1 Ganeti 2.0 OS Interface
2 =======================
9 We want to update the Ganeti OS Interface, which allows different operating
10 systems to be installed on instances in a Ganeti cluster. This interface has
11 been designed for Ganeti 1.2, and needs some changes to be compatible with the
12 changes introduced in Ganeti 2.0 (for a more through discussion of those please
13 check the Cluster Parameters design doc).
19 The current Ganeti OS interface, version 5, is tailored for Ganeti 1.2. The
20 interface is composed by a series of scripts which get called with certain
21 parameters to perform OS-dependent operations on the cluster. The current
25 called when a new instance is added to the cluster
27 called to export an instance disk to a stream
29 called to import from a stream to a new instance
31 called to perform the os-specific operations necessary for renaming an
34 Currently these scripts suffer from the limitations of Ganeti 1.2: for example
35 they accept exactly one block and one swap devices to operate on, rather than
36 any amount of generic block devices, they blindly assume that an instance will
37 have just one network interface to operate, they can not be configured to
38 optimise the instance for a particular hypervisor.
40 Since in Ganeti 2.0 we want to support multiple hypervisors, and a non-fixed
41 number of network and disks the OS interface need to change to transmit the
42 appropriate amount of information about an instance to its managing operating
43 system, when operating on it. Moreover since some old assumptions usually used
44 in OS scripts are no longer valid we need to re-establish a common knowledge on
45 what can be assumed and what cannot be regarding Ganeti environment.
51 When designing the new OS API our priorities are:
53 - future extensibility
54 - ease of porting from the old api
57 As such we want to limit the number of scripts that must be written to support
58 an OS, and make it easy to share code between them by uniforming their input.
59 We also will leave the current script structure unchanged, as far as we can,
60 and make a few of the scripts (import, export and rename) optional. Most
61 information will be passed to the script through environment variables, for
62 ease of access and at the same time ease of using only the information a script
72 As in Ganeti 1.2, every OS which wants to be installed in Ganeti needs to
73 support the following functionality, through scripts:
76 used to create a new instance running that OS. This script should prepare the
77 block devices, and install them so that the new OS can boot under the
80 used to export an installed instance using the given OS to a format which can
81 be used to import it back into a new instance.
83 used to import an exported instance into a new one. This script is similar to
84 create, but the new instance should have the content of the export, rather
85 than contain a pristine installation.
87 used to perform the internal OS-specific operations needed to rename an
90 If any optional script is not implemented Ganeti will refuse to perform the
91 given operation on instances using the non-implementing OS. Of course the
92 create script is mandatory, and it doesn't make sense to support the either the
93 export or the import operation but not both.
95 Incompatibilities with 1.2
96 ~~~~~~~~~~~~~~~~~~~~~~~~~~
98 We expect the following incompatibilities between the OS scripts for 1.2 and
101 - Input parameters: in 1.2 those were passed on the command line, in 2.0 we'll
102 use environment variables, as there will be a lot more information and not
103 all OSes may care about all of it.
104 - Input/Output: in 1.2 import scripts accepted the instance contents from
105 standard input, and export scripts exported it to standard output. This can't
106 be done in 2.0 as we can have more than one disk and so we wouldn't have a
107 way to distinguish them. A target/source node and directory will be instead
108 passed, for the import script to do the correct job. All scripts will get
109 nothing in input and are supposed to output only user-relevant messages.
110 - Some scripts are not compulsory: if such a script is missing the relevant
111 operations will be forbidden for instances of that os. This makes it easier
112 to distinguish between unsupported operations and no-op ones (if any).
118 Rather than using command line flags, as they do now, scripts will accept
119 inputs from environment variables. We expect the following input values:
122 The version of the OS api that the following parameters comply with;
123 this is used so that in the future we could have OSes supporting
124 multiple versions and thus Ganeti send the proper version in this
127 Name of the instance acted on
129 The hypervisor the instance should run on (eg. 'xen-pvm', 'xen-hvm', 'kvm')
131 The number of disks this instance will have
133 The number of nics this instance will have
135 Path to the Nth disk.
137 W if read/write, R if read only. OS scripts are not supposed to touch
138 read-only disks, but will be passed them to know.
139 DISK_<N>_FRONTEND_TYPE
140 Type of the disk as seen by the instance. Can be 'scsi', 'ide', 'virtio'
141 DISK_<N>_BACKEND_TYPE
142 Type of the disk as seen from the node. Can be 'block', 'file:loop' or
145 Mac address for the Nth network interface
147 Ip address for the Nth network interface, if available
149 Node bridge the Nth network interface will be connected to
150 NIC_<N>_FRONTEND_TYPE
151 Type of the Nth nic as seen by the instance. For example 'virtio', 'rtl8139', etc.
153 Whether more out should be produced, for debugging purposes. Currently the
154 only valid values are 0 and 1.
156 These are only the basic variables we are thinking of now, but more may come
157 during the implementation and they will be documented in the ganeti-os-api man
158 page. All these variables will be available to all scripts.
160 Some scripts will need a few more information to work. These will have
161 per-script variables, such as for example:
164 rename: the name the instance should be renamed from.
166 import/export: node where the export should be saved to or sourced from.
168 import/export: the place where exports are/should be saved to.
170 import/export: the type of compression for the exports.
172 (Rationale for the EXPORT_NODE, EXPORT_PATH decision: giving always a local
173 path would require either copying exports around, which takes a lot of
174 time/resources or depending on some type of remote filesystem to mount
175 resources on different nodes, which would add even more complexity to the
176 ganeti code, and can be added later, if really needed.)
178 (Rationale for INSTANCE_NAME as an environment variable: the instance name is
179 always needed and we could pass it on the command line. On the other hand,
180 though, this would force scripts to both access the environment and parse the
181 command line, so we'll move it for uniformity.)
187 As discussed scripts should only send user-targeted information to stdout. The
188 create and import scripts are supposed to format/initialise the given block
189 devices and install the correct instance data. The export script is supposed to
190 export instance data to a target node in a format understandable by the the
191 import script. The rename script should only modify the instance's knowledge of
194 Other declarative style features
195 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
197 Similar to Ganeti 1.2, OS specifications will need to provide a
198 'ganeti_api_version' containing list of numbers matching the version(s) of the
199 api they implement. Ganeti itself will always be compatible with one version of
200 the API and may maintain retrocompatibility if it's feasible to do so. The
201 numbers are one-per-line, so an OS supporting both version 5 and version 20
202 will have a file containing two lines. This is different from Ganeti 1.2, which
203 only supported one version number.
205 In addition to that an OS will be able to declare that it does support only a
206 subset of the ganeti hypervisors, by declaring them in the 'hypervisors' file.
212 We might want to have a "default" import/export behaviour that just dumps all
213 disks and restores them. This can save work as most systems will just do this,
214 while allowing flexibility for different systems.
216 Environment variables are limited in size, but we expect that there will be
217 enough space to store the information we need. If we discover that this is not
218 the case we may want to go to a more complex API such as storing those
219 information on the filesystem and providing the OS script with the path to a
220 file where they are encoded in some format.