From 0799782ce83d1057f262b44c979a15f9a1b05c72 Mon Sep 17 00:00:00 2001
From: Michal Nasiadka <mnasiadka@gmail.com>
Date: Thu, 12 Dec 2019 13:19:48 +0100
Subject: [PATCH] Fix keystone fernet bootstrap

There are cases when a multinode deployment ends up in unusable
keystone public wsgi on some nodes.

The root cause is that keystone public wsgi doesn't find fernet
keys on startup - and then persists on sending 500 errors to any
requests - due to a race condition between
fernet_setup/fernet-push.sh and keystone startup.

Depends-On: https://review.opendev.org/703742/
Change-Id: I63709c2e3f6a893db82a05640da78f492bf8440f
Closes-Bug: #1846789
---
 .../keystone/tasks/bootstrap_service.yml      | 38 ++++++++++++++++++-
 ansible/roles/keystone/tasks/deploy.yml       |  2 +-
 .../keystone/tasks/distribute_fernet.yml      | 19 ++++++++++
 ansible/roles/keystone/tasks/init_fernet.yml  | 27 -------------
 .../keystone/templates/fernet-node-sync.sh.j2 | 20 +++++++++-
 .../keystone/templates/fernet-push.sh.j2      |  3 ++
 .../keystone/templates/fernet-rotate.sh.j2    |  3 ++
 ...fix-fernet-bootstrap-36f87e36e4dc6ec9.yaml |  6 +++
 8 files changed, 87 insertions(+), 31 deletions(-)
 create mode 100644 ansible/roles/keystone/tasks/distribute_fernet.yml
 delete mode 100644 ansible/roles/keystone/tasks/init_fernet.yml
 create mode 100644 releasenotes/notes/bug-1846789-fix-fernet-bootstrap-36f87e36e4dc6ec9.yaml

diff --git a/ansible/roles/keystone/tasks/bootstrap_service.yml b/ansible/roles/keystone/tasks/bootstrap_service.yml
index 0a296fb2b8..8a20685edd 100644
--- a/ansible/roles/keystone/tasks/bootstrap_service.yml
+++ b/ansible/roles/keystone/tasks/bootstrap_service.yml
@@ -1,4 +1,15 @@
 ---
+- name: Checking for any running keystone_fernet containers
+  become: true
+  kolla_container_facts:
+    name:
+      - keystone_fernet
+  register: container_facts
+
+- name: Group nodes where keystone_fernet is running
+  group_by:
+    key: keystone_fernet_{{ container_facts['keystone_fernet'].State | default('bootstrap') }}
+
 - name: Running Keystone bootstrap container
   vars:
     keystone: "{{ keystone_services.keystone }}"
@@ -17,4 +28,29 @@
     restart_policy: no
     volumes: "{{ keystone.volumes|reject('equalto', '')|list }}"
   run_once: True
-  delegate_to: "{{ groups['keystone'][0] }}"
+
+- name: Running Keystone fernet bootstrap container
+  vars:
+    keystone_fernet: "{{ keystone_services['keystone-fernet'] }}"
+  become: true
+  kolla_docker:
+    action: "start_container"
+    common_options: "{{ docker_common_options }}"
+    detach: False
+    environment:
+      KOLLA_CONFIG_STRATEGY: "{{ config_strategy }}"
+    image: "{{ keystone_fernet.image }}"
+    labels:
+      BOOTSTRAP:
+    command: >
+      bash -c 'sudo -E kolla_set_configs &&
+      keystone-manage --config-file /etc/keystone/keystone.conf
+      fernet_setup --keystone-user {{ keystone_username }} --keystone-group {{ keystone_groupname }} && ls -l /etc/keystone/fernet-keys/'
+    name: "bootstrap_keystone_fernet"
+    restart_policy: no
+    volumes: "{{ keystone_fernet.volumes|reject('equalto', '')|list }}"
+  run_once: True
+  delegate_to: "{{ groups['keystone_fernet_bootstrap'][0] }}"
+  when:
+    - keystone_token_provider == 'fernet'
+    - groups['keystone_fernet_running'] is not defined
diff --git a/ansible/roles/keystone/tasks/deploy.yml b/ansible/roles/keystone/tasks/deploy.yml
index cf24dfb30b..7a73af2b53 100644
--- a/ansible/roles/keystone/tasks/deploy.yml
+++ b/ansible/roles/keystone/tasks/deploy.yml
@@ -10,7 +10,7 @@
 - name: Flush handlers
   meta: flush_handlers
 
-- include_tasks: init_fernet.yml
+- include_tasks: distribute_fernet.yml
   when:
     - keystone_token_provider == 'fernet'
 
diff --git a/ansible/roles/keystone/tasks/distribute_fernet.yml b/ansible/roles/keystone/tasks/distribute_fernet.yml
new file mode 100644
index 0000000000..184acac840
--- /dev/null
+++ b/ansible/roles/keystone/tasks/distribute_fernet.yml
@@ -0,0 +1,19 @@
+---
+- name: Waiting for Keystone SSH port to be UP
+  wait_for:
+    host: "{{ api_interface_address }}"
+    port: "{{ keystone_ssh_port }}"
+    connect_timeout: 1
+  register: check_keystone_ssh_port
+  until: check_keystone_ssh_port is success
+  retries: 10
+  delay: 5
+
+- name: Run key distribution
+  become: true
+  command: docker exec -t keystone_fernet /usr/bin/fernet-push.sh
+  run_once: True
+  delegate_to: >-
+    {% if groups['keystone_fernet_running'] is defined -%}
+    {{ groups['keystone_fernet_running'][0] }}
+    {%- else -%}{{ groups['keystone'][0] }}{%- endif %}
diff --git a/ansible/roles/keystone/tasks/init_fernet.yml b/ansible/roles/keystone/tasks/init_fernet.yml
deleted file mode 100644
index 9fa0769468..0000000000
--- a/ansible/roles/keystone/tasks/init_fernet.yml
+++ /dev/null
@@ -1,27 +0,0 @@
----
-- name: Waiting for Keystone SSH port to be UP
-  wait_for:
-    host: "{{ api_interface_address }}"
-    port: "{{ keystone_ssh_port }}"
-    connect_timeout: 1
-  register: check_keystone_ssh_port
-  until: check_keystone_ssh_port is success
-  retries: 10
-  delay: 5
-
-- name: Initialise fernet key authentication
-  become: true
-  command: "docker exec -t keystone_fernet kolla_keystone_bootstrap {{ keystone_username }} {{ keystone_groupname }}"
-  register: fernet_create
-  changed_when: fernet_create.stdout.find('localhost | SUCCESS => ') != -1 and (fernet_create.stdout.split('localhost | SUCCESS => ')[1]|from_json).changed
-  until: fernet_create.stdout.split()[2] == 'SUCCESS' or fernet_create.stdout.find('Key repository is already initialized') != -1
-  retries: 10
-  delay: 5
-  run_once: True
-  delegate_to: "{{ groups['keystone'][0] }}"
-
-- name: Run key distribution
-  become: true
-  command: docker exec -t keystone_fernet /usr/bin/fernet-push.sh
-  run_once: True
-  delegate_to: "{{ groups['keystone'][0] }}"
diff --git a/ansible/roles/keystone/templates/fernet-node-sync.sh.j2 b/ansible/roles/keystone/templates/fernet-node-sync.sh.j2
index 050198f0d8..7556eed518 100644
--- a/ansible/roles/keystone/templates/fernet-node-sync.sh.j2
+++ b/ansible/roles/keystone/templates/fernet-node-sync.sh.j2
@@ -4,10 +4,26 @@ set -o errexit
 set -o pipefail
 
 # Get data on the fernet tokens
-TOKEN_CHECK=$(/usr/bin/python{{ distro_python_version }} /usr/bin/fetch_fernet_tokens.py -t {{ fernet_token_expiry }} -n {{ (groups['keystone'] | length) + 1 }})
+# NOTE(mnasiadka): Check for existence of at least two tokens (should exist after bootstrap)
+TOKEN_CHECK=$(/usr/bin/python{{ distro_python_version }} /usr/bin/fetch_fernet_tokens.py -t {{ fernet_token_expiry }} -n 2)
+
+# Ensure tokens are populated
+n=0
+while /usr/bin/python{{ distro_python_version }} /usr/bin/fetch_fernet_tokens.py -t 86400 -n 1 | grep -q '"populated": false'; do
+    if [ $n -lt 10 ]; then
+        n=$(( n + 1 ))
+        echo "ERROR: Fernet tokens have not been populated, rechecking in 1 minute"
+        echo "DEBUG: /etc/keystone/fernet-keys contents:"
+        ls -l /etc/keystone/fernet-keys/
+        sleep 60
+    else
+        echo "CRITICAL: Waited for 10 minutes - failing"
+        exit 1
+    fi
+done
 
 # Ensure the primary token exists and is not stale
-if $(echo "$TOKEN_CHECK" | grep -q '"update_required":"false"'); then
+if $(echo "$TOKEN_CHECK" | grep -q '"update_required": false'); then
     exit 0;
 fi
 
diff --git a/ansible/roles/keystone/templates/fernet-push.sh.j2 b/ansible/roles/keystone/templates/fernet-push.sh.j2
index 6179cb2a81..5aeda018fc 100644
--- a/ansible/roles/keystone/templates/fernet-push.sh.j2
+++ b/ansible/roles/keystone/templates/fernet-push.sh.j2
@@ -1,5 +1,8 @@
 #!/bin/bash
 
+set -o errexit
+set -o pipefail
+
 {% for host in groups['keystone'] %}
 {% if inventory_hostname != host %}
 /usr/bin/rsync -az -e 'ssh -i /var/lib/keystone/.ssh/id_rsa -p {{ hostvars[host]['keystone_ssh_port'] }} -F /var/lib/keystone/.ssh/config' --delete /etc/keystone/fernet-keys/ keystone@{{ 'api' | kolla_address(host) | put_address_in_context('url') }}:/etc/keystone/fernet-keys
diff --git a/ansible/roles/keystone/templates/fernet-rotate.sh.j2 b/ansible/roles/keystone/templates/fernet-rotate.sh.j2
index 3ef7a0e63c..a67aa19e26 100644
--- a/ansible/roles/keystone/templates/fernet-rotate.sh.j2
+++ b/ansible/roles/keystone/templates/fernet-rotate.sh.j2
@@ -1,5 +1,8 @@
 #!/bin/bash
 
+set -o errexit
+set -o pipefail
+
 keystone-manage --config-file /etc/keystone/keystone.conf fernet_rotate --keystone-user {{ keystone_username }} --keystone-group {{ keystone_groupname }}
 
 /usr/bin/fernet-push.sh
diff --git a/releasenotes/notes/bug-1846789-fix-fernet-bootstrap-36f87e36e4dc6ec9.yaml b/releasenotes/notes/bug-1846789-fix-fernet-bootstrap-36f87e36e4dc6ec9.yaml
new file mode 100644
index 0000000000..c60bd36344
--- /dev/null
+++ b/releasenotes/notes/bug-1846789-fix-fernet-bootstrap-36f87e36e4dc6ec9.yaml
@@ -0,0 +1,6 @@
+---
+fixes:
+  - |
+    Rework keystone fernet bootstrap which had tendencies to fail on multinode
+    setups. See `bug 1846789 <https://launchpad.net/bugs/1846789>`__ for
+    details.