1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
|
.. _build_guest_images:
.. role:: bash(code)
:language: bash
====================
Building guest image
====================
Overview
========
When Trove receives a command to create a database instance, it does so by
launching a Nova instance based on the appropriate guest image that is
stored in Glance. This document shows you the steps to build the guest images.
.. note::
For testing purpose, the Trove guest images of some specific databases are
periodically built and published in
http://tarballs.openstack.org/trove/images/ in Trove upstream CI.
Since Victoria release, Trove supports to run database service as docker
container inside the guest instance, so that we don't need to maintain
multiple images for different database service. That's the reason that you
can see images for MySQL and MariaDB for Ussuri and Train releases in
http://tarballs.openstack.org/trove/images/.
Additionally, if you install Trove in devstack environment, the guest image
is created and registered in Glance automatically, unless it's disabled by
setting ``TROVE_ENABLE_IMAGE_BUILD=false`` in devstack local.conf file.
Since Yoga release, Trove sets default use_nova_server_config_drive to
``False``. This change will break instance provisioning in some cases.
To avoid that, make sure your ``DIB_CLOUD_INIT_DATASOURCES`` environment
contains ``OpenStack`` to configure cloud-init to use metadata serivce.
If cloud-init still doesn't inject files after changing the environment,
please ask your IaaS provider admin to enable OpenStack metadata service.
High Level Overview of a Trove Guest Instance
=============================================
At the most basic level, a Trove Guest Instance is a Nova instance launched by
Trove in response to a create command. This section describes the various
components of a Trove Guest Instance.
----------------
Operating System
----------------
The officially supported operating system is Ubuntu, based on which the
functional tests are running.
------
Docker
------
Since Vitoria, the database service is running as docker container inside the
trove guest instance, so docker should be installed when building the guest
image. This also means the trove guest instance should be able to pull docker
images from the image registry(either from user port or trove management port),
the related options for container images are:
.. code-block:: ini
[mysql]
docker_image
backup_docker_image
[postgresql]
docker_image
backup_docker_image
[mariadb]
docker_image
backup_docker_image
-----------------
Trove Guest Agent
-----------------
The guest agent runs inside the Nova instances that are used to run the
database engines. The agent listens to the messaging bus for the topic and is
responsible for actually translating and executing the commands that are sent
to it by the task manager component for the particular datastore.
Trove guest agent is responsible for datastore docker container management.
------------------------------------------
Injected Configuration for the Guest Agent
------------------------------------------
When TaskManager launches the guest VM it injects config files into the
VM, including:
* ``/etc/trove/conf.d/guest_info.conf``: Contains some information about
the guest, e.g. the guest identifier, the tenant ID, etc.
* ``/etc/trove/conf.d/trove-guestagent.conf``: The config file for the
guest agent service.
In addition to these config files, Trove supports to inject user data when
launching the instance for customization on boot time, e.g. network
configuration, hosts file settings, etc. The user data files are located inside
the directory configured by ``cloudinit_location``, for mysql, the file name is
``mysql.cloudinit``
------------------------------
Persistent Storage, Networking
------------------------------
The database stores data on persistent storage on Cinder (if
``CONF.volume_support=True``) or ephemeral storage on the Nova instance. The
database service is accessible over the tenant network provided when creating
the database instance.
The cloud administrator is able to config management
networks(``CONF.management_networks``) that is invisible to the cloud tenants,
but used for communication between database instance and the control plane
services(e.g. the message queue).
Building Guest Images
=====================
Since Victoria release, a single trove guest image can be used for different
datastores, it's unnecessary to maintain different images for differnt
datastores.
-----------------------------
Build images using trovestack
-----------------------------
``trovestack`` is the recommended tooling provided by Trove community to build
the guest images. Before running ``trovestack`` command:
.. code-block:: console
git clone https://opendev.org/openstack/trove
cd trove/integration/scripts
The trove guest image could be created by running the following command:
.. code-block:: console
$ ./trovestack build-image \
${guest_os} \
${guest_os_release} \
${dev_mode} \
${guest_username} \
${output_image_path}
* Currently, only ``guest_os=ubuntu`` and ``guest_os_release=focal`` are fully
tested and supported.
* Default input values:
.. code-block:: ini
guest_os=ubuntu
guest_os_release=focal
dev_mode=true
guest_username=ubuntu
output_image_path=$HOME/images/trove-guest-${guest_os}-${guest_os_release}-dev.qcow2
* ``dev_mode=true`` is mainly for testing purpose for trove developers and it's
necessary to build the image on the trove controller host, because the host
and the guest VM need to ssh into each other without password. In this mode,
when the trove guest agent code is changed, the image doesn't need to be
rebuilt which is convenient for debugging. Trove guest agent will ssh into
the controller node and download trove code during the service initialization.
* If ``dev_mode=false``, the trove code for guest agent is injected into the
image at the building time.
* Some other global variables:
* ``HOST_SCP_USERNAME``: Only used in dev mode, this is the user name used by
guest agent to connect to the controller host, e.g. in devstack
environment, it should be the ``stack`` user.
* The image type can be easily changed by specifying a different image file
extension, e.g. to build a raw image, you can specify
``$your-image-name.raw`` as the ``output_image_path`` parameter.
For example, in order to build a guest image for Ubuntu focal operating
system in development mode:
.. code-block:: console
$ ./trovestack build-image ubuntu focal true ubuntu
Once the image build is finished, the cloud administrator needs to register the
image in Glance and register a new datastore or version in Trove using
``trove-manage`` command, e.g. after building an image for MySQL 5.7.29:
.. code-block:: console
$ openstack image create trove-guest-ubuntu-focal \
--private \
--disk-format qcow2 \
--container-format bare \
--tag trove --tag mysql \
--file ~/images/trove-guest-ubuntu-focal-dev.qcow2
$ openstack datastore version create 5.7.29 mysql mysql "" \
--image-tags trove,mysql \
--active --default
$ trove-manage db_load_datastore_config_parameters mysql 5.7.29 ${trove_repo_dir}/trove/templates/mysql/validation-rules.json
.. note::
The command ``trove-manage`` needs to run on Trove controller node.
If you see anything error or need help for the image creation, please ask help
either in ``#openstack-trove`` IRC channel or sending emails to
openstack-discuss@lists.openstack.org mailing list.
|
