root / doc / design-os.rst @ 36399e2a
History | View | Annotate | Download (20.9 kB)
1 |
=============================== |
---|---|
2 |
Ganeti OS installation redesign |
3 |
=============================== |
4 |
|
5 |
.. contents:: :depth: 3 |
6 |
|
7 |
This is a design document detailing a new OS installation procedure, which is |
8 |
more secure, able to provide more features and easier to use for many common |
9 |
tasks w.r.t. the current one. |
10 |
|
11 |
Current state and shortcomings |
12 |
============================== |
13 |
|
14 |
As of Ganeti 2.10, each instance is associated with an OS definition. An OS |
15 |
definition is a set of scripts (``create``, ``export``, ``import``, ``rename``) |
16 |
that are executed with root privileges on the primary host of the instance to |
17 |
perform all the OS-related functionality (setting up an operating system inside |
18 |
the disks of the instance being created, exporting/importing the instance, |
19 |
renaming it). |
20 |
|
21 |
These scripts receive, as environment variables, a fixed set of parameters |
22 |
related to the instance (such as the hypervisor, the name of the instance, the |
23 |
number of disks, and their location) and a set of user defined parameters. |
24 |
These parameters are also written in the configuration file of Ganeti, to allow |
25 |
future reinstalls of the instance, and in various log files, namely: |
26 |
|
27 |
* node daemon log file: contains DEBUG strings of the ``/os_validate``, |
28 |
``/instance_os_add`` and ``/instance_start`` RPC calls. |
29 |
|
30 |
* master daemon log file: DEBUG strings related to the same RPC calls are stored |
31 |
here as well. |
32 |
|
33 |
* commands log: the CLI commands that create a new instance, including their |
34 |
parameters, are logged here. |
35 |
|
36 |
* RAPI log: the RAPI commands that create a new instance, including their |
37 |
parameters, are logged here. |
38 |
|
39 |
* job logs: the job files stored in the job queue, or in its archive, contain |
40 |
the parameters. |
41 |
|
42 |
The current situation presents a number of shortcomings: |
43 |
|
44 |
* Having the installation scripts run as root on the nodes doesn't allow |
45 |
user-defined OS scripts, as they would pose a huge security issue. |
46 |
Furthermore, even a script without malicious intentions might end up |
47 |
distrupting a node because of a bug in it. |
48 |
|
49 |
* Ganeti cannot be used to create instances starting from user provided disk |
50 |
images: even in the (hypothetical) case where the scripts are completely |
51 |
secure and run not by root but by an unprivileged user with only the power to |
52 |
mount arbitrary files as disk images, this is a security issue. It has been |
53 |
proven that a carefully crafted file system might exploit kernel |
54 |
vulnerabilities to gain control of the system. Therefore, directly mounting |
55 |
images on the Ganeti nodes is not an option. |
56 |
|
57 |
* There is no way to inject files into an existing disk image. A common use case |
58 |
is for the system administrator to provide a standard image of the system, to |
59 |
be later personalized with the network configuration, private keys identifying |
60 |
the machine, ssh keys of the users and so on. A possible workaround would be |
61 |
for the scripts to mount the image (only if this is trusted!) and to receive |
62 |
the configurations and ssh keys as user defined OS parameters. Unfortunately, |
63 |
this is also not an option for security sensitive material (such as the ssh |
64 |
keys) because the OS parameters are stored in many places on the system, as |
65 |
already described above. |
66 |
|
67 |
* Most other virtualization software simply work with instance images, not with |
68 |
installation scripts. This difference makes the interaction of Ganeti with |
69 |
other software difficult. |
70 |
|
71 |
Proposed changes |
72 |
================ |
73 |
|
74 |
In order to fix the shortcomings of the current state, we plan to introduce the |
75 |
following changes: |
76 |
|
77 |
* Change the OS parameters to have three categories: |
78 |
|
79 |
* ``public``: the current behavior. The parameter is logged and stored freely. |
80 |
|
81 |
* ``private``: the parameter is saved inside the Ganeti configuration (to allow |
82 |
for instance reinstall) but it is not shown in logs, job logs, or passed back |
83 |
via RAPI. |
84 |
|
85 |
* ``secret``: the parameter is not saved inside the Ganeti configuration. |
86 |
Reinstalls are impossible unless the data is passed again. The parameter will |
87 |
not appear in any log file. When a functionality is performed jointly by |
88 |
multiple daemons (such as MasterD and LuxiD), currently Ganeti sometimes |
89 |
serializes jobs on disk and later reloads them. Secret parameters will not be |
90 |
serialized to disk. They will be passed around as part of the LUXI calls |
91 |
exchanged by the daemons, and only kept in memory, in order to reduce their |
92 |
accessibility as much as possible. In case of failure of the master node, |
93 |
these parameters will be lost and cannot be recovered because they are not |
94 |
serialized. As a result, the job cannot be taken over by the new master. |
95 |
This is an expected and accepted side effect of jobs with secret parameters: |
96 |
if they fail, they'll have to be restarted manually. |
97 |
|
98 |
* A new OS installation procedure, based on a safe virtualized environment. |
99 |
This virtualized environment will run with the same hardware parameter as the |
100 |
actual instance being installed, as much as possible. This will also allow to |
101 |
reduce the memory usage in the host (specifically, in Dom0 for Xen |
102 |
installations). Each instance will have these possible execution modes: |
103 |
|
104 |
* ``run``: the default mode, used when the machine is running normally and |
105 |
the OS installation procedure is run before starting the instance for the |
106 |
first time. |
107 |
|
108 |
* ``self_install``: the first run of the instance will be with a different set |
109 |
of parameters w.r.t. all the successive runs. This set of "install |
110 |
parameters" will allow, e.g., to attach an installation |
111 |
floppy/cdrom/network, change the boot device order, or specify an OS image |
112 |
to be used. Through this set of parameters, the administrator will have to |
113 |
provide the hypervisor a way to find an installation medium for the instance |
114 |
(e.g., a boot disk, a network image, etc). This medium will then install the |
115 |
instance itself on the disks and will then be responsible to get the |
116 |
parameters for configuring it (its network interfaces, IP address, hostname, |
117 |
etc.) from a set of metadata provided by Ganeti (e.g.: using an approach |
118 |
comparable to the one of the ``cloud-init`` tool). When this installation |
119 |
mode is used, no OS installation script is required. In order for the |
120 |
installation of an OS from an image to be possible, the ``--os-type`` |
121 |
parameter will be extended to support a new additional format: ``--os-type |
122 |
image:<URL>`` will instruct Ganeti to take an image from the specified |
123 |
position. For the initial implementation, URL can be either a filename or a |
124 |
publically accessible HTTP or FTP resource. Once the instance image is |
125 |
received, it will be dd-ed onto the first disk of the instance. When an |
126 |
image is specified, ``--os-parameters`` can still be used, and its content |
127 |
will be passed to the instance as part of the metadata. Note that, as part |
128 |
of the OS scripts, there is a file specifying what parameters are |
129 |
expected. With OS images, though, none of the traditional structure of OS |
130 |
scripts is in place, so there will be no check regarding what parameters can |
131 |
be specified: they will all be passed, as long as the ``--os-parameters`` |
132 |
string is syntactically valid. The set of ``self_install`` parameters will |
133 |
be stored as part of the instance configuration, so that they can be used to |
134 |
reinstall the instance. It will be the user's responsibility to ensure that |
135 |
the OS image or any installation media is still available in the proper |
136 |
position when a reinstall happens. After the first run, the instance will |
137 |
revert to ``run`` mode. |
138 |
|
139 |
* ``install``: Ganeti will start the instance using a virtual appliance |
140 |
specifically made for installing Ganeti instances. Scripts analogous to the |
141 |
current ones will run inside this instance. The disks of the instance being |
142 |
installed will be connected to this virtual appliance, so that the scripts |
143 |
can mount them and modify them as needed, as currently happens, but with the |
144 |
additional protection given by this happening in a VM. The disk of the |
145 |
virtual appliance will be read only, so that a pristine copy of the |
146 |
appliance can be started every time a new instance needs to be created, to |
147 |
further increase security. The data the instance needs to write at runtime |
148 |
will only be stored in RAM, and disappear as soon as the instance is |
149 |
stopped. Metadata will be provided also to this virtual applicance, that |
150 |
will take care of converting them to environment variables for the |
151 |
installation scripts. After the first run, the instance will revert to |
152 |
``run`` mode. |
153 |
|
154 |
* In order to allow for the metadata to be sent inside the instance, a |
155 |
communication mechanism between the instance and the host will be created. |
156 |
This mechanism will be bidirectional (e.g.: to allow the setup process going |
157 |
on inside the instance to communicate its progress to the host). Each instance |
158 |
will have access exclusively to its own metadata, and it will be only able to |
159 |
communicate with its host over this channel. More details will be provided in |
160 |
the `Communication mechanism and metadata service`_ section. |
161 |
|
162 |
* As part of the instance creation command it will be possible to indicate a URL |
163 |
for a "personalization package", that is an archive containing a set of files |
164 |
meant to be overlayed on top of the operating system file system at the end of |
165 |
the setup process, before the VM is started for the first time in ``run`` |
166 |
mode. Ganeti will provide a mechanism for receiving and unpacking this |
167 |
archive as part of the ``install`` execution mode, whereas in ``self_install`` |
168 |
mode it will only be provided as a metadata for the instance to use. The |
169 |
archive will be in TAR-GZIP format (with extension ``.tar.gz`` or ``.tgz``) |
170 |
and will contain the files according to the directory structure that will be |
171 |
recreated on the installation disk. Files contained in this archive will |
172 |
overwrite files with the same path created during the install procedure (if |
173 |
any). The URL of the "personalization package" will have to specify an |
174 |
extesion to identify the file format (in order to allow for more formats to be |
175 |
supported in the future). The URL will be stored as part of the configuration |
176 |
of the instance (therefore, the URL should not contain confidential |
177 |
information, but the files there available can). It is up to the system |
178 |
administrator to ensure that a package is actually available at that URL at |
179 |
install and reinstall time. The content of the package is allowed to change. |
180 |
E.g.: a system administrator might create a package containing the private |
181 |
keys of the instance being created. When the instance is reinstalled, a new |
182 |
package with new keys can be made available there, therefore allowing instance |
183 |
reinstall without the need to store keys. Together with the URL, a username |
184 |
and a password can be specified to. If the URL is a HTTP(S) URL, they will be |
185 |
used as basic access authentication credentials to access that URL. The |
186 |
username and password will not be saved in the config, and will have to be |
187 |
provided again in case a reinstall is requested. The downloaded |
188 |
personalization package will not be stored locally on the node for longer than |
189 |
it is needed while unpacking it and adding its files to the instance being |
190 |
created. The personalization package will be overlayed on top of the instance |
191 |
filesystem after the scripts that created it have been executed. In order for |
192 |
the files in the package to be automatically overlayed on top of the instance |
193 |
filesystem it is required that the appliance is actually able to mount the |
194 |
instance disks, therefore this will not work for every filesystem. |
195 |
|
196 |
Implementation |
197 |
============== |
198 |
|
199 |
The implementation of this design will happen as an ordered sequence of steps, |
200 |
of increasing impact on the system and, in some cases, dependent on each other: |
201 |
|
202 |
#. Private and secret instance parameters |
203 |
#. Communication mechanism between host and instance |
204 |
#. Metadata service |
205 |
#. Personalization package (inside a virtualization environment) |
206 |
#. ``self_install`` mode |
207 |
#. ``install`` mode (inside a virtualization environment) |
208 |
|
209 |
Some of these steps need to be more deeply specified w.r.t. what is already |
210 |
written in the `Proposed changes`_ Section. Extra details will be provided in |
211 |
the following subsections. |
212 |
|
213 |
Communication mechanism and metadata service |
214 |
++++++++++++++++++++++++++++++++++++++++++++ |
215 |
|
216 |
The communication mechanism and the metadata service are described together |
217 |
because they are deeply tied. On the other hand, the communication mechanism |
218 |
will need to be more generic because it can be used for other reasons in the |
219 |
future (like allowing instances to explicitly send commands to Ganeti, or to let |
220 |
Ganeti control a helper instance, like the one hereby introduced for performing |
221 |
OS installs inside a safe environment). |
222 |
|
223 |
The communication mechanism will be enabled automatically when the instance is |
224 |
in ``self_install`` or ``install`` mode, but for backwards compatibility it will |
225 |
be disabled when the instance is in ``run`` mode unless it is explicitly |
226 |
requested. Specifically, a new parameter ``--communication`` (short version: |
227 |
``-C``), with possible values ``true`` or ``false`` will be added to |
228 |
``gnt-instance add`` and ``gnt-instance modify``. It will determine whether the |
229 |
instance will have a communication channel set up to interact with the host and |
230 |
to receive metadata. The value of this parameter will be saved as part of the |
231 |
configuration of the instance. |
232 |
|
233 |
When the communication mechanism is enabled, Ganeti will create a new network |
234 |
interface inside the instance. This additional network interface will be the |
235 |
last one in the instance, after all the user defined ones. On the host side, |
236 |
this interface will only be accessible to the host itself, and not routed |
237 |
outside the machine. |
238 |
On this network interface, the instance will connect using the IP: |
239 |
169.254.169.253 and netmask 255.255.255.0. |
240 |
The host will be on the same network, with the IP address: 169.254.169.254. |
241 |
|
242 |
The way to create this interface depends on the specific hypervisor being used. |
243 |
In KVM, it is possible to create a network interface inside the instance without |
244 |
having a corresponding interface created on the host. Using a command like:: |
245 |
|
246 |
kvm -net nic -net \ |
247 |
user,restrict=on,net=169.254.169.0/24,host=169.254.169.253, |
248 |
guestfwd=tcp:169.254.169.254:80-tcp:127.0.0.1:8080 |
249 |
|
250 |
a network interface will be created inside the VM, part of the 169.254.169.0/24 |
251 |
network, where the VM will have IP address .253 and the host port 8080 will be |
252 |
reachable on port 80. |
253 |
|
254 |
In Xen, unfortunately, such a capability is not present, and an actual network |
255 |
interface has to be created on the host (using the ``vif`` parameter of the Xen |
256 |
configuration file). Each instance will have its corresponding ``vif`` network |
257 |
interface on the host. These interfaces will not be connected to each other in |
258 |
any way, and Ganeti will not configure them to allow traffic to be forwarded |
259 |
beyond the host machine. The ``vif-route`` script of Xen might be helpful in |
260 |
implementing this. |
261 |
It will be the system administrator's responsibility to ensure that the extra |
262 |
firewalling and routing rules specified on the host don't allow this |
263 |
accidentally. |
264 |
|
265 |
The instance will be able to connect to 169.254.169.254:80, and issue GET |
266 |
requests to an HTTP server that will provide the instance metadata. |
267 |
|
268 |
The choice of this IP address and port for accessing the metadata is done for |
269 |
compatibility reasons with OpenStack's and Amazon EC2's ways of providing |
270 |
metadata to the instance. The metadata will be provided by a single daemon, |
271 |
which will determine what instance the request comes from and reply with the |
272 |
metadata specific for that instance. |
273 |
|
274 |
Where possible, the metadata will be provided in a way compatible with Amazon |
275 |
EC2, at:: |
276 |
|
277 |
http://169.254.169.254/<version>/meta-data/* |
278 |
|
279 |
If some metadata are Ganeti-specific and don't fit this structure, they will be |
280 |
provided at:: |
281 |
|
282 |
http://169.254.169.254/ganeti/<version>/meta_data.json |
283 |
|
284 |
``<version>`` is either a date in YYYY-MM-DD format, or ``latest`` to indicate |
285 |
the most recent available protocol version. |
286 |
|
287 |
If needed in the future, this structure also allows us to support OpenStack's |
288 |
metadata at:: |
289 |
|
290 |
http://169.254.169.254/openstack/<version>/meta_data.json |
291 |
|
292 |
A bi-directional, pipe-like communication channel will be provided. The instance |
293 |
will be able to receive data from the host by a GET request at:: |
294 |
|
295 |
http://169.254.169.254/ganeti/<version>/read |
296 |
|
297 |
and to send data to the host by a POST request at:: |
298 |
|
299 |
http://169.254.169.254/ganeti/<version>/write |
300 |
|
301 |
As in a pipe, once the data are read, they will not be in the buffer anymore, so |
302 |
subsequent GET requests to ``read`` will not return the same data twice. |
303 |
Unlike a pipe, though, it will not be possible to perform blocking I/O |
304 |
operations. |
305 |
|
306 |
The OS parameters will be accessible through a GET |
307 |
request at:: |
308 |
|
309 |
http://169.254.169.254/ganeti/<version>/os/parameters.json |
310 |
|
311 |
as a JSON serialized dictionary having the parameter name as the key, and the |
312 |
pair ``(<value>, <visibility>)`` as the value, where ``<value>`` is the |
313 |
user-provided value of the parameter, and ``<visibility>`` is either ``public``, |
314 |
``private`` or ``secret``. |
315 |
|
316 |
The installation scripts to be run inside the virtualized environment while the |
317 |
instance is run in ``install`` mode will be available at:: |
318 |
|
319 |
http://169.254.169.254/<version>/ganeti/os/scripts/<script_name> |
320 |
|
321 |
where ``<script_name>`` is the name of the script. |
322 |
|
323 |
|
324 |
Rationale |
325 |
--------- |
326 |
|
327 |
The choice of using a network interface for instance-host communication, as |
328 |
opposed to VirtIO, XenBus or other methods, is due to the will of having a |
329 |
generic, hypervisor-independent way of creating a communication channel, that |
330 |
doesn't require unusual (para)virtualization drivers. |
331 |
At the same time, a network interface was preferred over solutions involving |
332 |
virtual floppy or USB devices because the latter tend to be detected and |
333 |
configured by the guest operating systems, sometimes even in prominent positions |
334 |
in the user interface, whereas it is fairly common to have an unconfigured |
335 |
network interface in a system, usually without any negative side effects. |
336 |
|
337 |
|
338 |
Installation process in a virtualized environment |
339 |
+++++++++++++++++++++++++++++++++++++++++++++++++ |
340 |
|
341 |
In the new OS installation scenario, we distinguish between trusted and |
342 |
untrusted code. |
343 |
|
344 |
The trusted installation code maintains the behavior of the current one and |
345 |
requires no modifications, with the scripts running on the node the instance is |
346 |
being created on. The untrusted code is stored in a subdirectory of the OS |
347 |
definition called ``untrusted``. This directory contains scripts that are |
348 |
equivalent to the already existing ones (``create``, ``export``, ``import``, |
349 |
``rename``) but that will be run inside an virtualized environment, to protect |
350 |
the host from malicious tampering. |
351 |
|
352 |
The ``untrusted`` code is meant to either be untrusted itself, or to be trusted |
353 |
code running operations that might be dangerous (such as mounting a |
354 |
user-provided image). |
355 |
|
356 |
By default, all new OS definitions will have to be explicitly marked as trusted |
357 |
by the cluster administrator (with a new ``gnt-os modify`` command) before they |
358 |
can run code on the host. Otherwise, only the untrusted part of the code will be |
359 |
allowed to run, inside the virtual appliance. For backwards compatibility |
360 |
reasons, when upgrading an existing cluster, all the installed OSes will be |
361 |
marked as trusted, so that they can keep running with no changes. |
362 |
|
363 |
In order to allow for the highest flexibility, if both a trusted and an |
364 |
untrusted script are provided for the same operation (i.e. ``create``), both of |
365 |
them will be executed at the same time, one on the host, and one inside the |
366 |
installation appliance. They will be allowed to communicate with each other |
367 |
through the already described communication mechanism, in order to orchestrate |
368 |
their execution (e.g.: the untrusted code might execute the installation, while |
369 |
the trusted one receives status updates from it and delivers them to a user |
370 |
interface). |
371 |
|
372 |
The cluster administrator will have an option to completely disable scripts |
373 |
running on the host, leaving only the ones running in the VM. |
374 |
|
375 |
Ganeti will provide a script to be run at install time that can be used to |
376 |
create the virtualized environment that will perform the OS installation of new |
377 |
instances. |
378 |
This script will build a debootstrapped basic debian system including a software |
379 |
that will read the metadata, setup the environment variables and launch the |
380 |
installation scripts inside the virtualized environment. The script will also |
381 |
provide hooks for personalization. |
382 |
|
383 |
It will also be possible to use other self-made virtualized environments, as |
384 |
long as they connect to Ganeti over the described communication mechanism and |
385 |
they know how to read and use the provided metadata to create a new instance. |
386 |
|
387 |
While performing an installation in the virtualized environment, a |
388 |
personalizable timeout will be used to detect possible problems with the |
389 |
installation process, and to kill the virtualized environment. The timeout will |
390 |
be optional and set on a cluster basis by the administrator. If set, it will be |
391 |
the total time allowed to setup an instance inside the appliance. It is mainly |
392 |
meant as a safety measure to prevent an instance taken over by malicious scripts |
393 |
to be available for a long time. |
394 |
|
395 |
.. vim: set textwidth=72 : |
396 |
.. Local Variables: |
397 |
.. mode: rst |
398 |
.. fill-column: 72 |
399 |
.. End: |