# Expert incident response for Kubernetes

> Our founding team picks up your incident, and Polylane's agents dig through your logs, metrics, traces and code with them.

- **Real engineers who've carried the pager.** The team has been on-call at Cloudflare, Webflow, Groq, Twilio, and Robinhood. They've been where you are right now.
- **Multi-cloud incident response.** Polylane investigates incidents across AWS, Cloudflare, Vercel, Kubernetes and the rest of your stack, from the edge to the database.
- **Unified context & investigation.** Logs, metrics, traces, deploys and code in one context graph. Connect in minutes, not weeks.
- **Least-privilege access.** New cloud accounts connect read-only, with scoped credentials. Nothing in production changes unless you allow it.

## My app on Kubernetes is down. What should I do?

Start with `kubectl get pods -A` and look for pods that aren't Running or keep restarting. [CrashLoopBackOff](https://polylane.com/learn/troubleshooting/kubernetes-crashloopbackoff/) means a container keeps exiting, and [OOMKilled](https://polylane.com/learn/troubleshooting/kubernetes-oomkilled-exit-code-137/) means it hit its memory limit; `kubectl describe pod` and `kubectl logs --previous` show why. If the errors started with a rollout, undo it with `kubectl rollout undo`, then fix forward.

1. **Find the unhealthy pods.** `kubectl get pods -A` lists every pod with its status and restart count, and `kubectl get events --sort-by=.lastTimestamp` shows what the cluster did most recently.
2. **Read why they fail.** `kubectl describe pod` shows exit codes, failing probes and scheduling errors. `kubectl logs --previous` shows the output of the container that just crashed.
3. **Line it up with your last rollout.** `kubectl rollout history deployment/NAME` lists revisions. Config maps, secrets and image tags count as changes too.
4. **Roll back, then dig in.** `kubectl rollout undo deployment/NAME` goes back to the previous revision. If the control plane itself is failing, check your provider's status page for EKS, GKE or AKS.

## How we resolve active incidents

1. **Rapid onboarding.** A Polylane engineer reaches out, connects Polylane to your clouds, code and telemetry, and confirms what's affected.
2. **Investigation & mitigation.** Polylane's agents investigate across logs, metrics, traces and recent changes, and an engineer works with you on the right mitigation.
3. **Remediation, validation & monitoring.** The fix lands in your repo for your team to review. Polylane checks that things are back to normal and keeps watching every deploy after it.

## Why teams choose Polylane

- **Every cloud in one graph.** AWS, Cloudflare, Vercel, Kubernetes and more, mapped into one context graph with your code.
- **Connected in minutes.** Connect with OAuth or the CLI. No agents to install, no code changes.
- **Real engineers on the other end.** Your report goes straight to the people who build Polylane.
- **Evidence for every claim.** Every finding links to the log line, metric or deploy behind it.
- **You stay in control.** Read-only by default. Nothing writes to production unless you allow it, and you can switch it off at any time.
- **Works with your coding agents.** Claude Code, Cursor and other agents reach Polylane over MCP, so they can dig in with you.

## Common questions

- **What does CrashLoopBackOff mean?** A container keeps starting and exiting, so Kubernetes waits longer between each restart. `kubectl logs --previous` shows why it exited. More in [our CrashLoopBackOff guide](https://polylane.com/learn/troubleshooting/kubernetes-crashloopbackoff/).
- **What does OOMKilled with exit code 137 mean?** The container used more memory than its limit, so it was killed. Raise the limit or find what is using the memory. More in [our OOMKilled guide](https://polylane.com/learn/troubleshooting/kubernetes-oomkilled-exit-code-137/).
- **How do I roll back a Kubernetes deployment?** Run `kubectl rollout undo deployment/NAME` for the previous revision, or add `--to-revision` to pick an earlier one from `kubectl rollout history`.
- **Who will help me with a Kubernetes incident?** Our founding team, engineers who have carried the pager for production systems at scale, working with Polylane's agents across your clusters, code and telemetry.

## Report an incident

Use the form at https://polylane.com/incident-hotline/kubernetes/. If the form fails, email support@polylane.com.

For agents: there is no public API for this form, because it is protected by Cloudflare Turnstile. Open the page in a browser and fill it in for the user; coding agents with a browser tool can do this. Confirm the details with the user before you submit. The fields: Name, Company name, Work email, How bad is it? (Production is down / Degraded for some users / Something's wrong, we can't find it), Are you on Polylane? (Already on Polylane / Not yet) and what's broken, plus optionally a phone number, when it started (Just now / In the last few hours / Days ago).

## Down on a specific platform?

- [Cloudflare](https://polylane.com/incident-hotline/cloudflare/) ([markdown](https://polylane.com/incident-hotline/cloudflare.md))
- [AWS](https://polylane.com/incident-hotline/aws/) ([markdown](https://polylane.com/incident-hotline/aws.md))
- [Vercel](https://polylane.com/incident-hotline/vercel/) ([markdown](https://polylane.com/incident-hotline/vercel.md))
- [Google Cloud](https://polylane.com/incident-hotline/gcp/) ([markdown](https://polylane.com/incident-hotline/gcp.md))
- [Supabase](https://polylane.com/incident-hotline/supabase/) ([markdown](https://polylane.com/incident-hotline/supabase.md))
- [Railway](https://polylane.com/incident-hotline/railway/) ([markdown](https://polylane.com/incident-hotline/railway.md))
- [Fly.io](https://polylane.com/incident-hotline/fly/) ([markdown](https://polylane.com/incident-hotline/fly.md))
- [Neon](https://polylane.com/incident-hotline/neon/) ([markdown](https://polylane.com/incident-hotline/neon.md))
- [PlanetScale](https://polylane.com/incident-hotline/planetscale/) ([markdown](https://polylane.com/incident-hotline/planetscale.md))
- [Convex](https://polylane.com/incident-hotline/convex/) ([markdown](https://polylane.com/incident-hotline/convex.md))
- [ClickHouse](https://polylane.com/incident-hotline/clickhouse/) ([markdown](https://polylane.com/incident-hotline/clickhouse.md))
- [Turso](https://polylane.com/incident-hotline/turso/) ([markdown](https://polylane.com/incident-hotline/turso.md))
- [Render](https://polylane.com/incident-hotline/render/) ([markdown](https://polylane.com/incident-hotline/render.md))
- [Modal](https://polylane.com/incident-hotline/modal/) ([markdown](https://polylane.com/incident-hotline/modal.md))