Changelog
Every release of Kubernetes Notes, newest first. A release is a container image tagged v0.1.<n>, where n is the CI run number, and it goes live a few minutes after the merge. Changes that touch only research notes ship with the next release.
v0.1.163
- Larger text across the notes. A reader said it was too small, and measuring agreed: the main reading text was 15.5px on a desktop and 14.5px on a phone, and a quarter of the site was under 14px. It is now 18px and 16px, nothing is under 14px except a few diagram labels, and the page is a little wider so diagrams grow with it. Headings are unchanged.
- cert-manager: seven corrections from Field Guide 09's research and lab, pinned to v1.21.3. The fifth built-in issuer is Venafi (now CyberArk's), the webhook validates and only stamps the requester on a CertificateRequest,
ca.crtis in the Secret only when the issuer hands over its CA (ACME does not), the Gateway API switch isconfig.gatewayAPI.enabled, and a Gateway listener must also use protocol HTTPS or TLS. - Behind the scenes: the changelog names v0.1.160, the release that put Field Guide 09 in the shop.
v0.1.160
- Behind the scenes: Field Guide 09, who signed this cert?, on cert-manager, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.157
- Behind the scenes: Field Guide 09, who signed this cert?, is written: 200 pages on cert-manager v1.21.3 on No. 00's kind lab: the first certificate and everything it leaves behind, a CA of your own with a root and an intermediate, the life of a Certificate from issue to renewal, HTTPS at the Gateway, ACME against Pebble inside the lab over HTTP-01 and DNS-01, mTLS between two Pods and a trust bundle from trust-manager, a certificate for each Pod from the csi driver, who may ask with approver-policy, an issuer and a DNS-01 solver of your own in Go, GitOps with Flux on a second cluster, and day two, with a CA rotated under running apps and everything taken off again. It is not in the shop yet.
v0.1.156
- Behind the scenes: the changelog names v0.1.154, the release that added the Nodes chapter.
v0.1.154
- New Core chapter: Nodes, three sheets. What a Node object is and how a kubelet registers one, capacity against allocatable, the two heartbeats, and what happens second by second when a node dies:
NotReadyabout 50 seconds after the last heartbeat, Pods deleted and replaced about five minutes after that, by each Pod's own toleration. Then pressure eviction, cordon and drain with the messageskubectl drainreally prints, removing a node, the version skew rule and a "why is it NotReady" table. Every timing and message was checked against the v1.37 source and run on a lab cluster. Contents, Architecture, Advanced Scheduling, DaemonSets, Karpenter and Cluster API link to it.
v0.1.153
- Behind the scenes: the changelog names v0.1.152, the release that corrected the Crossplane chapter.
v0.1.152
- Crossplane: ten corrections from Field Guide 08's research and its lab. The latest core is v2.4.3 (2026-10-22), and the functions table has today's releases: patch-and-transform v0.11.2, go-templating v0.13.2, environment-configs v0.9.2, auto-ready v0.7.3. An activation policy picks CRDs only for a provider that declares safe-start; any other provider's MRDs are all Active at install. The rbac-manager covers XRs and managed resources, so another composed kind (a CloudNativePG Cluster, an HTTPRoute) may need a ClusterRole of your own. function-sequencer (v0.6.1) runs after auto-ready, which is no longer "always last". provider-upjet-aws v2.9.0 ships 1063 namespaced and 1063 legacy service CRDs. The CLI install line pins
XP_VERSION=v2.5.0. Deleting a Provider that still has managed resources is not refused, and on the lab it could take their data with it, so the sheet now says to delete the managed resources first. And the uninstall order ends with the CRDs, which Helm leaves behind.
v0.1.151
- Behind the scenes: the changelog names v0.1.150, the release that put Field Guide 08 in the shop.
v0.1.150
- Behind the scenes: Field Guide 08, who made all this?, on Crossplane, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.147
- Behind the scenes: Field Guide 08, who made all this?, is written: 300 pages on Crossplane v2.4.3 on No. 00's kind lab: packages and their revisions, managed resources, Helm releases, PostgreSQL through provider-sql on a CloudNativePG cluster, drift, adoption and deletion, an API of your own with its Composition built a step at a time, a function written in Go, two teams with RBAC and Usages, day two, GitOps with Flux on a second cluster, a failure gallery and taking it all off. It is not in the shop yet.
v0.1.145
- CloudNativePG: thirteen corrections from Field Guide 07's research, and a third sheet on backups. CloudNativePG is CNCF Incubating since a TOC vote on 2026-11-02. Helm names the operator's Deployment cnpg-cloudnative-pg, and the CRD table gains ClusterImageCatalog and FailoverQuorum. The Barman Cloud plugin needs cert-manager and the operator's namespace, and nothing is archived until a Cluster lists a plugin. The in-tree barmanObjectStore goes in 1.32.0. A failover sets the target to pending, waits for every WAL receiver and takes the primary Lease (spec.primaryLease, a promotion gate, not a fence). A recreated instance keeps its name (the lowest free serial, since 1.30). kubectl cnpg backup needs -m plugin on a plugin Cluster, and the chart repo moved to cloudnative-pg.io/charts. A deleted Cluster takes its PVCs and data, and the safety net is the bucket and the snapshots.
v0.1.144
- Behind the scenes: the changelog names v0.1.143, the release that put Field Guide 07 in the shop.
v0.1.143
- Behind the scenes: Field Guide 07, who holds the primary?, on the CloudNativePG operator and its Barman Cloud Plugin, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.140
- Behind the scenes: Field Guide 07, who holds the primary?, is written: 194 pages on the CloudNativePG operator 1.30.1 and its Barman Cloud Plugin on No. 00's kind lab: failover and switchover, acknowledged writes lost on purpose and then kept, backups to RustFS, recovery to a point in time, volume snapshots, a replica cluster in a second namespace, and upgrades from 17 to 18. It goes on sale in a later release. The zine kit gains three crew: a lantern for an instance manager, a crate for an object store and a funnel for a connection pooler.
v0.1.138
- KEDA: eight corrections from Field Guide 06's research. v2.21.0 has three breaking changes, and the token audience is set in the operator's settings, not the TriggerAuthentication. A second HPA on a scaled Deployment does not fight: the webhook refuses a ScaledObject over an existing HPA, and a plain HPA made after it stalls both with AmbiguousSelector. A bad TriggerAuthentication shows only the source's own error, and a missing Secret key is not logged. Since 2.21,
pollingIntervalpaces only status when the minimum is above 0. Pausing deletes KEDA's HPA. Delete KEDA's objects beforehelm uninstall, or they hang onfinalizer.keda.sh. Scale to zero goes straight from any count to 0. The latest release is still v2.21.0; a 2.21.1 patch is agreed.
v0.1.137
- Behind the scenes: the changelog names v0.1.136, the release that put Field Guide 06 in the shop.
v0.1.136
- Behind the scenes: Field Guide 06, who scales my workers?, on KEDA and its HTTP Add-on, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.133
- Behind the scenes: Field Guide 06, who scales my workers?, is written: 148 pages on KEDA 2.21.0 and its HTTP Add-on 0.16.0 (beta) on No. 00's kind lab, with metrics-server and Redis, RabbitMQ, Kafka, Prometheus and PostgreSQL to scale on, and the add-on behind No. 00's kgateway. It goes on sale in a later release. The page checks no longer read an escaped
\$NAME, such as awk's$NF, as a shell variable, and a zine cover title that wraps onto two lines keeps its spaces in the PDF outline.
v0.1.131
- Behind the scenes: the changelog names v0.1.130, the release that fixed the Knative and Knative Operator sheets.
v0.1.130
- Knative and the Knative Operator: fifteen corrections from Field Guide 05's research. The latest release is v1.24.0, and Knative graduated in the CNCF by a TOC vote on 2025-09-11 (announced 2025-10-08). The Operator example pins "1.24": Operator 1.24.0 carries Knative 1.21 to 1.24 only, and a skipped minor is refused, though one minor down passes. Since Operator 1.22,
kubectl get knativeservingshows only TARGET CLUSTER. The Operator puts back only the ConfigMap keys its CR sets, and runs as two Deployments. The activator also stays in a Service's path while spare room is short, until three idle Pods are ready. Gateway API is the fourth networking layer; with none, a Knative Service still gets a URL but is not Ready. Magic DNS needs a LoadBalancer IP, and the hello example's image is pinned by digest.
v0.1.129
- Behind the scenes: the changelog names v0.1.128, the release that put Field Guide 05 in the shop.
v0.1.128
- Behind the scenes: Field Guide 05, who woke my pod?, on the Knative Operator and Knative, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.125
- Behind the scenes: Field Guide 05, who woke my pod?, is written: 150 pages on the Knative Operator and Knative 1.24 on No. 00's kind lab, from Kourier to Gateway API on No. 00's kgateway, with Eventing and Functions. It goes on sale in a later release. The page checks now accept a shell variable with a default, such as
${URL:-$B}, before the page that sets it.
v0.1.123
- Behind the scenes: Field Guide 04, who vouches for this pod?, is now edition 1.2. Page 46 now says a new identity can take a few seconds to reach the agent on another node, and to ask again if it shows only one. No command or printout changed. Everyone who has it gets the new edition in their library.
v0.1.121
- Behind the scenes: the changelog names v0.1.120, the release that put Field Guide 04 1.1 in the shop.
v0.1.120
- Behind the scenes: Field Guide 04, who vouches for this pod?, is now edition 1.1. A fresh lab checked the two facts its captures never showed. Page 077 now says the chart's default identity also skips the namespaces the controller manager ignores. Page 029 now says an agent retries other errors for up to a minute, then gives up too. No command or printout changed. Everyone who has it gets the new edition in their library.
v0.1.118
- SPIFFE and SPIRE: eight corrections from Field Guide 04's research. The versions are SPIRE v1.15.5 and spiffe-csi v0.3.0 (the chart still installs 0.2.7). The SVID lifetimes are SPIRE's defaults, and the Helm chart sets 4h and 1h. The install uses the chart's recommended layout (
spire-mgmt,spire-server,spire-system), with the commands Field Guide 04's lab ran. A ClusterSPIFFEID needsclassName. The CSI driver's namespace must be privileged, but Pods that mount it passrestricted. The trust domain must be lowercase. The bundle refresh hint is the producer's interval in seconds. Graduation was voted on 2022-08-23 and announced on 2022-09-20.
v0.1.117
- Behind the scenes: the changelog names v0.1.116, the release that put Field Guide 04 in the shop.
v0.1.116
- Behind the scenes: Field Guide 04, who vouches for this pod?, on SPIFFE and SPIRE, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.113
- Behind the scenes: Field Guide 04, who vouches for this pod?, is written: 136 pages on SPIFFE and SPIRE v1.15.5 on No. 00's kind lab, with two trust domains on one cluster. It goes on sale in a later release. The page checks now read a drawing's crew captions in place and a listed time, and every field guide's lab waits read a long log correctly.
v0.1.111
- Behind the scenes: Field Guide 03, who bumped my tag?, is now edition 1.1. No page has more than two highlights, as the series' style asks; nothing it says or prints changed. Everyone who has it gets the new edition in their library.
v0.1.109
- Behind the scenes: the changelog names v0.1.108, the release that put Field Guide 03 in the shop.
v0.1.108
- Behind the scenes: Field Guide 03, who bumped my tag?, on Argo CD Image Updater, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.105
- Behind the scenes: the source of Field Guide 03, who bumped my tag?, on Argo CD Image Updater, is in the repo. The site and the shop are unchanged: the shop lists only the books in its catalog.
v0.1.103
- Behind the scenes: Field Guide 02, who pressed sync?, on Argo CD, is in the shop at $20, with its own page and its desk card as a free sample. It joins the bundle Every field guide, so everyone who bought the bundle finds it in their library.
v0.1.100
- Behind the scenes: the source of Field Guide 02, who pressed sync?, on Argo CD, is in the repo. The site and the shop are unchanged: the shop lists only the books in its catalog.
v0.1.98
- Behind the scenes: the shop has a Field guides section. Field Guide 01 is $20, and the bundle Every field guide is $200: it holds Field Guide 01 now and every later field guide as it comes out. Team licences take $1 off every copy at each size, as for the zines. A field guide's page, labels and checkout keys say Field Guide, so they no longer clash with No. 01's.
v0.1.95
- Behind the scenes: the source of Field Guide 01, is git still the truth?, the first Beyond Core book, is in the repo. The site and the shop are unchanged: the shop lists only the books in its catalog.
v0.1.86
- Behind the scenes: the shop has its prices. No. 00 is $1, and each of No. 01 to No. 13 is $7. A team licence takes another $1 off every copy at each size, so a team of 5 pays $30 for a book and a team of 100 pays $300. The whole shelf is $65, $5 a Core book, and the Core shelf bundle is gone.
v0.1.83
- Behind the scenes: the shop's privacy page is dated the day its version 3 went live, and says that deleting a library doesn't delete Stripe's own record of the payment.
v0.1.82
- Behind the scenes: the shop's privacy page says that deleting a library deletes every address it has had and the record of its moves, how a deletion is confirmed, and what stays (the bare order record, and backups for up to six weeks).
v0.1.79
- Behind the scenes: the shop's page for No. 00, who answers my curl?, now offers edition 1.2. Its home-print kit turns the back cover the right way up when you turn the folded booklet over.
v0.1.75
- Behind the scenes: the shop's terms, privacy and refund pages say how support moves a library to another address, for a buyer who lost theirs or is leaving a school address, and that every move is recorded.
v0.1.71
- Behind the scenes: switching the shop's academic price on now needs the day it starts, and new versions of the terms and privacy pages dated that day, so their dates tell buyers which terms an order was placed under.
v0.1.64
- Behind the scenes: the shop can offer teachers, students and school staff 50% off every zine, licence and bundle with a school email address (
.edu, or.eduor.acand a country, such as.ac.uk). It is behind a switch, off for now.
v0.1.60
- Behind the scenes: the shop's page for No. 00, who answers my curl?, now offers edition 1.1. Its PDFs no longer print a faint second frame through the first line of each code box.
v0.1.58
- Behind the scenes: the shop's library marks a book that was refunded with a red Refunded pill, or Charged back after a lost dispute.
v0.1.43
- Behind the scenes: small wording fixes on pages nothing links to yet.
v0.1.41
- Behind the scenes: the link check now covers images too, the build has a few new pages and styles that nothing links to yet, and the security policy allows a form to post to this site.
v0.1.35
- The license page says what the MIT license covers: this site and the source that builds it. The Spec & Status zines have a license of their own.
v0.1.33
- Version refresh: the kgateway pages name v2.4.7 (released 2026-10-16) as the latest stable and pin it in the chart command, and the Gateway API CRDs pages name and pin v1.6.3, the current standard-channel release (v1.7.0 is still a release candidate).
v0.1.32
- Facts the hands-on zines' research found that had not reached the notes yet, each checked on a real kind cluster or in the Kubernetes v1.37 source. Pods:
kubectl describe podlists the five conditions in the order the kubelet writes them, PodReadyToStartContainers (GA in v1.37) first. Services: a Service with no Pod to pick still resolves and refuses calls after about a second, while aninternalTrafficPolicy: Localdrop times out;kubectl get endpointsliceslists every endpoint, butdescribe svcand the deprecatedkubectl get endpoints(its exact warning is quoted) list only ready ones; BusyBox 1.37's nslookup gives only dotless names the search list, sends dotted ones as typed and exits 1 on any NXDOMAIN; an SRV query without_http._tcplists every port; a barewebfrom another namespace leaves the cluster, because web is a public top-level domain; andexternalTrafficPolicy: Localstill answers Pods and the node itself. The-o jsonpathnodePort command is gone, sincekubectl get svcalready shows the port. Volumes and Storage: a local-path PV patched back to Delete is cleaned only while it is Released, never once it is Available. Jobs and CronJobs:kubectl describe jobshows no conditions, so read them withkubectl get job -o yaml.
v0.1.31
- Corrections from the hands-on zines' research on storage (No. 12) and NetworkPolicy (No. 13), each checked on a real kind cluster. Volumes and Storage: a claim's binding is recorded on both sides; a provisioner need not be CSI; ReadWriteOncePod is not CSI only; ReadWriteOnce keeps an attached or local volume on one node, and only ReadWriteOncePod limits it to one Pod; a hostPath PV with node affinity works beyond one node; a grow needs a driver that can grow, and since v1.34 a failed grow can be lowered again, to above the old size;
kubectl delete pvcwaits while the claim is in use; and removing onlyclaimRef.uidkeeps a PV for a claim of the same name. Network Policy in Depth: kind v0.33.0 embeds an older kube-network-policies that enforces no ClusterNetworkPolicy, takes no flags (so it logs no verdicts), keeps open connections after a new deny, lets a new Pod take calls until its node lists it, matches Pod IPs in anipBlock, and lets only root's sockets on a node skip the check.
v0.1.30
- The last corrections from the hands-on zines' research (No. 09 ConfigMaps, No. 10 Secrets, No. 11 RBAC), each checked on a real kind cluster or in the Kubernetes v1.37 source. ConfigMaps and Secrets: a mounted file follows an edit within 60 to 90 s, Pod by Pod, and annotating the Pod refreshes it at once; a container restart re-reads env vars; an immutable object needs a new name, because reusing the old one can leave a node on the old copy;
describeprints a service account token Secret's token in full; static Pods can never use Secrets or ConfigMaps; encryption at rest needs a restart for each edit unless its automatic reload flag is on, and a local key should usesecretbox, notaescbc;dockercfgalso works for image pulls. Volumes and Storage: subPath files update on a restart. RBAC and Service Accounts:pods/execneedscreate; the special verbs now includesign,attestand the betaimpersonate-on:form; every ServiceAccount may read ClusterTrustBundles; kubeadm bindscluster-admintokubeadm:cluster-admins, not onlysystem:masters;viewflows intoeditandeditintoadmin; and a revoked ServiceAccount's token can work for up to 10 s.
v0.1.29
- More corrections from the hands-on zines' research, each checked against the Kubernetes v1.37 source or a real kind cluster. The Jobs and CronJobs page no longer says
kubectl describe jobshows why a Job failed: describe prints no conditions, so the reason shows only in its events. A deleted Job Pod counts as failed and is replaced about 10 s later. Its CronJobs part notes that a schedule edit can run a minute that already passed (open issue 128155) and that events thin out over time. The Rollouts and Probes page says an unready Pod stays in its EndpointSlice, marked not ready, and gets no new Service traffic, and that every probe delay counts from container start. The Services pages say Pods still serving while they stop get traffic when none is ready, and that an empty Service refuses calls after about a second; the SRV note now says which form needs a named port. The Pods page says the first restart after a crash happens at once.
v0.1.28
- Five corrections found while researching the hands-on zines, each checked against the Kubernetes v1.37 source or a real cluster. The Deployments page now lists all eight rules a ReplicaSet uses to pick which Pods go first on a scale-down: the deletion cost is fourth, after Pods that are not yet on a node, not running or not ready. The DaemonSets page now says broken Pods are replaced first when you push a fixed template, but still use up the update budget (since v1.31). The Admission page marks NamespaceLifecycle as mutating, since it runs with the mutating plugins. The Namespaces page says deleting a namespace removes its Pods first, and names kind's kindnet as a plugin that enforces NetworkPolicy. The StatefulSet
maxUnavailabletable notes that 60% was measured again on a kind cluster.
v0.1.27
- Every arrow and hyphen is now in the handwriting font. Eight arrows on the External Secrets, Image Policy and Network Policy pages, and the hyphens in "Add-ons" and "Server-Side" on the part dividers, used characters the font does not have, so they showed in whatever font the device fell back to. The arrows are now written
->, like the rest of the notes.
v0.1.26
- The chapter pills on the Part 2 and Part 3 divider pages are now links: click anywhere on a pill to open its chapter.
v0.1.25
- Links to other chapters now look like links: blue ink, a solid underline and a soft highlight, in both themes. This covers the cross-links inside chapters and every title in the contents lists, which used to read as plain text.
v0.1.24
- Every version on the changelog is now a link: each heading points to its own anchor (for example
/changelog#v0.1.17), so a specific release can be shared, and a linked release is highlighted.
v0.1.23
- Requests are now traced end to end: the site's nginx runs the official OpenTelemetry module, so each page view adds a
kubernetes-notesspan under the gateway's trace in Tempo, with nginx's own route, status and timing. The image moves to the-otelvariant of the same nginx release.
v0.1.22
- Search engines can now list the site:
robots.txtallows crawling and points to a new/sitemap.xmlwith every page, and responses no longer carrynoindex, nofollow.
v0.1.21
- New group in Beyond Core, CI and builds, right after GitOps and delivery: Actions Runner Controller; Gitea Actions Runners; Building Images in the Cluster. Three sheets each, written from three new verified research dossiers (researched with Opus at the highest effort, built and checked with Sonnet).
- Beyond Core now has 29 chapters in 10 groups. The Part 2 contents and divider list the group, and the Pod Security, Image Signing, Attestations, Flux CD v2 and What is Kubernetes sheets link the new chapters.
v0.1.20
- The site's request log now records visitors' addresses only to their /24 network (
a.b.c.0), not in full. Visit counts come from the gateway's logs, which store a salted hash instead.
v0.1.19
- Flux CD v2 gains a fourth sheet, Flux Operator and its Web UI: the FluxInstance and FluxReport objects, what the Web UI shows and which actions it allows, anonymous versus OIDC login with impersonation, its RBAC roles and the Secrets read pitfall, and how it compares to Argo CD's UI. Written from a new verified research dossier.
- The Flux vs Argo CD tables in both chapters no longer say Flux has no UI, and the contents, Flux deep dive and OIDC chapters link the new sheet.
v0.1.18
- Added this changelog at
/changelog, linked from every page footer, with an entry for every version since v0.0.0.
v0.1.17
- New section in Advanced Concepts, Resources and autoscaling, after Scheduling (#15): Requests, Limits and QoS; In-Place Pod Resize; Vertical Pod Autoscaler; Horizontal Pod Autoscaler in Depth. Three sheets each, written from four new verified research dossiers.
- Advanced Concepts now has 34 chapters in 11 groups. The Part 3 contents and divider list the group, and Pods, Deployments and Scaling, Quick Revision, Advanced Scheduling, KEDA, Karpenter, kube-state-metrics and metrics-server link the new chapters.
v0.1.16
- The "One real cluster" notes in the Observability chapters became Seen in practice boxes (#14): each names a common pitfall and a "Do better" hint instead of describing one specific cluster.
- The Flux SOPS and Network Policy sheets lose their last homelab-specific wording.
v0.1.15
- New section in Advanced Concepts, Identity and trust, after Security (#13): OIDC Authentication for the API Server; Keycloak and Single Sign-On; OpenBao: Secrets as a Service; Private PKI and Trust Distribution; Pod Certificates (stable in v1.37).
- Advanced Concepts now has 30 chapters in 10 groups. Fourteen existing chapters link the new ones.
- The section's homelab notes are generic Seen in practice boxes from the start.
v0.1.14
- Every chapter is readable at phone width (#12): wide diagrams in 21 chapters were redrawn narrower or set beside their text, tables use stacked phone layouts, and long code tokens wrap at
_,/and.instead of mid-word.
v0.1.13
- New group in Beyond Core, Observability, next to the OpenTelemetry Operator (#11): Prometheus and ServiceMonitors; Alertmanager Routing; Loki and Log Pipelines with Alloy; Kubernetes Events as a Signal; Tracing with Tempo; Continuous Profiling with Pyroscope.
- Beyond Core now has 26 chapters. The contents grow to a fourth sheet.
- Research dossiers for the section (#10) ship with this release.
v0.1.12
- New chapter in the GitOps group: Secrets in Git with SOPS, how Flux decrypts SOPS-encrypted Secrets, with age and cloud KMS keys.
v0.1.11
- One page per chapter. The single 1.46 MB page became one page per chapter (about 11 to 13 KB gzipped each) plus one shared stylesheet that is cached for good. The whole notebook is still on one page at
/all. - Header menus mark the current chapter, every chapter has previous and next links, and old one-page links are forwarded to their new page.
v0.1.10
- New section in Advanced Concepts, Storage internals: CSI Drivers and the Volume Lifecycle; Volume Topology and Scheduling; Volume Operations and Failure Modes; Snapshots, Clones and Backup.
v0.1.9
- New section in Advanced Concepts, Networking internals: Pod Networking and CNI; The Service Dataplane; Cluster DNS in Depth; Network Policy in Depth.
v0.1.8
- New section in Advanced Concepts, Supply chain security: Images, Registries and Pulls; Image Signing: Sigstore and Notation; SBOMs, Provenance and Attestations; Enforcing Image Policy.
v0.1.7
- New chapter in a new GitOps group: GitOps with Flux: Deep Dive, covering repo layout, dependency layers, drift, multi-tenancy, OCI artifacts and troubleshooting.
v0.1.6
- Release pipeline: the deploy notification now fails the CI run when it does not land, instead of only warning, so a release can no longer sit undeployed unnoticed.
v0.1.5
- The notes are MIT licensed. The license is at
/licenseand linked from every page footer. - Project guide for contributors and agents (AGENTS.md); changes now land only through pull requests.
v0.1.4
- Release pipeline: CI notifies the cluster right after publishing, so a release goes live in about two minutes instead of waiting for the next poll.
v0.1.3
- Link previews show a summary of the notes and the theme colour instead of a bare title.
v0.1.2
- First published release. The site ships as a multi-architecture container image (amd64 and arm64) served by unprivileged nginx with a strict Content Security Policy, built and published by CI on every change.
v0.1.1
- Not published. The first CI run failed (the nginx config check could not write to stderr under the build sandbox), so no image exists for this version.
- Added the container build and the publish workflow that v0.1.2 shipped with.
v0.0.0
- The notebook before the site. Kubernetes Notes started as a hand-drawn style notebook and a claude.ai artifact: Core (10), Beyond Core (16) and Advanced Concepts (9) chapters, the cover, contents and part dividers, the style kit, and the paper and website builds.
- Before the first release it gained Dynamic Resource Allocation, Cluster API, Jobs and CronJobs, Argo CD, Argo CD Image Updater, Karpenter, Crossplane, and StatefulSets and DaemonSets, and one shared chapter list that every build reads.