fix: restore tree dropped by 6d224861; land stem95su gdrive-sync (10m) [ci skip]
6d224861 came from a --no-checkout worktree whose empty index made the
commit drop every file except two. This restores 05b50d2b's full tree and
correctly adds stacks/stem95su/gdrive-sync.tf + the service-catalog stem95su
entry. Forward-only (parent=6d224861, no force-push); [ci skip] since the
live infra was never applied from the broken commit.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>This commit is contained in:
Viktor Barzin
parent
6d224861c4
commit
fd0f4a0365
1166 changed files with 358546 additions and 0 deletions
23
stacks/rbac/main.tf
Normal file
23
stacks/rbac/main.tf
Normal file
|
|
@ -0,0 +1,23 @@
|
|||
variable "tls_secret_name" { type = string }
|
||||
variable "ssh_private_key" {
|
||||
type = string
|
||||
default = ""
|
||||
sensitive = true
|
||||
}
|
||||
|
||||
data "vault_kv_secret_v2" "secrets" {
|
||||
mount = "secret"
|
||||
name = "platform"
|
||||
}
|
||||
|
||||
locals {
|
||||
k8s_users = jsondecode(data.vault_kv_secret_v2.secrets.data["k8s_users"])
|
||||
}
|
||||
|
||||
module "rbac" {
|
||||
source = "./modules/rbac"
|
||||
tier = local.tiers.cluster
|
||||
tls_secret_name = var.tls_secret_name
|
||||
k8s_users = local.k8s_users
|
||||
ssh_private_key = var.ssh_private_key
|
||||
}
|
||||
160
stacks/rbac/modules/rbac/apiserver-oidc.tf
Normal file
160
stacks/rbac/modules/rbac/apiserver-oidc.tf
Normal file
|
|
@ -0,0 +1,160 @@
|
|||
# Configure kube-apiserver authentication via a structured
|
||||
# AuthenticationConfiguration file (apiserver.config.k8s.io/v1, GA on k8s 1.30+).
|
||||
#
|
||||
# WHY structured config instead of the legacy --oidc-* flags: the apiserver can
|
||||
# only carry ONE legacy issuer, but we need TWO — the `kubernetes` app (kubectl
|
||||
# / kubelogin CLI) AND the `k8s-dashboard` app (oauth2-proxy in front of the
|
||||
# Kubernetes Dashboard). Structured config supports multiple JWT issuers.
|
||||
#
|
||||
# Both issuers map username<-email and groups<-groups with EMPTY prefixes, to
|
||||
# match the existing RBAC subjects (kind: User, name: <raw email>; group names
|
||||
# verbatim). Do NOT add a prefix or existing bindings break.
|
||||
#
|
||||
# DRIFT WARNING: this edits the kube-apiserver static-pod manifest on the single
|
||||
# master. A `kubeadm upgrade` regenerates that manifest and DROPS this flag (this
|
||||
# is exactly how OIDC silently broke before — the flag was wiped and the
|
||||
# content-hash trigger never re-fired). After any k8s control-plane upgrade,
|
||||
# re-apply the rbac stack to restore apiserver OIDC. See
|
||||
# docs/plans/2026-06-04-k8s-dashboard-sso-design.md.
|
||||
#
|
||||
# SAFETY: the remote script health-gates on /livez and AUTO-ROLLS-BACK the
|
||||
# manifest from a timestamped backup if the apiserver does not recover, so a
|
||||
# malformed config cannot leave the single master down.
|
||||
|
||||
variable "k8s_master_host" {
|
||||
type = string
|
||||
default = "10.0.20.100"
|
||||
}
|
||||
|
||||
variable "ssh_private_key" {
|
||||
type = string
|
||||
sensitive = true
|
||||
}
|
||||
|
||||
variable "oidc_issuer_url" {
|
||||
type = string
|
||||
default = "https://authentik.viktorbarzin.me/application/o/kubernetes/"
|
||||
}
|
||||
|
||||
variable "oidc_client_id" {
|
||||
type = string
|
||||
default = "kubernetes"
|
||||
}
|
||||
|
||||
variable "k8s_dashboard_issuer_url" {
|
||||
type = string
|
||||
default = "https://authentik.viktorbarzin.me/application/o/k8s-dashboard/"
|
||||
}
|
||||
|
||||
variable "k8s_dashboard_audience" {
|
||||
type = string
|
||||
default = "k8s-dashboard"
|
||||
}
|
||||
|
||||
locals {
|
||||
apiserver_auth_config_yaml = <<-YAML
|
||||
apiVersion: apiserver.config.k8s.io/v1
|
||||
kind: AuthenticationConfiguration
|
||||
jwt:
|
||||
- issuer:
|
||||
url: "${var.oidc_issuer_url}"
|
||||
audiences:
|
||||
- "${var.oidc_client_id}"
|
||||
claimMappings:
|
||||
username:
|
||||
claim: email
|
||||
prefix: ""
|
||||
groups:
|
||||
claim: groups
|
||||
prefix: ""
|
||||
- issuer:
|
||||
url: "${var.k8s_dashboard_issuer_url}"
|
||||
audiences:
|
||||
- "${var.k8s_dashboard_audience}"
|
||||
claimMappings:
|
||||
username:
|
||||
claim: email
|
||||
prefix: ""
|
||||
groups:
|
||||
claim: groups
|
||||
prefix: ""
|
||||
YAML
|
||||
|
||||
# Indentation-safe manifest editor: appends the --authentication-config flag
|
||||
# using the exact leading whitespace of the --authorization-mode line.
|
||||
apiserver_flag_insert_py = <<-PY
|
||||
import sys
|
||||
p = sys.argv[1]
|
||||
lines = open(p).read().splitlines(True)
|
||||
out, done = [], False
|
||||
for ln in lines:
|
||||
out.append(ln)
|
||||
if not done and '- --authorization-mode=' in ln:
|
||||
indent = ln[:len(ln) - len(ln.lstrip())]
|
||||
out.append(indent + '- --authentication-config=/etc/kubernetes/pki/auth-config.yaml\n')
|
||||
done = True
|
||||
open(p, 'w').writelines(out)
|
||||
print('flag-inserted' if done else 'ANCHOR-NOT-FOUND')
|
||||
PY
|
||||
|
||||
# Whole remote operation, base64-embedded for byte-exact transfer (no
|
||||
# heredoc/escaping hazards across SSH).
|
||||
apiserver_auth_remote_script = <<-SH
|
||||
MANIFEST=/etc/kubernetes/manifests/kube-apiserver.yaml
|
||||
AUTHCFG=/etc/kubernetes/pki/auth-config.yaml
|
||||
TS=$(date +%s)
|
||||
|
||||
# 1. Write the structured AuthenticationConfiguration (hot-reloaded by the
|
||||
# apiserver on change; mounted into the pod via the existing pki hostPath).
|
||||
echo '${base64encode(local.apiserver_auth_config_yaml)}' | base64 -d | sudo tee "$AUTHCFG" >/dev/null
|
||||
sudo chmod 600 "$AUTHCFG"
|
||||
|
||||
# 2. Ensure the apiserver references it. Only touch the manifest (→ restart)
|
||||
# when the flag is missing; otherwise the file write above hot-reloads.
|
||||
if ! sudo grep -q -- '--authentication-config=' "$MANIFEST"; then
|
||||
sudo cp "$MANIFEST" "$MANIFEST.bak.$TS"
|
||||
sudo sed -i '/--oidc-issuer-url/d;/--oidc-client-id/d;/--oidc-username-claim/d;/--oidc-groups-claim/d' "$MANIFEST"
|
||||
echo '${base64encode(local.apiserver_flag_insert_py)}' | base64 -d | sudo python3 - "$MANIFEST"
|
||||
fi
|
||||
|
||||
# 3. Fail loudly if the flag still isn't present (e.g. anchor not found).
|
||||
if ! sudo grep -q -- '--authentication-config=' "$MANIFEST"; then
|
||||
echo "ERROR: --authentication-config absent after edit"; exit 1
|
||||
fi
|
||||
|
||||
# 4. Health-gate on /livez; auto-rollback the manifest if it never recovers.
|
||||
echo "Waiting for kube-apiserver /livez ..."
|
||||
ok=0
|
||||
for i in $(seq 1 60); do
|
||||
sleep 2
|
||||
if curl -sk https://localhost:6443/livez 2>/dev/null | grep -q '^ok'; then ok=1; break; fi
|
||||
done
|
||||
if [ "$ok" != "1" ]; then
|
||||
echo "kube-apiserver UNHEALTHY after change — rolling back"
|
||||
BAK=$(ls -t "$MANIFEST".bak.* 2>/dev/null | head -1)
|
||||
if [ -n "$BAK" ]; then sudo cp "$BAK" "$MANIFEST"; fi
|
||||
for i in $(seq 1 60); do sleep 2; if curl -sk https://localhost:6443/livez 2>/dev/null | grep -q '^ok'; then break; fi; done
|
||||
echo "rolled back to previous manifest"; exit 1
|
||||
fi
|
||||
echo "kube-apiserver healthy with multi-issuer --authentication-config"
|
||||
SH
|
||||
}
|
||||
|
||||
resource "null_resource" "apiserver_oidc_config" {
|
||||
connection {
|
||||
type = "ssh"
|
||||
user = "wizard"
|
||||
host = var.k8s_master_host
|
||||
private_key = var.ssh_private_key
|
||||
}
|
||||
|
||||
provisioner "remote-exec" {
|
||||
inline = [
|
||||
"echo '${base64encode(local.apiserver_auth_remote_script)}' | base64 -d | bash",
|
||||
]
|
||||
}
|
||||
|
||||
triggers = {
|
||||
auth_config = sha256(local.apiserver_auth_config_yaml)
|
||||
}
|
||||
}
|
||||
174
stacks/rbac/modules/rbac/audit-policy.tf
Normal file
174
stacks/rbac/modules/rbac/audit-policy.tf
Normal file
|
|
@ -0,0 +1,174 @@
|
|||
# Deploy audit policy to k8s-master and configure kube-apiserver to use it.
|
||||
# Audit logs are written to /var/log/kubernetes/audit.log on the master node.
|
||||
# Alloy (log collector DaemonSet) will pick them up and ship to Loki.
|
||||
|
||||
resource "null_resource" "audit_policy" {
|
||||
connection {
|
||||
type = "ssh"
|
||||
user = "wizard"
|
||||
host = var.k8s_master_host
|
||||
private_key = var.ssh_private_key
|
||||
}
|
||||
|
||||
# Upload audit policy file
|
||||
provisioner "file" {
|
||||
content = yamlencode({
|
||||
apiVersion = "audit.k8s.io/v1"
|
||||
kind = "Policy"
|
||||
rules = [
|
||||
{
|
||||
# Don't log requests to the API discovery endpoints (very noisy)
|
||||
level = "None"
|
||||
resources = [{
|
||||
group = ""
|
||||
resources = ["endpoints", "services", "services/status"]
|
||||
}]
|
||||
users = ["system:kube-proxy"]
|
||||
},
|
||||
{
|
||||
# Don't log watch requests (very noisy)
|
||||
level = "None"
|
||||
verbs = ["watch"]
|
||||
},
|
||||
{
|
||||
# Don't log health checks
|
||||
level = "None"
|
||||
nonResourceURLs = ["/healthz*", "/readyz*", "/livez*"]
|
||||
},
|
||||
{
|
||||
# Log secret access at Metadata level only (no request/response bodies)
|
||||
level = "Metadata"
|
||||
resources = [{
|
||||
group = ""
|
||||
resources = ["secrets"]
|
||||
}]
|
||||
},
|
||||
{
|
||||
# Log all other mutating requests at RequestResponse level
|
||||
level = "RequestResponse"
|
||||
verbs = ["create", "update", "patch", "delete"]
|
||||
},
|
||||
{
|
||||
# Log read requests at Metadata level
|
||||
level = "Metadata"
|
||||
verbs = ["get", "list"]
|
||||
},
|
||||
]
|
||||
})
|
||||
destination = "/tmp/audit-policy.yaml"
|
||||
}
|
||||
|
||||
provisioner "remote-exec" {
|
||||
inline = [
|
||||
# Move audit policy to proper location
|
||||
"sudo mkdir -p /etc/kubernetes/policies",
|
||||
"sudo mv /tmp/audit-policy.yaml /etc/kubernetes/policies/audit-policy.yaml",
|
||||
"sudo chown root:root /etc/kubernetes/policies/audit-policy.yaml",
|
||||
|
||||
# Create audit log directory
|
||||
"sudo mkdir -p /var/log/kubernetes",
|
||||
|
||||
# Idempotently add audit flags, volumes, and volumeMounts using Python
|
||||
# to avoid sed duplication bugs on re-runs
|
||||
<<-SCRIPT
|
||||
sudo python3 -c "
|
||||
import yaml
|
||||
|
||||
path = '/etc/kubernetes/manifests/kube-apiserver.yaml'
|
||||
with open(path) as f:
|
||||
doc = yaml.safe_load(f)
|
||||
|
||||
container = doc['spec']['containers'][0]
|
||||
cmd = container['command']
|
||||
|
||||
# Add audit flags if missing
|
||||
audit_flags = {
|
||||
'--audit-policy-file=/etc/kubernetes/policies/audit-policy.yaml': True,
|
||||
'--audit-log-path=/var/log/kubernetes/audit.log': True,
|
||||
'--audit-log-maxage=7': True,
|
||||
'--audit-log-maxbackup=3': True,
|
||||
'--audit-log-maxsize=100': True,
|
||||
}
|
||||
existing = set(cmd)
|
||||
for flag in audit_flags:
|
||||
if flag not in existing:
|
||||
cmd.append(flag)
|
||||
|
||||
# Add volumes if missing (deduplicate by name)
|
||||
vol_names = {v['name'] for v in doc['spec']['volumes']}
|
||||
for vol in [
|
||||
{'name': 'audit-policy', 'hostPath': {'path': '/etc/kubernetes/policies', 'type': 'DirectoryOrCreate'}},
|
||||
{'name': 'audit-log', 'hostPath': {'path': '/var/log/kubernetes', 'type': 'DirectoryOrCreate'}},
|
||||
]:
|
||||
if vol['name'] not in vol_names:
|
||||
doc['spec']['volumes'].append(vol)
|
||||
vol_names.add(vol['name'])
|
||||
|
||||
# Add volumeMounts if missing (deduplicate by mountPath)
|
||||
mount_paths = {vm['mountPath'] for vm in container['volumeMounts']}
|
||||
for vm in [
|
||||
{'mountPath': '/etc/kubernetes/policies', 'name': 'audit-policy', 'readOnly': True},
|
||||
{'mountPath': '/var/log/kubernetes', 'name': 'audit-log'},
|
||||
]:
|
||||
if vm['mountPath'] not in mount_paths:
|
||||
container['volumeMounts'].append(vm)
|
||||
mount_paths.add(vm['mountPath'])
|
||||
|
||||
with open(path, 'w') as f:
|
||||
yaml.dump(doc, f, default_flow_style=False, sort_keys=False)
|
||||
|
||||
print('Audit config applied (idempotent)')
|
||||
"
|
||||
SCRIPT
|
||||
,
|
||||
|
||||
# Wait for API server to restart
|
||||
"echo 'Waiting for API server to restart with audit logging...'",
|
||||
"sleep 30",
|
||||
"sudo kubectl --kubeconfig=/etc/kubernetes/admin.conf get nodes || echo 'API server still restarting'",
|
||||
]
|
||||
}
|
||||
|
||||
triggers = {
|
||||
policy_version = "v1" # Bump to force re-apply of manifest flags
|
||||
policy_hash = sha256(yamlencode({
|
||||
apiVersion = "audit.k8s.io/v1"
|
||||
kind = "Policy"
|
||||
rules = [
|
||||
{
|
||||
level = "None"
|
||||
resources = [{
|
||||
group = ""
|
||||
resources = ["endpoints", "services", "services/status"]
|
||||
}]
|
||||
users = ["system:kube-proxy"]
|
||||
},
|
||||
{
|
||||
level = "None"
|
||||
verbs = ["watch"]
|
||||
},
|
||||
{
|
||||
level = "None"
|
||||
nonResourceURLs = ["/healthz*", "/readyz*", "/livez*"]
|
||||
},
|
||||
{
|
||||
level = "Metadata"
|
||||
resources = [{
|
||||
group = ""
|
||||
resources = ["secrets"]
|
||||
}]
|
||||
},
|
||||
{
|
||||
level = "RequestResponse"
|
||||
verbs = ["create", "update", "patch", "delete"]
|
||||
},
|
||||
{
|
||||
level = "Metadata"
|
||||
verbs = ["get", "list"]
|
||||
},
|
||||
]
|
||||
}))
|
||||
}
|
||||
|
||||
depends_on = [null_resource.apiserver_oidc_config]
|
||||
}
|
||||
97
stacks/rbac/modules/rbac/dashboard-sa.tf
Normal file
97
stacks/rbac/modules/rbac/dashboard-sa.tf
Normal file
|
|
@ -0,0 +1,97 @@
|
|||
# Per-namespace-owner ServiceAccount + long-lived token for Kubernetes Dashboard
|
||||
# token-paste access.
|
||||
#
|
||||
# WHY: seamless OIDC SSO into the dashboard is blocked — the apiserver currently
|
||||
# rejects all Authentik OIDC tokens (see docs/plans/2026-06-04-k8s-dashboard-sso-design.md
|
||||
# §12). Until that's solved, each namespace-owner gets a ServiceAccount scoped to
|
||||
# `admin` on their namespace(s) + cluster read-only, and a long-lived token they
|
||||
# paste into the dashboard "Token" login. Real per-namespace isolation, no OIDC
|
||||
# dependency. Rotate a token by deleting+recreating its `dashboard-<user>-token`
|
||||
# Secret. Retrieve with:
|
||||
# kubectl -n <ns> get secret dashboard-<user>-token -o jsonpath='{.data.token}' | base64 -d
|
||||
#
|
||||
# Driven by the same `local.namespace_owner_pairs` as the OIDC bindings, so every
|
||||
# namespace-owner in k8s_users automatically gets one.
|
||||
|
||||
resource "kubernetes_service_account" "dashboard_owner" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "dashboard-${each.value.user_key}"
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
}
|
||||
|
||||
# Full admin within the owner's namespace (same scope as their OIDC RoleBinding).
|
||||
resource "kubernetes_role_binding" "dashboard_owner_admin" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "dashboard-owner-${each.value.user_key}"
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = "admin"
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "ServiceAccount"
|
||||
name = kubernetes_service_account.dashboard_owner[each.key].metadata[0].name
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
}
|
||||
|
||||
# Minimal cluster-read for the dashboard nav ONLY: the namespace picker needs to
|
||||
# list namespaces, and the Nodes view needs nodes. Deliberately does NOT grant
|
||||
# cluster-wide read of pods/services/configmaps/etc — a namespace-owner can see
|
||||
# the namespace LIST but can only read resources INSIDE their own namespace
|
||||
# (where they have `admin`). Keeps tenants from reading each other's workloads
|
||||
# and configmaps. (Separate from the broader OIDC `namespace_owner_readonly`.)
|
||||
resource "kubernetes_cluster_role" "dashboard_nav_readonly" {
|
||||
metadata {
|
||||
name = "dashboard-nav-readonly"
|
||||
}
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["namespaces", "nodes"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
}
|
||||
|
||||
resource "kubernetes_cluster_role_binding" "dashboard_owner_readonly" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "dashboard-readonly-${each.value.user_key}-${each.value.namespace}"
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = kubernetes_cluster_role.dashboard_nav_readonly.metadata[0].name
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "ServiceAccount"
|
||||
name = kubernetes_service_account.dashboard_owner[each.key].metadata[0].name
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
}
|
||||
|
||||
# Long-lived (non-expiring) token the user pastes into the dashboard login.
|
||||
resource "kubernetes_secret" "dashboard_owner_token" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "dashboard-${each.value.user_key}-token"
|
||||
namespace = each.value.namespace
|
||||
annotations = {
|
||||
"kubernetes.io/service-account.name" = kubernetes_service_account.dashboard_owner[each.key].metadata[0].name
|
||||
}
|
||||
}
|
||||
type = "kubernetes.io/service-account-token"
|
||||
wait_for_service_account_token = true
|
||||
}
|
||||
50
stacks/rbac/modules/rbac/etcd-tuning.tf
Normal file
50
stacks/rbac/modules/rbac/etcd-tuning.tf
Normal file
|
|
@ -0,0 +1,50 @@
|
|||
# Tune etcd for reduced disk writes on k8s-master.
|
||||
# Increases snapshot-count from 10000 (default) to 50000 to reduce WAL snapshot frequency.
|
||||
# etcd writes ~37.5 GB/day; less frequent snapshots reduce this by ~30-40%.
|
||||
# This patches the kubeadm-managed static pod manifest. Note: kubeadm upgrades
|
||||
# will reset this, so re-apply after any kubeadm upgrade.
|
||||
|
||||
resource "null_resource" "etcd_tuning" {
|
||||
connection {
|
||||
type = "ssh"
|
||||
user = "wizard"
|
||||
host = var.k8s_master_host
|
||||
private_key = var.ssh_private_key
|
||||
}
|
||||
|
||||
provisioner "remote-exec" {
|
||||
inline = [
|
||||
<<-SCRIPT
|
||||
sudo python3 -c "
|
||||
import yaml
|
||||
|
||||
path = '/etc/kubernetes/manifests/etcd.yaml'
|
||||
with open(path) as f:
|
||||
doc = yaml.safe_load(f)
|
||||
|
||||
container = doc['spec']['containers'][0]
|
||||
args = container['command']
|
||||
|
||||
# Update or add --snapshot-count=50000
|
||||
new_args = [a for a in args if not a.startswith('--snapshot-count=')]
|
||||
new_args.append('--snapshot-count=50000')
|
||||
|
||||
# Update or add --quota-backend-bytes (256MB, default is 2GB which is fine)
|
||||
# Keep default for now
|
||||
|
||||
container['command'] = new_args
|
||||
|
||||
with open(path, 'w') as f:
|
||||
yaml.dump(doc, f, default_flow_style=False)
|
||||
|
||||
print('etcd manifest updated: --snapshot-count=50000')
|
||||
"
|
||||
SCRIPT
|
||||
]
|
||||
}
|
||||
|
||||
# Re-run if the configuration changes
|
||||
triggers = {
|
||||
snapshot_count = "50000"
|
||||
}
|
||||
}
|
||||
318
stacks/rbac/modules/rbac/main.tf
Normal file
318
stacks/rbac/modules/rbac/main.tf
Normal file
|
|
@ -0,0 +1,318 @@
|
|||
variable "tls_secret_name" {}
|
||||
variable "tier" { type = string }
|
||||
|
||||
variable "k8s_users" {
|
||||
type = map(object({
|
||||
role = string # "admin", "power-user", "namespace-owner"
|
||||
email = string # OIDC email claim
|
||||
namespaces = optional(list(string), []) # for namespace-owners
|
||||
domains = optional(list(string), []) # subdomains for user apps
|
||||
quota = optional(object({
|
||||
cpu_requests = optional(string, "2")
|
||||
memory_requests = optional(string, "4Gi")
|
||||
memory_limits = optional(string, "8Gi")
|
||||
pods = optional(string, "20")
|
||||
}), {})
|
||||
}))
|
||||
default = {}
|
||||
}
|
||||
|
||||
# --- Admin role ---
|
||||
# Binds to built-in cluster-admin ClusterRole
|
||||
|
||||
resource "kubernetes_cluster_role_binding" "admin_users" {
|
||||
for_each = nonsensitive({ for name, user in var.k8s_users : name => user if user.role == "admin" })
|
||||
|
||||
metadata {
|
||||
name = "oidc-admin-${each.key}"
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = "cluster-admin"
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "User"
|
||||
name = each.value.email
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
}
|
||||
}
|
||||
|
||||
# --- Power-user role (read+write+secrets) — RETAINED BUT UNBOUND ---
|
||||
# Superseded by oidc-power-user-readonly (below) per ADR-0005: power-users are bound
|
||||
# to the read-only role, NOT this one. Kept defined for reference/rollback; do NOT
|
||||
# bind it without a deliberate decision (it grants cluster-wide write + secrets).
|
||||
|
||||
resource "kubernetes_cluster_role" "power_user" {
|
||||
metadata {
|
||||
name = "oidc-power-user"
|
||||
}
|
||||
|
||||
# Core resources
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["pods", "pods/log", "pods/exec", "services", "endpoints", "configmaps", "secrets", "persistentvolumeclaims", "events", "namespaces"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["pods", "services", "configmaps", "secrets", "persistentvolumeclaims"]
|
||||
verbs = ["create", "update", "patch", "delete"]
|
||||
}
|
||||
|
||||
# Apps
|
||||
rule {
|
||||
api_groups = ["apps"]
|
||||
resources = ["deployments", "statefulsets", "daemonsets", "replicasets"]
|
||||
verbs = ["get", "list", "watch", "create", "update", "patch", "delete"]
|
||||
}
|
||||
|
||||
# Batch
|
||||
rule {
|
||||
api_groups = ["batch"]
|
||||
resources = ["jobs", "cronjobs"]
|
||||
verbs = ["get", "list", "watch", "create", "update", "patch", "delete"]
|
||||
}
|
||||
|
||||
# Networking
|
||||
rule {
|
||||
api_groups = ["networking.k8s.io"]
|
||||
resources = ["ingresses", "networkpolicies"]
|
||||
verbs = ["get", "list", "watch", "create", "update", "patch", "delete"]
|
||||
}
|
||||
|
||||
# Autoscaling
|
||||
rule {
|
||||
api_groups = ["autoscaling"]
|
||||
resources = ["horizontalpodautoscalers"]
|
||||
verbs = ["get", "list", "watch", "create", "update", "patch", "delete"]
|
||||
}
|
||||
|
||||
# Read-only on cluster-level resources
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["nodes"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["storage.k8s.io"]
|
||||
resources = ["storageclasses"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["rbac.authorization.k8s.io"]
|
||||
resources = ["clusterroles", "clusterrolebindings", "roles", "rolebindings"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
}
|
||||
|
||||
# --- Power-user READ-ONLY role (ADR-0005) ---
|
||||
# Cluster-wide get/list/watch, explicitly NO secrets and NO pods/exec. This is the
|
||||
# role power-users are actually bound to (workstation tier: "cluster-wide read,
|
||||
# no Secrets"). Mirrors power_user's resource breadth minus writes/secrets/exec.
|
||||
resource "kubernetes_cluster_role" "power_user_readonly" {
|
||||
metadata {
|
||||
name = "oidc-power-user-readonly"
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["pods", "pods/log", "services", "endpoints", "configmaps", "persistentvolumeclaims", "events", "namespaces", "nodes"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["apps"]
|
||||
resources = ["deployments", "statefulsets", "daemonsets", "replicasets"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["batch"]
|
||||
resources = ["jobs", "cronjobs"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["networking.k8s.io"]
|
||||
resources = ["ingresses", "networkpolicies"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["autoscaling"]
|
||||
resources = ["horizontalpodautoscalers"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["storage.k8s.io"]
|
||||
resources = ["storageclasses"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["rbac.authorization.k8s.io"]
|
||||
resources = ["clusterroles", "clusterrolebindings", "roles", "rolebindings"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
}
|
||||
|
||||
# Power-users are bound to the READ-ONLY role above (NOT the read+write+secrets one).
|
||||
resource "kubernetes_cluster_role_binding" "power_users" {
|
||||
for_each = nonsensitive({ for name, user in var.k8s_users : name => user if user.role == "power-user" })
|
||||
|
||||
metadata {
|
||||
name = "oidc-power-user-readonly-${each.key}"
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = kubernetes_cluster_role.power_user_readonly.metadata[0].name
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "User"
|
||||
name = each.value.email
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
}
|
||||
}
|
||||
|
||||
# --- Namespace-owner role ---
|
||||
# Full admin within assigned namespaces + read-only cluster-wide
|
||||
|
||||
locals {
|
||||
# Flatten user->namespace pairs for iteration
|
||||
namespace_owner_pairs = flatten([
|
||||
for name, user in var.k8s_users : [
|
||||
for ns in user.namespaces : {
|
||||
user_key = name
|
||||
namespace = ns
|
||||
email = user.email
|
||||
quota = user.quota
|
||||
}
|
||||
] if user.role == "namespace-owner"
|
||||
])
|
||||
}
|
||||
|
||||
resource "kubernetes_role_binding" "namespace_owner" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "namespace-owner-${each.value.user_key}"
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = "admin" # Built-in ClusterRole with full namespace access
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "User"
|
||||
name = each.value.email
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
}
|
||||
}
|
||||
|
||||
# Read-only cluster-wide access for namespace owners
|
||||
resource "kubernetes_cluster_role" "namespace_owner_readonly" {
|
||||
metadata {
|
||||
name = "oidc-namespace-owner-readonly"
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["namespaces", "nodes"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = [""]
|
||||
resources = ["pods", "services", "configmaps", "events"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
|
||||
rule {
|
||||
api_groups = ["apps"]
|
||||
resources = ["deployments", "statefulsets", "daemonsets"]
|
||||
verbs = ["get", "list", "watch"]
|
||||
}
|
||||
}
|
||||
|
||||
resource "kubernetes_cluster_role_binding" "namespace_owner_readonly" {
|
||||
for_each = nonsensitive({ for name, user in var.k8s_users : name => user if user.role == "namespace-owner" })
|
||||
|
||||
metadata {
|
||||
name = "oidc-ns-owner-readonly-${each.key}"
|
||||
}
|
||||
|
||||
role_ref {
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
kind = "ClusterRole"
|
||||
name = kubernetes_cluster_role.namespace_owner_readonly.metadata[0].name
|
||||
}
|
||||
|
||||
subject {
|
||||
kind = "User"
|
||||
name = each.value.email
|
||||
api_group = "rbac.authorization.k8s.io"
|
||||
}
|
||||
}
|
||||
|
||||
# Resource quotas per user namespace
|
||||
resource "kubernetes_resource_quota" "user_namespace_quota" {
|
||||
for_each = nonsensitive({ for pair in local.namespace_owner_pairs : "${pair.user_key}-${pair.namespace}" => pair })
|
||||
|
||||
metadata {
|
||||
name = "user-quota"
|
||||
namespace = each.value.namespace
|
||||
}
|
||||
|
||||
spec {
|
||||
hard = {
|
||||
"requests.cpu" = each.value.quota.cpu_requests
|
||||
"requests.memory" = each.value.quota.memory_requests
|
||||
"limits.memory" = each.value.quota.memory_limits
|
||||
"pods" = each.value.quota.pods
|
||||
}
|
||||
}
|
||||
|
||||
depends_on = [kubernetes_role_binding.namespace_owner]
|
||||
}
|
||||
|
||||
# ConfigMap with user-role mapping for the self-service portal
|
||||
resource "kubernetes_config_map" "user_roles" {
|
||||
metadata {
|
||||
name = "k8s-user-roles"
|
||||
namespace = "k8s-portal"
|
||||
}
|
||||
|
||||
data = {
|
||||
"users.json" = jsonencode({
|
||||
for name, user in var.k8s_users : user.email => {
|
||||
role = user.role
|
||||
namespaces = user.namespaces
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
# TLS secret in each user namespace (so they can create HTTPS ingresses)
|
||||
module "user_namespace_tls" {
|
||||
for_each = nonsensitive(toset(flatten([
|
||||
for name, user in var.k8s_users : user.namespaces
|
||||
if user.role == "namespace-owner"
|
||||
])))
|
||||
|
||||
source = "../../../../modules/kubernetes/setup_tls_secret"
|
||||
namespace = each.value
|
||||
tls_secret_name = var.tls_secret_name
|
||||
}
|
||||
1
stacks/rbac/secrets
Symbolic link
1
stacks/rbac/secrets
Symbolic link
|
|
@ -0,0 +1 @@
|
|||
../../secrets
|
||||
8
stacks/rbac/terragrunt.hcl
Normal file
8
stacks/rbac/terragrunt.hcl
Normal file
|
|
@ -0,0 +1,8 @@
|
|||
include "root" {
|
||||
path = find_in_parent_folders()
|
||||
}
|
||||
|
||||
dependency "infra" {
|
||||
config_path = "../infra"
|
||||
skip_outputs = true
|
||||
}
|
||||
Loading…
Add table
Add a link
Reference in a new issue