code.grnet.gr Git - ganeti-local/blob - doc/hooks.sgml

   1 <!DOCTYPE article PUBLIC "-//OASIS//DTD DocBook V4.2//EN" [
   2 ]>
   3   <article class="specification">
   4   <articleinfo>
   5     <title>Ganeti customisation using hooks</title>
   6   </articleinfo>
   7   <para>Documents ganeti version 1.2</para>
   8   <section>
   9     <title>Introduction</title>
  10
  11     <para>
  12       In order to allow customisation of operations, ganeti will run
  13       scripts under <filename
  14       class="directory">/etc/ganeti/hooks</filename> based on certain
  15       rules.
  16     </para>
  17
  18       <para>This is similar to the <filename
  19       class="directory">/etc/network/</filename> structure present in
  20       Debian for network interface handling.</para>
  21
  22     </section>
  23
  24     <section>
  25       <title>Organisation</title>
  26
  27       <para>For every operation, two sets of scripts are run:
  28
  29       <itemizedlist>
  30           <listitem>
  31             <simpara>pre phase (for authorization/checking)</simpara>
  32           </listitem>
  33           <listitem>
  34             <simpara>post phase (for logging)</simpara>
  35           </listitem>
  36         </itemizedlist>
  37       </para>
  38
  39       <para>Also, for each operation, the scripts are run on one or
  40       more nodes, depending on the operation type.</para>
  41
  42       <para>Note that, even though we call them scripts, we are
  43       actually talking about any executable.</para>
  44
  45       <section>
  46         <title><emphasis>pre</emphasis> scripts</title>
  47
  48         <para>The <emphasis>pre</emphasis> scripts have a definite
  49         target: to check that the operation is allowed given the
  50         site-specific constraints. You could have, for example, a rule
  51         that says every new instance is required to exists in a
  52         database; to implement this, you could write a script that
  53         checks the new instance parameters against your
  54         database.</para>
  55
  56         <para>The objective of these scripts should be their return
  57         code (zero or non-zero for success and failure). However, if
  58         they modify the environment in any way, they should be
  59         idempotent, as failed executions could be restarted and thus
  60         the script(s) run again with exactly the same
  61         parameters.</para>
  62
  63       <para>
  64         Note that if a node is unreachable at the time a hooks is run,
  65         this will not be interpreted as a deny for the execution. In
  66         other words, only an actual error returned from a script will
  67         cause abort, and not an unreachable node.
  68       </para>
  69
  70       <para>
  71         Therefore, if you want to guarantee that a hook script is run
  72         and denies an action, it's best to put it on the master node.
  73       </para>
  74
  75       </section>
  76
  77       <section>
  78         <title><emphasis>post</emphasis> scripts</title>
  79
  80         <para>These scripts should do whatever you need as a reaction
  81         to the completion of an operation. Their return code is not
  82         checked (but logged), and they should not depend on the fact
  83         that the <emphasis>pre</emphasis> scripts have been
  84         run.</para>
  85
  86       </section>
  87
  88       <section>
  89         <title>Naming</title>
  90
  91         <para>The allowed names for the scripts consist of (similar to
  92         <citerefentry> <refentrytitle>run-parts</refentrytitle>
  93         <manvolnum>8</manvolnum> </citerefentry>) upper and lower
  94         case, digits, underscores and hyphens. In other words, the
  95         regexp
  96         <computeroutput>^[a-zA-Z0-9_-]+$</computeroutput>. Also,
  97         non-executable scripts will be ignored.
  98         </para>
  99       </section>
 100
 101       <section>
 102         <title>Order of execution</title>
 103
 104         <para>On a single node, the scripts in a directory are run in
 105         lexicographic order (more exactly, the python string
 106         comparison order). It is advisable to implement the usual
 107         <emphasis>NN-name</emphasis> convention where
 108         <emphasis>NN</emphasis> is a two digit number.</para>
 109
 110         <para>For an operation whose hooks are run on multiple nodes,
 111         there is no specific ordering of nodes with regard to hooks
 112         execution; you should assume that the scripts are run in
 113         parallel on the target nodes (keeping on each node the above
 114         specified ordering).  If you need any kind of inter-node
 115         synchronisation, you have to implement it yourself in the
 116         scripts.</para>
 117
 118       </section>
 119
 120       <section>
 121         <title>Execution environment</title>
 122
 123         <para>The scripts will be run as follows:
 124           <itemizedlist>
 125             <listitem>
 126               <simpara>no command line arguments</simpara>
 127             </listitem>
 128             <listitem>
 129               <simpara>no controlling <acronym>tty</acronym></simpara>
 130             </listitem>
 131             <listitem>
 132               <simpara><varname>stdin</varname> is
 133                 actually <filename>/dev/null</filename></simpara>
 134             </listitem>
 135             <listitem>
 136               <simpara><varname>stdout</varname> and
 137                 <varname>stderr</varname> are directed to
 138                 files</simpara>
 139             </listitem>
 140             <listitem>
 141               <simpara>the <varname>PATH</varname> is reset to
 142                 <literal>/sbin:/bin:/usr/sbin:/usr/bin</literal></simpara>
 143             </listitem>
 144             <listitem>
 145               <simpara>the environment is cleared, and only
 146                 ganeti-specific variables will be left</simpara>
 147             </listitem>
 148           </itemizedlist>
 149
 150         </para>
 151
 152       <para>All informations about the cluster is passed using
 153       environment variables. Different operations will have sligthly
 154       different environments, but most of the variables are
 155       common.</para>
 156
 157     </section>
 158
 159
 160     <section>
 161       <title>Operation list</title>
 162       <table>
 163         <title>Operation list</title>
 164         <tgroup cols="7">
 165           <colspec>
 166           <colspec>
 167           <colspec>
 168           <colspec>
 169           <colspec>
 170           <colspec colname="prehooks">
 171           <colspec colname="posthooks">
 172           <spanspec namest="prehooks" nameend="posthooks"
 173             spanname="bothhooks">
 174           <thead>
 175             <row>
 176               <entry>Operation ID</entry>
 177               <entry>Directory prefix</entry>
 178               <entry>Description</entry>
 179               <entry>Command</entry>
 180               <entry>Supported env. variables</entry>
 181               <entry><emphasis>pre</emphasis> hooks</entry>
 182               <entry><emphasis>post</emphasis> hooks</entry>
 183             </row>
 184           </thead>
 185           <tbody>
 186             <row>
 187               <entry>OP_INIT_CLUSTER</entry>
 188               <entry><filename class="directory">cluster-init</filename></entry>
 189               <entry>Initialises the cluster</entry>
 190               <entry><computeroutput>gnt-cluster init</computeroutput></entry>
 191               <entry><constant>CLUSTER</constant>, <constant>MASTER</constant></entry>
 192               <entry spanname="bothhooks">master node, cluster name</entry>
 193             </row>
 194             <row>
 195               <entry>OP_MASTER_FAILOVER</entry>
 196               <entry><filename class="directory">master-failover</filename></entry>
 197               <entry>Changes the master</entry>
 198               <entry><computeroutput>gnt-cluster master-failover</computeroutput></entry>
 199               <entry><constant>OLD_MASTER</constant>, <constant>NEW_MASTER</constant></entry>
 200               <entry>the new master</entry>
 201               <entry>all nodes</entry>
 202             </row>
 203             <row>
 204               <entry>OP_ADD_NODE</entry>
 205               <entry><filename class="directory">node-add</filename></entry>
 206               <entry>Adds a new node to the cluster</entry>
 207               <entry><computeroutput>gnt-node add</computeroutput></entry>
 208               <entry><constant>NODE_NAME</constant>, <constant>NODE_PIP</constant>, <constant>NODE_SIP</constant></entry>
 209               <entry>all existing nodes</entry>
 210               <entry>all existing nodes plus the new node</entry>
 211             </row>
 212             <row>
 213               <entry>OP_REMOVE_NODE</entry>
 214               <entry><filename class="directory">node-remove</filename></entry>
 215               <entry>Removes a node from the cluster</entry>
 216               <entry><computeroutput>gnt-node remove</computeroutput></entry>
 217               <entry><constant>NODE_NAME</constant></entry>
 218               <entry spanname="bothhooks">all existing nodes except the removed node</entry>
 219             </row>
 220             <row>
 221               <entry>OP_INSTANCE_ADD</entry>
 222               <entry><filename class="directory">instance-add</filename></entry>
 223               <entry>Creates a new instance</entry>
 224               <entry><computeroutput>gnt-instance add</computeroutput></entry>
 225               <entry><constant>INSTANCE_NAME</constant>, <constant>INSTANCE_PRIMARY</constant>, <constant>INSTANCE_SECONDARIES</constant>, <constant>DISK_TEMPLATE</constant>, <constant>MEM_SIZE</constant>, <constant>DISK_SIZE</constant>, <constant>SWAP_SIZE</constant>, <constant>VCPUS</constant>, <constant>INSTANCE_IP</constant>, <constant>INSTANCE_ADD_MODE</constant>, <constant>SRC_NODE</constant>, <constant>SRC_PATH</constant>, <constant>SRC_IMAGE</constant></entry>
 226               <entry spanname="bothhooks" morerows="4">master node, primary and
 227                    secondary nodes</entry>
 228             </row>
 229             <row>
 230               <entry>OP_BACKUP_EXPORT</entry>
 231               <entry><filename class="directory">instance-export</filename></entry>
 232               <entry>Export the instance</entry>
 233               <entry><computeroutput>gnt-backup export</computeroutput></entry>
 234               <entry><constant>INSTANCE_NAME</constant>, <constant>EXPORT_NODE</constant>, <constant>EXPORT_DO_SHUTDOWN</constant></entry>
 235             </row>
 236             <row>
 237               <entry>OP_INSTANCE_START</entry>
 238               <entry><filename class="directory">instance-start</filename></entry>
 239               <entry>Starts an instance</entry>
 240               <entry><computeroutput>gnt-instance start</computeroutput></entry>
 241               <entry><constant>INSTANCE_NAME</constant>, <constant>INSTANCE_PRIMARY</constant>, <constant>INSTANCE_SECONDARIES</constant>, <constant>FORCE</constant></entry>
 242             </row>
 243             <row>
 244               <entry>OP_INSTANCE_SHUTDOWN</entry>
 245               <entry><filename class="directory">instance-shutdown</filename></entry>
 246               <entry>Stops an instance</entry>
 247               <entry><computeroutput>gnt-instance shutdown</computeroutput></entry>
 248               <entry><constant>INSTANCE_NAME</constant>, <constant>INSTANCE_PRIMARY</constant>, <constant>INSTANCE_SECONDARIES</constant></entry>
 249             </row>
 250             <row>
 251               <entry>OP_INSTANCE_MODIFY</entry>
 252               <entry><filename class="directory">instance-modify</filename></entry>
 253               <entry>Modifies the instance parameters.</entry>
 254               <entry><computeroutput>gnt-instance modify</computeroutput></entry>
 255               <entry><constant>INSTANCE_NAME</constant>, <constant>MEM_SIZE</constant>, <constant>VCPUS</constant>, <constant>INSTANCE_IP</constant></entry>
 256             </row>
 257             <row>
 258               <entry>OP_INSTANCE_FAILOVER</entry>
 259               <entry><filename class="directory">instance-failover</filename></entry>
 260               <entry>Failover an instance</entry>
 261               <entry><computeroutput>gnt-instance start</computeroutput></entry>
 262               <entry><constant>INSTANCE_NAME</constant>, <constant>INSTANCE_PRIMARY</constant>, <constant>INSTANCE_SECONDARIES</constant>, <constant>IGNORE_CONSISTENCY</constant></entry>
 263             </row>
 264             <row>
 265               <entry>OP_INSTANCE_REMOVE</entry>
 266               <entry><filename class="directory">instance-remove</filename></entry>
 267               <entry>Remove an instance</entry>
 268               <entry><computeroutput>gnt-instance remove</computeroutput></entry>
 269               <entry><constant>INSTANCE_NAME</constant>, <constant>INSTANCE_PRIMARY</constant>, <constant>INSTANCE_SECONDARIES</constant></entry>
 270               <entry spanname="bothhooks">master node</entry>
 271             </row>
 272             <row>
 273               <entry>OP_INSTANCE_ADD_MDDRBD</entry>
 274               <entry><filename class="directory">mirror-add</filename></entry>
 275               <entry>Adds a mirror component</entry>
 276               <entry><computeroutput>gnt-instance add-mirror</computeroutput></entry>
 277               <entry><constant>INSTANCE_NAME</constant>, <constant>NEW_SECONDARY</constant>, <constant>DISK_NAME</constant></entry>
 278             </row>
 279             <row>
 280               <entry>OP_INSTANCE_REMOVE_MDDRBD</entry>
 281               <entry><filename class="directory">mirror-remove</filename></entry>
 282               <entry>Removes a mirror component</entry>
 283               <entry><computeroutput>gnt-instance remove-mirror</computeroutput></entry>
 284               <entry><constant>INSTANCE_NAME</constant>, <constant>OLD_SECONDARY</constant>, <constant>DISK_NAME</constant>, <constant>DISK_ID</constant></entry>
 285             </row>
 286             <row>
 287               <entry>OP_INSTANCE_REPLACE_DISKS</entry>
 288               <entry><filename class="directory">mirror-replace</filename></entry>
 289               <entry>Replace all mirror components</entry>
 290               <entry><computeroutput>gnt-instance replace-disks</computeroutput></entry>
 291               <entry><constant>INSTANCE_NAME</constant>, <constant>OLD_SECONDARY</constant>, <constant>NEW_SECONDARY</constant></entry>
 292
 293             </row>
 294           </tbody>
 295         </tgroup>
 296       </table>
 297     </section>
 298
 299     <section>
 300       <title>Environment variables</title>
 301
 302       <para>Note that all variables listed here are actually prefixed
 303       with <constant>GANETI_</constant> in order to provide a
 304       different namespace.</para>
 305
 306       <section>
 307         <title>Common variables</title>
 308
 309         <para>This is the list of environment variables supported by
 310         all operations:</para>
 311
 312         <variablelist>
 313           <varlistentry>
 314             <term>HOOKS_VERSION</term>
 315             <listitem>
 316               <para>Documents the hooks interface version. In case this
 317             doesnt match what the script expects, it should not
 318             run. The documents conforms to the version
 319             <literal>1</literal>.</para>
 320             </listitem>
 321           </varlistentry>
 322           <varlistentry>
 323             <term>HOOKS_PHASE</term>
 324             <listitem>
 325               <para>one of <constant>PRE</constant> or
 326               <constant>POST</constant> denoting which phase are we
 327               in.</para>
 328             </listitem>
 329           </varlistentry>
 330           <varlistentry>
 331             <term>CLUSTER</term>
 332             <listitem>
 333               <para>the cluster name</para>
 334             </listitem>
 335           </varlistentry>
 336           <varlistentry>
 337             <term>MASTER</term>
 338             <listitem>
 339               <para>the master node</para>
 340             </listitem>
 341           </varlistentry>
 342           <varlistentry>
 343             <term>OP_ID</term>
 344             <listitem>
 345               <para>one of the <constant>OP_*</constant> values from
 346               the table of operations</para>
 347             </listitem>
 348           </varlistentry>
 349           <varlistentry>
 350             <term>OBJECT_TYPE</term>
 351             <listitem>
 352               <para>one of <simplelist type="inline">
 353                   <member><constant>INSTANCE</constant></member>
 354                   <member><constant>NODE</constant></member>
 355                   <member><constant>CLUSTER</constant></member>
 356                 </simplelist>, showing the target of the operation.
 357              </para>
 358             </listitem>
 359           </varlistentry>
 360           <!-- commented out since it causes problems in our rpc
 361                multi-node optimised calls
 362           <varlistentry>
 363             <term>HOST_NAME</term>
 364             <listitem>
 365               <para>The name of the node the hook is run on as known by
 366             the cluster.</para>
 367             </listitem>
 368           </varlistentry>
 369           <varlistentry>
 370             <term>HOST_TYPE</term>
 371             <listitem>
 372               <para>one of <simplelist type="inline">
 373                   <member><constant>MASTER</constant></member>
 374                   <member><constant>NODE</constant></member>
 375                 </simplelist>, showing the role of this node in the cluster.
 376              </para>
 377             </listitem>
 378           </varlistentry>
 379           -->
 380         </variablelist>
 381       </section>
 382
 383       <section>
 384         <title>Specialised variables</title>
 385
 386         <para>This is the list of variables which are specific to one
 387         or more operations.</para>
 388         <variablelist>
 389           <varlistentry>
 390             <term>INSTANCE_NAME</term>
 391             <listitem>
 392               <para>The name of the instance which is the target of
 393               the operation.</para>
 394             </listitem>
 395           </varlistentry>
 396           <varlistentry>
 397             <term>INSTANCE_DISK_TYPE</term>
 398             <listitem>
 399               <para>The disk type for the instance.</para>
 400             </listitem>
 401           </varlistentry>
 402           <varlistentry>
 403             <term>INSTANCE_DISK_SIZE</term>
 404             <listitem>
 405               <para>The (OS) disk size for the instance.</para>
 406             </listitem>
 407           </varlistentry>
 408           <varlistentry>
 409             <term>INSTANCE_OS</term>
 410             <listitem>
 411               <para>The name of the instance OS.</para>
 412             </listitem>
 413           </varlistentry>
 414           <varlistentry>
 415             <term>INSTANCE_PRIMARY</term>
 416             <listitem>
 417               <para>The name of the node which is the primary for the
 418               instance.</para>
 419             </listitem>
 420           </varlistentry>
 421           <varlistentry>
 422             <term>INSTANCE_SECONDARIES</term>
 423             <listitem>
 424               <para>Space-separated list of secondary nodes for the
 425               instance.</para>
 426             </listitem>
 427           </varlistentry>
 428           <varlistentry>
 429             <term>NODE_NAME</term>
 430             <listitem>
 431               <para>The target node of this operation (not the node on
 432               which the hook runs).</para>
 433             </listitem>
 434           </varlistentry>
 435           <varlistentry>
 436             <term>NODE_PIP</term>
 437             <listitem>
 438               <para>The primary IP of the target node (the one over
 439               which inter-node communication is done).</para>
 440             </listitem>
 441           </varlistentry>
 442           <varlistentry>
 443             <term>NODE_SIP</term>
 444             <listitem>
 445               <para>The secondary IP of the target node (the one over
 446               which drbd replication is done). This can be equal to
 447               the primary ip, in case the cluster is not
 448               dual-homed.</para>
 449             </listitem>
 450           </varlistentry>
 451           <varlistentry>
 452             <term>OLD_MASTER</term>
 453             <term>NEW_MASTER</term>
 454             <listitem>
 455               <para>The old, respectively the new master for the
 456               master failover operation.</para>
 457             </listitem>
 458           </varlistentry>
 459           <varlistentry>
 460             <term>FORCE</term>
 461             <listitem>
 462               <para>This is provided by some operations when the user
 463               gave this flag.</para>
 464             </listitem>
 465           </varlistentry>
 466           <varlistentry>
 467             <term>IGNORE_CONSISTENCY</term>
 468             <listitem>
 469               <para>The user has specified this flag. It is used when
 470               failing over instances in case the primary node is
 471               down.</para>
 472             </listitem>
 473           </varlistentry>
 474           <varlistentry>
 475             <term>MEM_SIZE, DISK_SIZE, SWAP_SIZE, VCPUS</term>
 476             <listitem>
 477               <para>The memory, disk, swap size and the number of
 478               processor selected for the instance (in
 479               <command>gnt-instance add</command> or
 480               <command>gnt-instance modify</command>).</para>
 481             </listitem>
 482           </varlistentry>
 483           <varlistentry>
 484             <term>INSTANCE_IP</term>
 485             <listitem>
 486               <para>If defined, the instance IP in the
 487               <command>gnt-instance add</command> and
 488               <command>gnt-instance set</command> commands. If not
 489               defined, it means that no IP has been defined.</para>
 490             </listitem>
 491           </varlistentry>
 492           <varlistentry>
 493             <term>DISK_TEMPLATE</term>
 494             <listitem>
 495               <para>The disk template type when creating the instance.</para>
 496             </listitem>
 497           </varlistentry>
 498           <varlistentry>
 499             <term>INSTANCE_ADD_MODE</term>
 500             <listitem>
 501               <para>The mode of the create: either
 502               <constant>create</constant> for create from scratch or
 503               <constant>import</constant> for restoring from an
 504               exported image.</para>
 505             </listitem>
 506           </varlistentry>
 507           <varlistentry>
 508             <term>SRC_NODE, SRC_PATH, SRC_IMAGE</term>
 509             <listitem>
 510               <para>In case the instance has been added by import,
 511               these variables are defined and point to the source
 512               node, source path (the directory containing the image
 513               and the config file) and the source disk image
 514               file.</para>
 515             </listitem>
 516           </varlistentry>
 517           <varlistentry>
 518             <term>DISK_NAME</term>
 519             <listitem>
 520               <para>The disk name (either <filename>sda</filename> or
 521               <filename>sdb</filename>) in mirror operations
 522               (add/remove mirror).</para>
 523             </listitem>
 524           </varlistentry>
 525           <varlistentry>
 526             <term>DISK_ID</term>
 527             <listitem>
 528               <para>The disk id for mirror remove operations. You can
 529               look this up using <command>gnt-instance
 530               info</command>.</para>
 531             </listitem>
 532           </varlistentry>
 533           <varlistentry>
 534             <term>NEW_SECONDARY</term>
 535             <listitem>
 536               <para>The name of the node on which the new mirror
 537               componet is being added. This can be the name of the
 538               current secondary, if the new mirror is on the same
 539               secondary.</para>
 540             </listitem>
 541           </varlistentry>
 542           <varlistentry>
 543             <term>OLD_SECONDARY</term>
 544             <listitem>
 545               <para>The name of the old secondary. This is used in
 546               both <command>replace-disks</command> and
 547               <command>remove-mirror</command>. Note that this can be
 548               equal to the new secondary (only
 549               <command>replace-disks</command> has both variables) if
 550               the secondary node hasn't actually changed).</para>
 551             </listitem>
 552           </varlistentry>
 553           <varlistentry>
 554             <term>EXPORT_NODE</term>
 555             <listitem>
 556               <para>The node on which the exported image of the
 557               instance was done.</para>
 558             </listitem>
 559           </varlistentry>
 560           <varlistentry>
 561             <term>EXPORT_DO_SHUTDOWN</term>
 562             <listitem>
 563               <para>This variable tells if the instance has been
 564               shutdown or not while doing the export. In the "was
 565               shutdown" case, it's likely that the filesystem is
 566               consistent, whereas in the "did not shutdown" case, the
 567               filesystem would need a check (journal replay or full
 568               fsck) in order to guarantee consistency.</para>
 569             </listitem>
 570           </varlistentry>
 571         </variablelist>
 572
 573       </section>
 574
 575     </section>
 576
 577   </section>
 578   </article>