Combine openshift_node and openshift_node_dnsmasq

This commit combines these two roles.  This will
prevent openshift_node_facts from running twice.

11 files changed, 251 insertions, 3 deletions

diff --git a/roles/openshift_node/defaults/main.yml b/roles/openshift_node/defaults/main.yml
index 62208c155..5a0c09f5c 100644
--- a/roles/openshift_node/defaults/main.yml
+++ b/roles/openshift_node/defaults/main.yml
@@ -1,6 +1,13 @@
 ---
 openshift_node_debug_level: "{{ debug_level | default(2) }}"
 
+openshift_node_dnsmasq_install_network_manager_hook: true
+
+# lo must always be present in this list or dnsmasq will conflict with
+# the node's dns service.
+openshift_node_dnsmasq_except_interfaces:
+- lo
+
 r_openshift_node_firewall_enabled: "{{ os_firewall_enabled | default(True) }}"
 r_openshift_node_use_firewalld: "{{ os_firewall_use_firewalld | default(False) }}"
 
diff --git a/roles/openshift_node/files/networkmanager/99-origin-dns.sh b/roles/openshift_node/files/networkmanager/99-origin-dns.sh
new file mode 100755
index 000000000..f4e48b5b7
--- /dev/null
+++ b/roles/openshift_node/files/networkmanager/99-origin-dns.sh
@@ -0,0 +1,128 @@
+#!/bin/bash -x
+# -*- mode: sh; sh-indentation: 2 -*-
+
+# This NetworkManager dispatcher script replicates the functionality of
+# NetworkManager's dns=dnsmasq  however, rather than hardcoding the listening
+# address and /etc/resolv.conf to 127.0.0.1 it pulls the IP address from the
+# interface that owns the default route. This enables us to then configure pods
+# to use this IP address as their only resolver, where as using 127.0.0.1 inside
+# a pod would fail.
+#
+# To use this,
+# - If this host is also a master, reconfigure master dnsConfig to listen on
+#   8053 to avoid conflicts on port 53 and open port 8053 in the firewall
+# - Drop this script in /etc/NetworkManager/dispatcher.d/
+# - systemctl restart NetworkManager
+# - Configure node-config.yaml to set dnsIP: to the ip address of this
+#   node
+#
+# Test it:
+# host kubernetes.default.svc.cluster.local
+# host google.com
+#
+# TODO: I think this would be easy to add as a config option in NetworkManager
+# natively, look at hacking that up
+
+cd /etc/sysconfig/network-scripts
+. ./network-functions
+
+[ -f ../network ] && . ../network
+
+if [[ $2 =~ ^(up|dhcp4-change|dhcp6-change)$ ]]; then
+  # If the origin-upstream-dns config file changed we need to restart
+  NEEDS_RESTART=0
+  UPSTREAM_DNS='/etc/dnsmasq.d/origin-upstream-dns.conf'
+  # We'll regenerate the dnsmasq origin config in a temp file first
+  UPSTREAM_DNS_TMP=`mktemp`
+  UPSTREAM_DNS_TMP_SORTED=`mktemp`
+  CURRENT_UPSTREAM_DNS_SORTED=`mktemp`
+  NEW_RESOLV_CONF=`mktemp`
+  NEW_NODE_RESOLV_CONF=`mktemp`
+
+
+  ######################################################################
+  # couldn't find an existing method to determine if the interface owns the
+  # default route
+  def_route=$(/sbin/ip route list match 0.0.0.0/0 | awk '{print $3 }')
+  def_route_int=$(/sbin/ip route get to ${def_route} | awk '{print $3}')
+  def_route_ip=$(/sbin/ip route get to ${def_route} | awk '{print $5}')
+  if [[ ${DEVICE_IFACE} == ${def_route_int} ]]; then
+    if [ ! -f /etc/dnsmasq.d/origin-dns.conf ]; then
+      cat << EOF > /etc/dnsmasq.d/origin-dns.conf
+no-resolv
+domain-needed
+server=/cluster.local/172.30.0.1
+server=/30.172.in-addr.arpa/172.30.0.1
+enable-dbus
+dns-forward-max=5000
+cache-size=5000
+EOF
+      # New config file, must restart
+      NEEDS_RESTART=1
+    fi
+
+    # If network manager doesn't know about the nameservers then the best
+    # we can do is grab them from /etc/resolv.conf but only if we've got no
+    # watermark
+    if ! grep -q '99-origin-dns.sh' /etc/resolv.conf; then
+      if [[ -z "${IP4_NAMESERVERS}" || "${IP4_NAMESERVERS}" == "${def_route_ip}" ]]; then
+            IP4_NAMESERVERS=`grep '^nameserver ' /etc/resolv.conf | awk '{ print $2 }'`
+      fi
+      ######################################################################
+      # Write out default nameservers for /etc/dnsmasq.d/origin-upstream-dns.conf
+      # and /etc/origin/node/resolv.conf in their respective formats
+      for ns in ${IP4_NAMESERVERS}; do
+        if [[ ! -z $ns ]]; then
+          echo "server=${ns}" >> $UPSTREAM_DNS_TMP
+          echo "nameserver ${ns}" >> $NEW_NODE_RESOLV_CONF
+        fi
+      done
+      # Sort it in case DNS servers arrived in a different order
+      sort $UPSTREAM_DNS_TMP > $UPSTREAM_DNS_TMP_SORTED
+      sort $UPSTREAM_DNS > $CURRENT_UPSTREAM_DNS_SORTED
+      # Compare to the current config file (sorted)
+      NEW_DNS_SUM=`md5sum ${UPSTREAM_DNS_TMP_SORTED} | awk '{print $1}'`
+      CURRENT_DNS_SUM=`md5sum ${CURRENT_UPSTREAM_DNS_SORTED} | awk '{print $1}'`
+      if [ "${NEW_DNS_SUM}" != "${CURRENT_DNS_SUM}" ]; then
+        # DNS has changed, copy the temp file to the proper location (-Z
+        # sets default selinux context) and set the restart flag
+        cp -Z $UPSTREAM_DNS_TMP $UPSTREAM_DNS
+        NEEDS_RESTART=1
+      fi
+      # compare /etc/origin/node/resolv.conf checksum and replace it if different
+      NEW_NODE_RESOLV_CONF_MD5=`md5sum ${NEW_NODE_RESOLV_CONF}`
+      OLD_NODE_RESOLV_CONF_MD5=`md5sum /etc/origin/node/resolv.conf`
+      if [ "${NEW_NODE_RESOLV_CONF_MD5}" != "${OLD_NODE_RESOLV_CONF_MD5}" ]; then
+        cp -Z $NEW_NODE_RESOLV_CONF /etc/origin/node/resolv.conf
+      fi
+    fi
+
+    if ! `systemctl -q is-active dnsmasq.service`; then
+      NEEDS_RESTART=1
+    fi
+
+    ######################################################################
+    if [ "${NEEDS_RESTART}" -eq "1" ]; then
+      systemctl restart dnsmasq
+    fi
+
+    # Only if dnsmasq is running properly make it our only nameserver and place
+    # a watermark on /etc/resolv.conf
+    if `systemctl -q is-active dnsmasq.service`; then
+      if ! grep -q '99-origin-dns.sh' /etc/resolv.conf; then
+          echo "# nameserver updated by /etc/NetworkManager/dispatcher.d/99-origin-dns.sh" >> ${NEW_RESOLV_CONF}
+      fi
+      sed -e '/^nameserver.*$/d' /etc/resolv.conf >> ${NEW_RESOLV_CONF}
+      echo "nameserver "${def_route_ip}"" >> ${NEW_RESOLV_CONF}
+      if ! grep -qw search ${NEW_RESOLV_CONF}; then
+        echo 'search cluster.local' >> ${NEW_RESOLV_CONF}
+      elif ! grep -q 'search.*cluster.local' ${NEW_RESOLV_CONF}; then
+        sed -i '/^search/ s/$/ cluster.local/' ${NEW_RESOLV_CONF}
+      fi
+      cp -Z ${NEW_RESOLV_CONF} /etc/resolv.conf
+    fi
+  fi
+
+  # Clean up after yourself
+  rm -f $UPSTREAM_DNS_TMP $UPSTREAM_DNS_TMP_SORTED $CURRENT_UPSTREAM_DNS_SORTED $NEW_RESOLV_CONF
+fi
diff --git a/roles/openshift_node/handlers/main.yml b/roles/openshift_node/handlers/main.yml
index b102c1b18..229c6bbed 100644
--- a/roles/openshift_node/handlers/main.yml
+++ b/roles/openshift_node/handlers/main.yml
@@ -1,4 +1,15 @@
 ---
+- name: restart NetworkManager
+  systemd:
+    name: NetworkManager
+    state: restarted
+    enabled: True
+
+- name: restart dnsmasq
+  systemd:
+    name: dnsmasq
+    state: restarted
+
 - name: restart openvswitch
   systemd:
     name: openvswitch
diff --git a/roles/openshift_node/meta/main.yml b/roles/openshift_node/meta/main.yml
index c32aa1600..373e0e1bd 100644
--- a/roles/openshift_node/meta/main.yml
+++ b/roles/openshift_node/meta/main.yml
@@ -22,6 +22,5 @@ dependencies:
 - role: openshift_docker
 - role: openshift_cloud_provider
   when: not (openshift_node_upgrade_in_progress | default(False))
-- role: openshift_node_dnsmasq
 - role: lib_utils
   when: openshift_node_upgrade_in_progress | default(False)
diff --git a/roles/openshift_node/tasks/dnsmasq.yml b/roles/openshift_node/tasks/dnsmasq.yml
new file mode 100644
index 000000000..22bdce6c6
--- /dev/null
+++ b/roles/openshift_node/tasks/dnsmasq.yml
@@ -0,0 +1,67 @@
+---
+- name: Check for NetworkManager service
+  command: >
+    systemctl show NetworkManager
+  register: nm_show
+  changed_when: false
+  ignore_errors: True
+
+- name: Set fact using_network_manager
+  set_fact:
+    network_manager_active: "{{ True if 'ActiveState=active' in nm_show.stdout else False }}"
+
+- name: Install dnsmasq
+  package: name=dnsmasq state=installed
+  when: not openshift.common.is_atomic | bool
+
+- name: ensure origin/node directory exists
+  file:
+    state: directory
+    path: "{{ item }}"
+    owner: root
+    group: root
+    mode: '0700'
+  with_items:
+  - /etc/origin
+  - /etc/origin/node
+
+# this file is copied to /etc/dnsmasq.d/ when the node starts and is removed
+# when the node stops. A dbus-message is sent to dnsmasq to add the same entries
+# so that dnsmasq doesn't need to be restarted. Once we can use dnsmasq 2.77 or
+# newer we can use --server-file option to update the servers dynamically and
+# reload them by sending dnsmasq a SIGHUP. We write the file in case someone else
+# triggers a restart of dnsmasq but not a node restart.
+- name: Install node-dnsmasq.conf
+  template:
+    src: node-dnsmasq.conf.j2
+    dest: /etc/origin/node/node-dnsmasq.conf
+
+- name: Install dnsmasq configuration
+  template:
+    src: origin-dns.conf.j2
+    dest: /etc/dnsmasq.d/origin-dns.conf
+  notify: restart dnsmasq
+
+- name: Deploy additional dnsmasq.conf
+  template:
+    src: "{{ openshift_node_dnsmasq_additional_config_file }}"
+    dest: /etc/dnsmasq.d/openshift-ansible.conf
+    owner: root
+    group: root
+    mode: 0644
+  when: openshift_node_dnsmasq_additional_config_file is defined
+  notify: restart dnsmasq
+
+- name: Enable dnsmasq
+  systemd:
+    name: dnsmasq
+    enabled: yes
+    state: started
+
+# Dynamic NetworkManager based dispatcher
+- include_tasks: dnsmasq/network-manager.yml
+  when: network_manager_active | bool
+
+# Relies on ansible in order to configure static config
+- include_tasks: dnsmasq/no-network-manager.yml
+  when: not network_manager_active | bool
diff --git a/roles/openshift_node/tasks/dnsmasq/network-manager.yml b/roles/openshift_node/tasks/dnsmasq/network-manager.yml
new file mode 100644
index 000000000..e5a92a630
--- /dev/null
+++ b/roles/openshift_node/tasks/dnsmasq/network-manager.yml
@@ -0,0 +1,10 @@
+---
+- name: Install network manager dispatch script
+  copy:
+    src: networkmanager/99-origin-dns.sh
+    dest: /etc/NetworkManager/dispatcher.d/
+    mode: 0755
+  notify: restart NetworkManager
+  when: openshift_node_dnsmasq_install_network_manager_hook | default(true) | bool
+
+- meta: flush_handlers
diff --git a/roles/openshift_node/tasks/dnsmasq/no-network-manager.yml b/roles/openshift_node/tasks/dnsmasq/no-network-manager.yml
new file mode 100644
index 000000000..dede2fb8f
--- /dev/null
+++ b/roles/openshift_node/tasks/dnsmasq/no-network-manager.yml
@@ -0,0 +1,11 @@
+---
+- fail: msg="Currently, NetworkManager must be installed and enabled prior to installation."
+  when: not openshift_node_bootstrap | bool
+
+- name: Install NetworkManager during node_bootstrap provisioning
+  package:
+    name: NetworkManager
+    state: present
+  notify: restart NetworkManager
+
+- include_tasks: network-manager.yml
diff --git a/roles/openshift_node/tasks/main.yml b/roles/openshift_node/tasks/main.yml
index 8e9d1d1b5..d46b1f9c3 100644
--- a/roles/openshift_node/tasks/main.yml
+++ b/roles/openshift_node/tasks/main.yml
@@ -6,6 +6,8 @@
     - deployment_type == 'openshift-enterprise'
     - not openshift_use_crio | default(false)
 
+- include: dnsmasq.yml
+
 - name: setup firewall
   import_tasks: firewall.yml
 
diff --git a/roles/openshift_node/tasks/upgrade.yml b/roles/openshift_node/tasks/upgrade.yml
index fb21b39a1..ff3478800 100644
--- a/roles/openshift_node/tasks/upgrade.yml
+++ b/roles/openshift_node/tasks/upgrade.yml
@@ -179,5 +179,4 @@
   retries: 24
   delay: 5
 
-- include_role:
-    name: openshift_node_dnsmasq
+- include_tasks: dnsmasq.yml
diff --git a/roles/openshift_node/templates/node-dnsmasq.conf.j2 b/roles/openshift_node/templates/node-dnsmasq.conf.j2
new file mode 100644
index 000000000..3caa3bd4a
--- /dev/null
+++ b/roles/openshift_node/templates/node-dnsmasq.conf.j2
@@ -0,0 +1,2 @@
+server=/in-addr.arpa/127.0.0.1
+server=/{{ openshift.common.dns_domain }}/127.0.0.1
diff --git a/roles/openshift_node/templates/origin-dns.conf.j2 b/roles/openshift_node/templates/origin-dns.conf.j2
new file mode 100644
index 000000000..6543c7c3e
--- /dev/null
+++ b/roles/openshift_node/templates/origin-dns.conf.j2
@@ -0,0 +1,12 @@
+no-resolv
+domain-needed
+no-negcache
+max-cache-ttl=1
+enable-dbus
+dns-forward-max=5000
+cache-size=5000
+bind-dynamic
+{% for interface in openshift_node_dnsmasq_except_interfaces %}
+except-interface={{ interface }}
+{% endfor %}
+# End of config