Tink vs Kube-router
Kube-router is a Kubernetes CNI with BGP, IPVS, and NetworkPolicy. It also exposes controller and session gauges for Prometheus. That is not a mechanic. Tink installs in one command, watches the Linux server, explains what is wrong, and helps you fix it.
Kube-router scrapes BGP. Tink answers “why is this VPS sick?”
Accidental sysadmins land on Kube-router because it is the lighter CNI that also replaces kube-proxy. Then the volume fills, nginx dies, and Grafana still shows Established peers — no PromQL matched host fill. A filling /var is a Tink warning; it is not a GoBGP job unless the BGP session itself is the failure.
Tink is the other job: detect the issue on the machine, say why it happened in plain English, propose the command, and run it only after you approve. Keep scraping Kube-router if you already operate Kubernetes and need CNI SLOs. Use Tink if you run Linux servers and want a mechanic.
| Feature | Tink | Kube-router |
|---|---|---|
| Setup time | 30 seconds (one curl | sh command) | Hours — kube-router DaemonSet, CNI conf, BGP peers, IPVS vs iptables, NetworkPolicy, and prometheus.yml scrape |
| What you get | Working monitoring, diagnosis, and approved fixes | A CNI scrape target. Host disk fill, nginx, Postgres, and certs stay your problem until PromQL fires and you SSH in |
| Pricing | Free (Scout) / $9 / $29 per machine per month | Open-source (Apache 2.0). You still pay cluster nodes, BGP routers, Prometheus, Grafana, and the engineer who owns GoBGP |
| Hidden costs | None — fully managed | BGP peer flaps, IPVS conntrack surprises, hairpin SNAT, kube-router CrashLoops, and a second tool to tell you the box is sick |
| Monitoring approach | Agent on the server — CPU, disk, services, logs, certs, ports | controller, BGP session, and IPVS gauges. Kube-router does not watch host fill or a quiet cert unless you added another exporter |
| Configuration | None after install — heuristics and AI | advertise-cluster-ip, peer-router-ips, enable-overlay, hairpin-mode, and which PromQL actually pages you |
| Plain-English diagnosis | Yes — AI explains root cause, impact, and fix | You graph kube_router_controller_bgp_peers in Grafana, then SSH in to change the box |
| Fix execution | Proposes and executes approved commands with an audit trail | Metrics only — Kube-router cannot free disk, restart nginx, or rotate a cert |
| Alerting | Built-in across 8 channels | Not a Kube-router job — Prometheus rules and Alertmanager have to fire |
| Predictive alerts | Yes — disk fills in ~6 days, memory and CPU trends | Not a Kube-router job — a CNI scrape does not forecast a quiet volume filling up |
| SSH brute-force detection | Built-in — parses auth.log every scan | Never — Kube-router does not read auth.log, journald, or host firewall state |
| Machine offline detection | Agent presence monitoring with multi-channel alerts | Silence if the agent dies — unless you built an up{job="kube-router"} deadman check |
| Public status page | Shareable URL with 90-day history | None — Kube-router has no customer-facing status page |
| Weekly fleet digest | Automated Monday digest + daily brief when issues are open | None — a BGP scrape target is not a plain-English fleet narrative |
| On-call tracking | Built-in /oncall command + incident acknowledgment | Not included — wire Alertmanager into PagerDuty or another incident tool |
| Conversation interface | Telegram, WhatsApp, web dashboard, CLI | kubectl and Grafana. No mechanic you text when /var filled and BGP still showed Established |
| Learning curve | None — works after install | Steep — GoBGP, IPVS, overlay vs direct routing, NetworkPolicy, PromQL, and which peer counter actually pages |
| Best for | Freelancers, small teams, accidental sysadmins (1-50 Linux servers) | Teams that already run Kubernetes and need CNI/BGP scrapes, not a sick-VPS mechanic |
When Kube-router metrics are the right choice
Keep scraping Kube-router when you already want that exact job:
- Kubernetes CNI SLOs — you need BGP peers, IPVS, and NetworkPolicy counters, not a mechanic.
- Pod routing at cluster scale — GoBGP advertisements and hairpin mode matter more than host nginx, certs, and SSH on a quiet VPS.
- An existing Prometheus + Grafana stack — kube-router already clusters, BGP already scrapes, and someone owns the peer list.
- Hosts that only page on Kube-router — you already run exporters and PromQL, not a sick-box SSH loop.
The real cost of “just scrape Kube-router”
Kube-router is a serious Kubernetes CNI. A working ops loop for five Linux boxes is not:
- A healthy BGP Established still means you SSH in and change the box by hand
- Host fill, nginx, certs, and SSH brute-force never appear unless Prometheus scraped those signals and a rule fired
- A filling /var is a Tink disk warning on every scan — Kube-router will stay quiet until the BGP session itself is the failure
- A single VPS with a dead kube-router still needs a mechanic, not another CNI scrape
For a 5-server team, Tink Mechanic at $45/month is cheaper than Kube-router plus the engineer who keeps Prometheus scrapes and BGP peer rules alive, and every scan includes a diagnosis a quiet Grafana board will not type.
No BGP homework. No silent host. One command install.
Start MechanicAlso compare: Tink vs Calico · Tink vs Cilium · Tink vs Prometheus