Merge pull request #911 from bogdando/DROP_CAPS

Drop linux capabilities and rework users/groups

docs/security.md

@@ -0,0 +1,31 @@
+Users and groups
+================
+
+There are following users and groups defined by the addusers role:
+
+* Kube user, group from the ``kubelet_user`` and ``kubelet_group`` vars.
+* Etcd user, group from the ``etcd_user`` and ``etcd_group`` vars.
+* Network plugin user, group from the ``netplug_user`` and ``netplug_group`` vars.
+
+There are additional certificate access groups for kube and etcd users defined.
+For example, kubelet and network plugins require read access to the
+etcd certs and keys. This is defined via the corresponding ``etcd_cert_group``
+var. Members of that group (defaults to `kube` and `netplug` users) will read
+etcd secret keys and certs. Same applies to the ``kube_cert_group``
+(defaults to `kube` user) members. You may want to share kube certs via that
+group with bastion proxies or the like.
+
+Linux capabilites
+=================
+
+Kargo allows to control dropped Linux capabilities for unprivileged docker
+containers it configures for deployments. For examle, etcd or some networking
+related systemd units or k8s workloads, like kubedns, dnsmasq or netchecker apps.
+
+Dropped capabilites are represented by the ``apps_drop_cap``, ``dnsmasq_drop_cap``,
+``etcd_drop_cap``, ``calico_drop_cap``  vars.
+
+Be carefull changing defaults - different kube components and k8s apps might
+expect specific capabilities to be present and can only run as root! Also note
+that kublet, kube-proxy and network plugins require privileged mode and ignore
+dropped capabilities.

inventory/group_vars/all.yml

@@ -36,10 +36,19 @@ retry_stagger: 5
 # Directory where python binary is installed
 # ansible_python_interpreter: "/opt/bin/python"
 
-# This is the group that the cert creation scripts chgrp the
-# cert files to. Not really changable...
+# This is the users/groups to own files and groups that the cert creation
+# scripts chgrp the cert files to
+kubelet_user: kube
+kubelet_group: kube
 kube_cert_group: kube-cert
 
+netplug_user: netplug
+netplug_group: netplug
+
+etcd_user: etcd
+etcd_group: etcd
+etcd_cert_group: etcd-cert
+
 # Cluster Loglevel configuration
 kube_log_level: 2
 

roles/adduser/defaults/main.yml

@@ -1,24 +1,40 @@
 ---
 addusers:
   etcd:
-    name: etcd
+    name: "{{ etcd_user }}"
     comment: "Etcd user"
-    createhome: yes
+    createhome: >-
+      {% if ansible_os_family in ["CoreOS", "Container Linux by CoreOS"] %}no{% else %}yes{% endif %}
     home: "/var/lib/etcd"
     system: yes
-    shell: /bin/nologin
+    shell: /usr/sbin/nologin
+    group: "{{ etcd_group }}"
+    groups: "{{ etcd_cert_group }}"
+    type: >-
+      {% if ansible_os_family in ["CoreOS", "Container Linux by CoreOS"] %}cloud-init{% endif %}
   kube:
-    name: kube
+    name: "{{ kubelet_user }}"
     comment: "Kubernetes user"
-    shell: /sbin/nologin
+    shell: /usr/sbin/nologin
     system: yes
-    group: "{{ kube_cert_group }}"
+    group: "{{ kubelet_group }}"
+    groups: "{{ etcd_cert_group }},{{ kube_cert_group }}"
     createhome: no
+  netplug:
+    name: "{{ netplug_user }}"
+    comment: "Network plugin user"
+    createhome: no
+    system: yes
+    shell: /usr/sbin/nologin
+    group: "{{ netplug_group }}"
+    groups: "{{ etcd_cert_group }}"
 
 adduser:
   name: "{{ user.name }}"
   group: "{{ user.name|default(None) }}"
+  groups: "{{ user.groups|default(None) }}"
   comment: "{{ user.comment|default(None) }}"
   shell: "{{ user.shell|default(None) }}"
   system: "{{ user.system|default(None) }}"
   createhome: "{{ user.createhome|default(None) }}"
+  type: "{{ user.type|default(None) }}"

roles/adduser/handlers/main.yml

@@ -0,0 +1,4 @@
+---
+- name:  User | update users for cloud-init
+  command: /usr/bin/coreos-cloudinit --from-file /etc/{{ user.name }}_user_cloud_init_conf
+  when: ansible_os_family in ["CoreOS", "Container Linux by CoreOS"]

roles/adduser/tasks/main.yml

@@ -1,13 +1,35 @@
 ---
+- name: User | Create Certificate Access Groups
+  group: name={{ item }} system=yes
+  with_items: "{{ user.groups.split(',') }}"
+
 - name: User | Create User Group
   group: name={{user.group|default(user.name)}} system={{user.system|default(omit)}}
 
+- name: User | Create cloud-init user
+  template:
+    dest: /etc/{{ user.name }}_user_cloud_init_conf
+    src: users.j2
+    owner: root
+    mode: 0640
+  notify: User | update users for cloud-init
+  when: "{{ user.type|default('standard') == 'cloud-init' }}"
+
+- meta: flush_handlers
+
+- name: User | Hack groups for existing cloud-init users CoreOS
+  command: /usr/sbin/usermod -aG {{ item }} {{ user.name }}
+  with_items: "{{ user.groups.split(',') }}"
+  when: "{{ ansible_os_family in ['CoreOS', 'Container Linux by CoreOS'] and user.type|default('standard') == 'cloud-init' }}"
+
 - name: User | Create User
   user:
     comment: "{{user.comment|default(omit)}}"
-    createhome: "{{user.create_home|default(omit)}}"
+    createhome: "{{user.createhome|default(omit)}}"
     group: "{{user.group|default(user.name)}}"
+    groups: "{{user.groups|default(omit)}}"
     home: "{{user.home|default(omit)}}"
     shell: "{{user.shell|default(omit)}}"
     name: "{{user.name}}"
     system: "{{user.system|default(omit)}}"
+  when: "{{ user.type|default('standard') != 'cloud-init' }}"

roles/adduser/templates/users.j2

@@ -0,0 +1,15 @@
+#cloud-config
+users:
+- name: {{ user.name  }}
+  gecos: {{ user.comment }}
+  system: {{ user.system|bool }}
+  no-log-init: {{ user.system|bool }}
+  primary-group: {{ user.group }}
+  no-create-home: {{ not user.createhome|bool }}
+  homedir: {{ user.home }}
+  shell: {{ user.shell }}
+  groups: |
+    {% for g in user.groups.split(',') %}
+    - {{ g }}
+    {% endfor %}
+    #

roles/adduser/vars/coreos.yml

@@ -1,8 +0,0 @@
----
-addusers:
-  - name: kube
-    comment: "Kubernetes user"
-    shell: /sbin/nologin
-    system: yes
-    group: "{{ kube_cert_group }}"
-    createhome: no

roles/adduser/vars/debian.yml

@@ -1,15 +0,0 @@
----
-addusers:
-  - name: etcd
-    comment: "Etcd user"
-    createhome: yes
-    home: "/var/lib/etcd"
-    system: yes
-    shell: /bin/nologin
-
-  - name: kube
-    comment: "Kubernetes user"
-    shell: /sbin/nologin
-    system: yes
-    group: "{{ kube_cert_group }}"
-    createhome: no

roles/adduser/vars/redhat.yml

@@ -1,15 +0,0 @@
----
-addusers:
-  - name: etcd
-    comment: "Etcd user"
-    createhome: yes
-    home: "/var/lib/etcd"
-    system: yes
-    shell: /bin/nologin
-
-  - name: kube
-    comment: "Kubernetes user"
-    shell: /sbin/nologin
-    system: yes
-    group: "{{ kube_cert_group }}"
-    createhome: no

roles/dnsmasq/defaults/main.yml

@@ -26,3 +26,16 @@ dns_cpu_limit: 100m
 dns_memory_limit: 170Mi
 dns_cpu_requests: 70m
 dns_memory_requests: 70Mi
+
+# Linux capabilities to be dropped for dnsmasq k8s app ran container engines
+dnsmasq_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setpcap
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap

roles/dnsmasq/templates/dnsmasq-ds.yml

@@ -26,6 +26,10 @@ spec:
             capabilities:
               add:
                 - NET_ADMIN
+              drop:
+{% for c in dnsmasq_drop_cap %}
+                - {{ c.upper() }}
+{% endfor %}
           imagePullPolicy: IfNotPresent
           resources:
             limits:

roles/etcd/defaults/main.yml

@@ -3,10 +3,26 @@ etcd_bin_dir: "{{ local_release_dir }}/etcd/etcd-{{ etcd_version }}-linux-amd64/
 
 etcd_config_dir: /etc/ssl/etcd
 etcd_cert_dir: "{{ etcd_config_dir }}/ssl"
-etcd_cert_group: root
 
 etcd_script_dir: "{{ bin_dir }}/etcd-scripts"
 
+# Linux capabilities to be dropped for container engines
+etcd_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setgid
+  - setuid
+  - setpcap
+  - net_bind_service
+  - net_raw
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap
+
 # Limits
 etcd_memory_limit: 512M
 etcd_cpu_limit: 300m

roles/etcd/files/make-ssl-etcd.sh

@@ -94,5 +94,8 @@ if [ -n "$HOSTS" ]; then
     done
 fi
 
+# Grant the group read access
+chmod g+r *.pem
+
 # Install certs
 mv *.pem ${SSLDIR}/

roles/etcd/meta/main.yml

@@ -2,7 +2,7 @@
 dependencies:
   - role: adduser
     user: "{{ addusers.etcd }}"
-    when: not ansible_os_family in ['CoreOS', 'Container Linux by CoreOS']
+    tags: bootstrap-os
   - role: download
     file: "{{ downloads.etcd }}"
     tags: download

roles/etcd/tasks/gen_certs.yml

@@ -4,14 +4,15 @@
     path={{ etcd_cert_dir }}
     group={{ etcd_cert_group }}
     state=directory
-    owner=root
+    mode=0750
+    owner={{ etcd_user }}
     recurse=yes
 
 - name: "Gen_certs | create etcd script dir (on {{groups['etcd'][0]}})"
   file:
     path: "{{ etcd_script_dir }}"
     state: directory
-    owner: root
+    owner: "{{ etcd_user }}"
   run_once: yes
   delegate_to: "{{groups['etcd'][0]}}"
 
@@ -20,7 +21,8 @@
     path={{ etcd_cert_dir }}
     group={{ etcd_cert_group }}
     state=directory
-    owner=root
+    mode=0750
+    owner={{ etcd_user }}
     recurse=yes
   run_once: yes
   delegate_to: "{{groups['etcd'][0]}}"
@@ -124,12 +126,12 @@
     path={{ etcd_cert_dir }}
     group={{ etcd_cert_group }}
     state=directory
-    owner=kube
+    owner={{ etcd_user }}
     recurse=yes
   tags: facts
 
-- name: Gen_certs | set permissions on keys
-  shell: chmod 0600 {{ etcd_cert_dir}}/*key.pem
+- name: Gen_certs | set shared group permissions on keys
+  shell: chmod 0640 {{ etcd_cert_dir}}/*.pem
   when: inventory_hostname in groups['etcd']
   changed_when: false
 

roles/etcd/tasks/main.yml

@@ -1,6 +1,8 @@
 ---
 - include: pre_upgrade.yml
   tags: etcd-pre-upgrade
+- include: set_facts.yml
+  tags: [bootstrap-os, facts]
 - include: check_certs.yml
   tags: [etcd-secrets, facts]
 - include: gen_certs.yml

roles/etcd/tasks/pre_upgrade.yml

@@ -1,3 +1,4 @@
+---
 - name: "Pre-upgrade | check for etcd-proxy unit file"
   stat:
     path: /etc/systemd/system/etcd-proxy.service
@@ -49,3 +50,7 @@
     awk -F"[: =]" '{print "{{ bin_dir }}/etcdctl --peers={{ etcd_access_addresses | regex_replace('https','http') }} member update "$1" https:"$7":"$8}' | bash
   run_once: true
   when: 'etcd_member_list.rc == 0 and "http://" in etcd_member_list.stdout'
+
+- name: "Pre-upgrade | share access to etcd certs for its users"
+  shell: chmod g+r {{ etcd_cert_dir }}/*.pem
+  failed_when: false

roles/etcd/tasks/set_facts.yml

@@ -0,0 +1,17 @@
+---
+- name: Etcd | get etcd user ID
+  shell: /usr/bin/id -u {{ etcd_user }} || echo 0
+  register: etcd_uid
+
+- name: Etcd | get etcd group ID
+  shell: /usr/bin/getent group {{ etcd_group }} | cut -d':' -f3 || echo 0
+  register: etcd_gid
+
+- name: Etcd | get etcd cert group ID
+  shell: /usr/bin/getent group {{ etcd_cert_group }} | cut -d':' -f3 || echo 0
+  register: etcd_cert_gid
+
+- set_fact:
+    etcd_user_id: "{{ etcd_uid.stdout }}"
+    etcd_group_id: "{{ etcd_gid.stdout }}"
+    etcd_cert_group_id: "{{ etcd_cert_gid.stdout }}"

roles/etcd/templates/etcd-docker.service.j2

@@ -14,8 +14,12 @@ ExecStart={{ docker_bin_dir }}/docker run --restart=on-failure:5 \
 -v /etc/ssl/certs:/etc/ssl/certs:ro \
 -v {{ etcd_cert_dir }}:{{ etcd_cert_dir }}:ro \
 -v /var/lib/etcd:/var/lib/etcd:rw \
+{% for c in etcd_drop_cap %}
+--cap-drop={{ c }} \
+{% endfor %}
 --memory={{ etcd_memory_limit|regex_replace('Mi', 'M') }} --cpu-shares={{ etcd_cpu_limit|regex_replace('m', '') }} \
 --name={{ etcd_member_name | default("etcd") }} \
+-u {{ etcd_user_id }}:{{ etcd_group_id }} --group-add {{ etcd_cert_group_id }} \
 {{ etcd_image_repo }}:{{ etcd_image_tag }} \
 {% if etcd_after_v3 %}
 {{ etcd_container_bin_dir }}etcd

roles/etcd/templates/etcd-rkt.service.j2

@@ -8,6 +8,9 @@ Restart=on-failure
 RestartSec=10s
 TimeoutStartSec=0
 LimitNOFILE=40000
+User=root
+Group={{ etcd_group_id }}
+SupplementaryGroups={{ etcd_cert_group_id }}
 
 ExecStart=/usr/bin/rkt run \
 --uuid-file-save=/var/run/etcd.uuid \
@@ -20,6 +23,11 @@ ExecStart=/usr/bin/rkt run \
 --set-env-file=/etc/etcd.env \
 --stage1-from-dir=stage1-fly.aci \
 {{ etcd_image_repo }}:{{ etcd_image_tag }} \
+{% for c in etcd_drop_cap %}
+--caps-remove=CAP_{{ c.upper() }} \
+{% endfor %}
+--memory={{ etcd_memory_limit }} --cpu={{ etcd_cpu_limit }} \
+--user={{ etcd_user_id }} --group={{ etcd_group_id }} \
 --name={{ etcd_member_name | default("etcd") }}
 
 ExecStartPre=-/usr/bin/rkt rm --uuid-file=/var/run/etcd.uuid

roles/kubernetes-apps/ansible/defaults/main.yml

@@ -51,3 +51,18 @@ netchecker_kubectl_memory_requests: 64M
 etcd_cert_dir: "/etc/ssl/etcd/ssl"
 calico_cert_dir: "/etc/calico/certs"
 canal_cert_dir: "/etc/canal/certs"
+
+# Linux capabilities to be dropped for k8s apps ran by container engines
+apps_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setgid
+  - setuid
+  - setpcap
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap

roles/kubernetes-apps/ansible/templates/calico-policy-controller.yml.j2

@@ -25,6 +25,12 @@ spec:
         - name: calico-policy-controller
           image: {{ calico_policy_image_repo }}:{{ calico_policy_image_tag }}
           imagePullPolicy: {{ k8s_image_pull_policy }}
+          securityContext:
+            capabilities:
+              drop:
+{% for c in apps_drop_cap %}
+                - {{ c.upper() }}
+{% endfor %}
           resources:
             limits:
               cpu: {{ calico_policy_controller_cpu_limit }}

roles/kubernetes-apps/ansible/templates/netchecker-agent-ds.yml

@@ -23,6 +23,12 @@ spec:
             - name: REPORT_INTERVAL
               value: '{{ agent_report_interval }}'
           imagePullPolicy: {{ k8s_image_pull_policy }}
+          securityContext:
+            capabilities:
+              drop:
+{% for c in apps_drop_cap %}
+                - {{ c.upper() }}
+{% endfor %}
           resources:
             limits:
               cpu: {{ netchecker_agent_cpu_limit }}

roles/kubernetes-apps/ansible/templates/netchecker-agent-hostnet-ds.yml

@@ -24,6 +24,12 @@ spec:
             - name: REPORT_INTERVAL
               value: '{{ agent_report_interval }}'
           imagePullPolicy: {{ k8s_image_pull_policy }}
+          securityContext:
+            capabilities:
+              drop:
+{% for c in apps_drop_cap %}
+                - {{ c.upper() }}
+{% endfor %}
           resources:
             limits:
               cpu: {{ netchecker_agent_cpu_limit }}

roles/kubernetes-apps/ansible/templates/netchecker-server-pod.yml

@@ -33,3 +33,9 @@ spec:
           memory: {{ netchecker_kubectl_memory_requests }}
       args:
         - proxy
+      securityContext:
+        capabilities:
+          drop:
+{% for c in apps_drop_cap %}
+            - {{ c.upper() }}
+{% endfor %}

roles/kubernetes/master/defaults/main.yml

@@ -13,6 +13,21 @@ kube_apiserver_node_port_range: "30000-32767"
 etcd_config_dir: /etc/ssl/etcd
 etcd_cert_dir: "{{ etcd_config_dir }}/ssl"
 
+# Linux capabilities to be dropped for k8s apps ran by container engines
+apps_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setgid
+  - setuid
+  - setpcap
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap
+
 # Limits for kube components
 kube_controller_memory_limit: 512M
 kube_controller_cpu_limit: 250m

roles/kubernetes/master/tasks/main.yml

@@ -2,6 +2,9 @@
 - include: pre-upgrade.yml
   tags: k8s-pre-upgrade
 
+- include: set_facts.yml
+  tags: facts
+
 - name: Copy kubectl from hyperkube container
   command: "{{ docker_bin_dir }}/docker run --rm -v {{ bin_dir }}:/systembindir {{ hyperkube_image_repo }}:{{ hyperkube_image_tag }} /bin/cp /hyperkube /systembindir/kubectl"
   register: kube_task_result

roles/kubernetes/master/tasks/set_facts.yml

@@ -0,0 +1,22 @@
+---
+- name: Master | get kube user ID
+  shell: /usr/bin/id -u {{ kubelet_user }} || echo 0
+  register: kube_uid
+
+- name: Master | get kube group ID
+  shell: /usr/bin/getent group {{ kubelet_group }} | cut -d':' -f3 || echo 0
+  register: kube_gid
+
+- name: Master | get kube cert group ID
+  shell: /usr/bin/getent group {{ kube_cert_group }} | cut -d':' -f3 || echo 0
+  register: kube_cert_gid
+
+- name: Master | get etcd cert group ID
+  shell: /usr/bin/getent group {{ etcd_cert_group }} | cut -d':' -f3 || echo 0
+  register: etcd_cert_gid
+
+- set_fact:
+    kubelet_user_id: "{{ kube_uid.stdout }}"
+    kubelet_group_id: "{{ kube_gid.stdout }}"
+    kube_cert_group_id: "{{ kube_cert_gid.stdout }}"
+    etcd_cert_group_id: "{{ etcd_cert_gid.stdout }}"

roles/kubernetes/master/templates/manifests/kube-apiserver.manifest.j2

@@ -12,6 +12,14 @@ spec:
   - name: kube-apiserver
     image: {{ hyperkube_image_repo }}:{{ hyperkube_image_tag }}
     imagePullPolicy: {{ k8s_image_pull_policy }}
+    securityContext:
+      capabilities:
+        drop:
+{% for c in apps_drop_cap %}
+          - {{ c.upper() }}
+{% endfor %}
+        add:
+          - DAC_OVERRIDE
     resources:
       limits:
         cpu: {{ kube_apiserver_cpu_limit }}

roles/kubernetes/master/templates/manifests/kube-controller-manager.manifest.j2

@@ -11,6 +11,17 @@ spec:
   - name: kube-controller-manager
     image: {{ hyperkube_image_repo }}:{{ hyperkube_image_tag }}
     imagePullPolicy: {{ k8s_image_pull_policy }}
+    securityContext:
+      runAsUser: {{ kubelet_user_id }}
+      fsGroup: {{ kubelet_group_id }}
+      supplementalGroups:
+        - {{ kube_cert_group_id }}
+        - {{ etcd_cert_group_id }}
+      capabilities:
+        drop:
+{% for c in apps_drop_cap %}
+          - {{ c.upper() }}
+{% endfor %}
     resources:
       limits:
         cpu: {{ kube_controller_cpu_limit }}

roles/kubernetes/master/templates/manifests/kube-scheduler.manifest.j2

@@ -11,6 +11,17 @@ spec:
   - name: kube-scheduler
     image: {{ hyperkube_image_repo }}:{{ hyperkube_image_tag }}
     imagePullPolicy: {{ k8s_image_pull_policy }}
+    securityContext:
+      runAsUser: {{ kubelet_user_id }}
+      fsGroup: {{ kubelet_group_id }}
+      supplementalGroups:
+        - {{ kube_cert_group_id }}
+        - {{ etcd_cert_group_id }}
+      capabilities:
+        drop:
+{% for c in apps_drop_cap %}
+          - {{ c.upper() }}
+{% endfor %}
     resources:
       limits:
         cpu: {{ kube_scheduler_cpu_limit }}

roles/kubernetes/node/defaults/main.yml

@@ -29,3 +29,18 @@ nginx_image_repo: nginx
 nginx_image_tag: 1.11.4-alpine
 
 etcd_config_dir: /etc/ssl/etcd
+
+# Linux capabilities to be dropped for container engines
+apps_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setgid
+  - setuid
+  - setpcap
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap

roles/kubernetes/node/tasks/install.yml

@@ -26,6 +26,6 @@
   notify: restart kubelet
 
 - name: install | Install kubelet launch script
-  template: src=kubelet-container.j2 dest="{{ bin_dir }}/kubelet" owner=kube mode=0755 backup=yes
+  template: src=kubelet-container.j2 dest="{{ bin_dir }}/kubelet" owner={{ kubelet_user }} mode=0755 backup=yes
   notify: restart kubelet
   when: kubelet_deployment_type == "docker"

roles/kubernetes/node/tasks/main.yml

@@ -4,6 +4,9 @@
       {%- if inventory_hostname in groups['kube-master'] and inventory_hostname not in groups['kube-node'] -%}true{%- else -%}false{%- endif -%}
   tags: facts
 
+- include: pre-upgrade.yml
+  tags: k8s-pre-upgrade
+
 - include: install.yml
   tags: kubelet
 

roles/kubernetes/node/tasks/pre-upgrade.yml

@@ -0,0 +1,4 @@
+---
+- name: "Pre-upgrade | share access to kube certs for its users"
+  shell: chmod g+r {{ kube_cert_dir }}/*.pem
+  failed_when: false

roles/kubernetes/node/templates/kubelet.rkt.service.j2

@@ -44,6 +44,7 @@ ExecStart=/usr/bin/rkt run \
         --mount volume=var-log,target=/var/log \
         --stage1-from-dir=stage1-fly.aci \
         {{ hyperkube_image_repo }}:{{ hyperkube_image_tag }} \
+        --memory={{ kubelet_memory_limit }} --cpu={{ kubelet_cpu_limit }} \
         --uuid-file-save=/var/run/kubelet.uuid \
         --debug --exec=/kubelet -- \
                 $KUBE_LOGTOSTDERR \

roles/kubernetes/preinstall/meta/main.yml

@@ -2,4 +2,7 @@
 dependencies:
   - role: adduser
     user: "{{ addusers.kube }}"
-    tags: kubelet
+    tags: [bootstrap-os, kubelet]
+  - role: adduser
+    user: "{{ addusers.netplug }}"
+    tags: [bootstrap-os, network]

roles/kubernetes/preinstall/tasks/main.yml

@@ -23,6 +23,12 @@
 - include: set_facts.yml
   tags: facts
 
+- include: set_resolv_facts.yml
+  tags: [bootstrap-os, resolvconf, facts]
+
+- include: set_uid_facts.yml
+  tags: [bootstrap-os, facts]
+
 - name: gather os specific variables
   include_vars: "{{ item }}"
   with_first_found:
@@ -42,7 +48,7 @@
   file:
     path: "{{ kube_config_dir }}"
     state: directory
-    owner: kube
+    owner: "{{ kubelet_user }}"
   when: "{{ inventory_hostname in groups['k8s-cluster'] }}"
   tags: [kubelet, k8s-secrets, kube-controller-manager, kube-apiserver, bootstrap-os, apps, network, master, node]
 
@@ -50,7 +56,7 @@
   file:
     path: "{{ kube_script_dir }}"
     state: directory
-    owner: kube
+    owner: "{{ kubelet_user }}"
   when: "{{ inventory_hostname in groups['k8s-cluster'] }}"
   tags: [k8s-secrets, bootstrap-os]
 
@@ -58,7 +64,7 @@
   file:
     path: "{{ kube_manifest_dir }}"
     state: directory
-    owner: kube
+    owner: "{{ kubelet_user }}"
   when: "{{ inventory_hostname in groups['k8s-cluster'] }}"
   tags: [kubelet, bootstrap-os, master, node]
 
@@ -80,7 +86,7 @@
   file:
     path: "{{ item }}"
     state: directory
-    owner: kube
+    owner: "{{ kubelet_user }}"
   with_items:
     - "/etc/cni/net.d"
     - "/opt/cni/bin"

roles/kubernetes/preinstall/tasks/set_facts.yml

@@ -51,6 +51,3 @@
     etcd_container_bin_dir: "{% if etcd_after_v3 %}/usr/local/bin/{% else %}/{% endif %}"
 - set_fact:
     peer_with_calico_rr: "{{ 'calico-rr' in groups and groups['calico-rr']|length > 0 }}"
-
-- include: set_resolv_facts.yml
-  tags: [bootstrap-os, resolvconf, facts]

roles/kubernetes/preinstall/tasks/set_uid_facts.yml

@@ -0,0 +1,32 @@
+---
+- name: Preinstall | get kube user ID
+  shell: /usr/bin/id -u {{ kubelet_user }} || echo 0
+  register: kube_uid
+
+- name: Preinstall | get kube group ID
+  shell: /usr/bin/id -g {{ kubelet_group }} || echo 0
+  register: kube_gid
+
+- name: Preinstall | get kube cert group ID
+  shell: /usr/bin/id -g {{ kube_cert_group }} || echo 0
+  register: kube_cert_gid
+
+- name: Preinstall | get etcd cert group ID
+  shell: /usr/bin/id -g {{ etcd_cert_group }} || echo 0
+  register: etcd_cert_gid
+
+- name: Preinstall | get netplug user ID
+  shell: /usr/bin/id -u {{ netplug_user }} || echo 0
+  register: netplug_uid
+
+- name: Preinstall | get netplug group ID
+  shell: /usr/bin/getent group {{ netplug_group }} | cut -d':' -f3 || echo 0
+  register: netplug_gid
+
+- set_fact:
+    kubelet_user_id: "{{ kube_uid.stdout }}"
+    kubelet_group_id: "{{ kube_gid.stdout }}"
+    kube_cert_group_id: "{{ kube_cert_gid.stdout }}"
+    etcd_cert_group_id: "{{ etcd_cert_gid.stdout }}"
+    netplug_user_id: "{{ netplug_uid.stdout }}"
+    netplug_group_id: "{{ netplug_gid.stdout }}"

roles/kubernetes/secrets/files/make-ssl.sh

@@ -101,5 +101,8 @@ if [ -n "$HOSTS" ]; then
     done
 fi
 
+# Grant the group read access
+chmod g+r *.pem
+
 # Install certs
 mv *.pem ${SSLDIR}/

roles/kubernetes/secrets/tasks/gen_certs.yml

@@ -140,11 +140,11 @@
   file:
     path={{ kube_cert_dir }}
     group={{ kube_cert_group }}
-    owner=kube
+    owner={{ kubelet_user }}
     recurse=yes
 
-- name: Gen_certs | set permissions on keys
-  shell: chmod 0600 {{ kube_cert_dir}}/*key.pem
+- name: Gen_certs | set shared group permissions on keys
+  shell: chmod 0640 {{ kube_cert_dir}}/*.pem
   when: inventory_hostname in groups['kube-master']
   changed_when: false
 

roles/kubernetes/secrets/tasks/main.yml

@@ -9,6 +9,7 @@
     path={{ kube_cert_dir }}
     state=directory
     mode=o-rwx
+    owner={{ kubelet_user }}
     group={{ kube_cert_group }}
 
 - name: Make sure the tokens directory exits
@@ -16,14 +17,16 @@
     path={{ kube_token_dir }}
     state=directory
     mode=o-rwx
-    group={{ kube_cert_group }}
+    owner={{ kubelet_user }}
+    group={{ kubelet_group }}
 
 - name: Make sure the users directory exits
   file:
     path={{ kube_users_dir }}
     state=directory
     mode=o-rwx
-    group={{ kube_cert_group }}
+    owner={{ kubelet_user }}
+    group={{ kubelet_group }}
 
 - name: Populate users for basic auth in API
   lineinfile:

roles/network_plugin/calico/defaults/main.yml

@@ -20,6 +20,23 @@ global_as_num: "64512"
 # defaults. The value should be a number, not a string.
 # calico_mtu: 1500
 
+# Linux capabilities to be dropped for container engines
+calico_drop_cap:
+  - chown
+  - dac_override
+  - fowner
+  - fsetid
+  - kill
+  - setgid
+  - setuid
+  - setpcap
+  - net_bind_service
+  - net_raw
+  - sys_chroot
+  - mknod
+  - audit_write
+  - setfcap
+
 # Limits for apps
 calico_node_memory_limit: 500M
 calico_node_cpu_limit: 300m

roles/network_plugin/calico/rr/tasks/main.yml

@@ -12,8 +12,8 @@
     dest: "{{ calico_cert_dir }}"
     state: directory
     mode: 0750
-    owner: root
-    group: root
+    owner: "{{ netplug_user }}"
+    group: "{{ netplug_group }}"
 
 - name: Calico-rr | Link etcd certificates for calico-node
   file:
@@ -31,8 +31,8 @@
     path: /var/log/calico-rr
     state: directory
     mode: 0755
-    owner: root
-    group: root
+    owner: "{{ netplug_user }}"
+    group: "{{ netplug_group }}"
 
 - name: Calico-rr | Write calico-rr.env for systemd init file
   template: src=calico-rr.env.j2 dest=/etc/calico/calico-rr.env

roles/network_plugin/calico/rr/templates/calico-rr.service.j2

@@ -6,7 +6,7 @@ Requires=docker.service
 [Service]
 EnvironmentFile=/etc/calico/calico-rr.env
 ExecStartPre=-{{ docker_bin_dir }}/docker rm -f calico-rr
-ExecStart={{ docker_bin_dir }}/docker run --net=host --privileged \
+ExecStart={{ docker_bin_dir }}/docker run --net=host \
  --name=calico-rr \
  -e IP=${IP} \
  -e IP6=${IP6} \
@@ -16,6 +16,10 @@ ExecStart={{ docker_bin_dir }}/docker run --net=host --privileged \
  -e ETCD_KEY_FILE=${ETCD_KEY_FILE} \
  -v /var/log/calico-rr:/var/log/calico \
  -v {{ calico_cert_dir }}:{{ calico_cert_dir }}:ro \
+{% for c in calico_drop_cap %}
+ --cap-drop={{ c }} \
+{% endfor %}
+ -u {{ netplug_user_id }}:{{ netplug_group_id }} --group-add {{ etcd_cert_group }} \
  --memory={{ calico_rr_memory_limit|regex_replace('Mi', 'M') }} --cpu-shares={{ calico_rr_cpu_limit|regex_replace('m', '') }} \
  {{ calico_rr_image_repo }}:{{ calico_rr_image_tag }}
 

roles/network_plugin/calico/tasks/main.yml

@@ -9,15 +9,16 @@
   template:
     src: "cni-calico.conf.j2"
     dest: "/etc/cni/net.d/10-calico.conf"
-    owner: kube
+    owner: "{{ kubelet_user }}"
+    group: "{{ kubelet_group }}"
 
 - name: Calico | Create calico certs directory
   file:
     dest: "{{ calico_cert_dir }}"
     state: directory
     mode: 0750
-    owner: root
-    group: root
+    owner: "{{ netplug_user }}"
+    group: "{{ netplug_group }}"
 
 - name: Calico | Link etcd certificates for calico-node
   file:

roles/network_plugin/canal/tasks/main.yml

@@ -3,15 +3,16 @@
   template:
     src: "cni-canal.conf.j2"
     dest: "/etc/cni/net.d/10-canal.conf"
-    owner: kube
+    owner: "{{ kubelet_user }}"
+    group: "{{ kubelet_group }}"
 
 - name: Canal | Create canal certs directory
   file:
     dest: "{{ canal_cert_dir }}"
     state: directory
     mode: 0750
-    owner: root
-    group: root
+    owner: "{{ netplug_user }}"
+    group: "{{ netplug_group }}"
 
 - name: Canal | Link etcd certificates for canal-node
   file: