a 30-day trial. Request access at hello@flinkit.org.