Reference

Components

Every class in the library, rendered in both themes.

Buttons

ok warn danger info pending idle sha256:9f2ae41c region: jkt-01 state: running production ⌘K

UMRAKD+7

Banners and toasts

Running

28/32

▲ 2 since yesterday

Memory

92%

Spend

$12,480

Area and bars

ingress egress

Donut, gauge, meters

68%
healthy

node-jkt-0174%

node-jkt-0258%

node-jkt-0396%

Table variants

striped · bordered · compact

Name	State	Shape	Node	Age
web-01	running	c2.4x	node-jkt-01	14d
db-02	provisioning	m4.2x	node-jkt-01	2m
cache-01	failed	b1.1x	node-jkt-03	6h
queue-03	draining	c2.2x	node-jkt-02	31d

Lists and media

registry.marspanel.dev/api

12 tags · 4.2 GiB

signed

ci-deploy-key

No expiry set

Rania

Owner · last seen 4m ago

cluster-backup-2026-09-14.tar.zst1.8 GiB

Kadek09:41

Readiness probe timeout is 1s but the app takes 1.4s to warm. That is the whole bug.

Disclosure and steps

Image 2Shape 3Network

Why did this instance restart?

The liveness probe failed three consecutive times. Liveness restarts the container; readiness would only have pulled it from rotation.

Is a snapshot a backup?

No. A snapshot is copy-on-write on the same node and dies with it. A backup lives in object storage and survives node loss.

What counts as N+1?

Every workload still fits after losing any single node. At 92% memory quota this cluster no longer does.

Overview Logs Timeline

Instance ID: i-9f2ae41c7b03d8
Image digest: sha256:9f2ae41c7b03d8
Private address: 10.20.4.17/22
Created: 2026-08-31 09:14 UTC

Forms

Instance name Lowercase letters, digits and hyphens.

Hostname Underscores and capitals are not valid in a hostname.

Endpoint

https:// .marspanel.dev

Shape

Description42 / 240

CPU limit

Labels

tier=frontend env=prod

Placement

Automatic

Most free memory wins.

Pin to node

Never rescheduled.

Nightly snapshots Public address Monitoring

Cloud-init Drop a file here or click to browse YAML, max 64 KiB

Verification code

Estimated $74.20/month

Terminal and diff

node-jkt-01 · exec

$ marspanel instance list --state failed NAME NODE AGE REASON cache-01 node-jkt-03 6h ImagePullBackOff $

readinessProbe:
  httpGet: { path: /healthz, port: 8080 }
  timeoutSeconds: 1  # too tight

14 timeoutSeconds: 1

15- failureThreshold: 1

15+ failureThreshold: 3

Topology

lb-public-01203.0.113.10

web-01in rotation

api-04removed · readiness

Board

Triage3

cache-01 image pull backoff

p16h

Bucket quota alert is noisy

p32d

In progress2

Raise readiness failureThreshold

p2KD

Done7

Add N+1 capacity check

shipped

September 2026

Mon

Tue

Wed

Thu

Fri

Sat

Sun

2Patch window

4Backup verified

8Cert expires

11Migration+2 more

14cache-01 down

Kadek

That is the whole bug.

Rania

Quota bumped, try again.

#incidents

cache-01 is still failing.

KDKadektyping

Today

Readiness timeout is 1s but warmup takes 1.4s.

09:41

So it never passes. Raising failureThreshold to 3.

09:43

That is the whole bug.

09:44

No snapshots yet

Snapshots are copy-on-write and live on the same node.

No matches

Nothing matched state=failed in this region.

Not permitted

Your role viewer cannot delete instances.

Control plane unreachable

Last successful poll 4 minutes ago.

Upload an image

Raw, qcow2 and OCI archives are accepted.

Restore complete

All 14 volumes verified against their checksums.