5 This document describes the major changes in Ganeti 2.1 compared to
8 The 2.1 version will be a relatively small release. Its main aim is to
9 avoid changing too much of the core code, while addressing issues and
10 adding new features and improvements over 2.0, in a timely fashion.
12 .. contents:: :depth: 4
17 Ganeti 2.1 will add features to help further automatization of cluster
18 operations, further improve scalability to even bigger clusters, and
19 make it easier to debug the Ganeti core.
30 As for 2.0 we divide the 2.1 design into three areas:
32 - core changes, which affect the master daemon/job queue/locking or
33 all/most logical units
34 - logical unit/feature changes
35 - external interface changes (eg. command line, os api, hooks, ...)
40 Storage units modelling
41 ~~~~~~~~~~~~~~~~~~~~~~~
43 Currently, Ganeti has a good model of the block devices for instances
44 (e.g. LVM logical volumes, files, DRBD devices, etc.) but none of the
45 storage pools that are providing the space for these front-end
46 devices. For example, there are hardcoded inter-node RPC calls for
47 volume group listing, file storage creation/deletion, etc.
49 The storage units framework will implement a generic handling for all
50 kinds of storage backends:
52 - LVM physical volumes
54 - File-based storage directories
55 - any other future storage method
57 There will be a generic list of methods that each storage unit type
60 - list of storage units of this type
61 - check status of the storage unit
63 Additionally, there will be specific methods for each method, for
66 - enable/disable allocations on a specific PV
67 - file storage directory creation/deletion
68 - VG consistency fixing
70 This will allow a much better modeling and unification of the various
71 RPC calls related to backend storage pool in the future. Ganeti 2.1 is
72 intended to add the basics of the framework, and not necessarilly move
73 all the curent VG/FileBased operations to it.
75 Note that while we model both LVM PVs and LVM VGs, the framework will
76 **not** model any relationship between the different types. In other
77 words, we don't model neither inheritances nor stacking, since this is
78 too complex for our needs. While a ``vgreduce`` operation on a LVM VG
79 could actually remove a PV from it, this will not be handled at the
80 framework level, but at individual operation level. The goal is that
81 this is a lightweight framework, for abstracting the different storage
82 operation, and not for modelling the storage hierarchy.
88 Current State and shortcomings
89 ++++++++++++++++++++++++++++++
91 The class ``LockSet`` (see ``lib/locking.py``) is a container for one or
92 many ``SharedLock`` instances. It provides an interface to add/remove
93 locks and to acquire and subsequently release any number of those locks
96 Locks in a ``LockSet`` are always acquired in alphabetic order. Due to
97 the way we're using locks for nodes and instances (the single cluster
98 lock isn't affected by this issue) this can lead to long delays when
99 acquiring locks if another operation tries to acquire multiple locks but
100 has to wait for yet another operation.
102 In the following demonstration we assume to have the instance locks
103 ``inst1``, ``inst2``, ``inst3`` and ``inst4``.
105 #. Operation A grabs lock for instance ``inst4``.
106 #. Operation B wants to acquire all instance locks in alphabetic order,
107 but it has to wait for ``inst4``.
108 #. Operation C tries to lock ``inst1``, but it has to wait until
109 Operation B (which is trying to acquire all locks) releases the lock
111 #. Operation A finishes and releases lock on ``inst4``. Operation B can
112 continue and eventually releases all locks.
113 #. Operation C can get ``inst1`` lock and finishes.
115 Technically there's no need for Operation C to wait for Operation A, and
116 subsequently Operation B, to finish. Operation B can't continue until
117 Operation A is done (it has to wait for ``inst4``), anyway.
122 Non-blocking lock acquiring
123 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
125 Acquiring locks for OpCode execution is always done in blocking mode.
126 They won't return until the lock has successfully been acquired (or an
127 error occurred, although we won't cover that case here).
129 ``SharedLock`` and ``LockSet`` must be able to be acquired in a
130 non-blocking way. They must support a timeout and abort trying to
131 acquire the lock(s) after the specified amount of time.
133 Retry acquiring locks
134 ^^^^^^^^^^^^^^^^^^^^^
136 To prevent other operations from waiting for a long time, such as
137 described in the demonstration before, ``LockSet`` must not keep locks
138 for a prolonged period of time when trying to acquire two or more locks.
139 Instead it should, with an increasing timeout for acquiring all locks,
140 release all locks again and sleep some time if it fails to acquire all
143 A good timeout value needs to be determined. In any case should
144 ``LockSet`` proceed to acquire locks in blocking mode after a few
145 (unsuccessful) attempts to acquire all requested locks.
147 One proposal for the timeout is to use ``2**tries`` seconds, where
148 ``tries`` is the number of unsuccessful tries.
150 In the demonstration before this would allow Operation C to continue
151 after Operation B unsuccessfully tried to acquire all locks and released
152 all acquired locks (``inst1``, ``inst2`` and ``inst3``) again.
154 Other solutions discussed
155 +++++++++++++++++++++++++
157 There was also some discussion on going one step further and extend the
158 job queue (see ``lib/jqueue.py``) to select the next task for a worker
159 depending on whether it can acquire the necessary locks. While this may
160 reduce the number of necessary worker threads and/or increase throughput
161 on large clusters with many jobs, it also brings many potential
162 problems, such as contention and increased memory usage, with it. As
163 this would be an extension of the changes proposed before it could be
164 implemented at a later point in time, but we decided to stay with the
165 simpler solution for now.
167 Implementation details
168 ++++++++++++++++++++++
170 ``SharedLock`` redesign
171 ^^^^^^^^^^^^^^^^^^^^^^^
173 The current design of ``SharedLock`` is not good for supporting timeouts
174 when acquiring a lock and there are also minor fairness issues in it. We
175 plan to address both with a redesign. A proof of concept implementation
176 was written and resulted in significantly simpler code.
178 Currently ``SharedLock`` uses two separate queues for shared and
179 exclusive acquires and waiters get to run in turns. This means if an
180 exclusive acquire is released, the lock will allow shared waiters to run
181 and vice versa. Although it's still fair in the end there is a slight
182 bias towards shared waiters in the current implementation. The same
183 implementation with two shared queues can not support timeouts without
184 adding a lot of complexity.
186 Our proposed redesign changes ``SharedLock`` to have only one single
187 queue. There will be one condition (see Condition_ for a note about
188 performance) in the queue per exclusive acquire and two for all shared
189 acquires (see below for an explanation). The maximum queue length will
190 always be ``2 + (number of exclusive acquires waiting)``. The number of
191 queue entries for shared acquires can vary from 0 to 2.
193 The two conditions for shared acquires are a bit special. They will be
194 used in turn. When the lock is instantiated, no conditions are in the
195 queue. As soon as the first shared acquire arrives (and there are
196 holder(s) or waiting acquires; see Acquire_), the active condition is
197 added to the queue. Until it becomes the topmost condition in the queue
198 and has been notified, any shared acquire is added to this active
199 condition. When the active condition is notified, the conditions are
200 swapped and further shared acquires are added to the previously inactive
201 condition (which has now become the active condition). After all waiters
202 on the previously active (now inactive) and now notified condition
203 received the notification, it is removed from the queue of pending
206 This means shared acquires will skip any exclusive acquire in the queue.
207 We believe it's better to improve parallelization on operations only
208 asking for shared (or read-only) locks. Exclusive operations holding the
209 same lock can not be parallelized.
215 For exclusive acquires a new condition is created and appended to the
216 queue. Shared acquires are added to the active condition for shared
217 acquires and if the condition is not yet on the queue, it's appended.
219 The next step is to wait for our condition to be on the top of the queue
220 (to guarantee fairness). If the timeout expired, we return to the caller
221 without acquiring the lock. On every notification we check whether the
222 lock has been deleted, in which case an error is returned to the caller.
224 The lock can be acquired if we're on top of the queue (there is no one
225 else ahead of us). For an exclusive acquire, there must not be other
226 exclusive or shared holders. For a shared acquire, there must not be an
227 exclusive holder. If these conditions are all true, the lock is
228 acquired and we return to the caller. In any other case we wait again on
231 If it was the last waiter on a condition, the condition is removed from
234 Optimization: There's no need to touch the queue if there are no pending
235 acquires and no current holders. The caller can have the lock
238 .. image:: design-2.1-lock-acquire.png
244 First the lock removes the caller from the internal owner list. If there
245 are pending acquires in the queue, the first (the oldest) condition is
248 If the first condition was the active condition for shared acquires, the
249 inactive condition will be made active. This ensures fairness with
250 exclusive locks by forcing consecutive shared acquires to wait in the
253 .. image:: design-2.1-lock-release.png
259 The caller must either hold the lock in exclusive mode already or the
260 lock must be acquired in exclusive mode. Trying to delete a lock while
261 it's held in shared mode must fail.
263 After ensuring the lock is held in exclusive mode, the lock will mark
264 itself as deleted and continue to notify all pending acquires. They will
265 wake up, notice the deleted lock and return an error to the caller.
271 Note: This is not necessary for the locking changes above, but it may be
272 a good optimization (pending performance tests).
274 The existing locking code in Ganeti 2.0 uses Python's built-in
275 ``threading.Condition`` class. Unfortunately ``Condition`` implements
276 timeouts by sleeping 1ms to 20ms between tries to acquire the condition
277 lock in non-blocking mode. This requires unnecessary context switches
278 and contention on the CPython GIL (Global Interpreter Lock).
280 By using POSIX pipes (see ``pipe(2)``) we can use the operating system's
281 support for timeouts on file descriptors (see ``select(2)``). A custom
282 condition class will have to be written for this.
284 On instantiation the class creates a pipe. After each notification the
285 previous pipe is abandoned and re-created (technically the old pipe
286 needs to stay around until all notifications have been delivered).
288 All waiting clients of the condition use ``select(2)`` or ``poll(2)`` to
289 wait for notifications, optionally with a timeout. A notification will
290 be signalled to the waiting clients by closing the pipe. If the pipe
291 wasn't closed during the timeout, the waiting function returns to its
295 Node daemon availability
296 ~~~~~~~~~~~~~~~~~~~~~~~~
298 Current State and shortcomings
299 ++++++++++++++++++++++++++++++
301 Currently, when a Ganeti node suffers serious system disk damage, the
302 migration/failover of an instance may not correctly shutdown the virtual
303 machine on the broken node causing instances duplication. The ``gnt-node
304 powercycle`` command can be used to force a node reboot and thus to
305 avoid duplicated instances. This command relies on node daemon
306 availability, though, and thus can fail if the node daemon has some
307 pages swapped out of ram, for example.
313 The proposed solution forces node daemon to run exclusively in RAM. It
314 uses python ctypes to to call ``mlockall(MCL_CURRENT | MCL_FUTURE)`` on
315 the node daemon process and all its children. In addition another log
316 handler has been implemented for node daemon to redirect to
317 ``/dev/console`` messages that cannot be written on the logfile.
319 With these changes node daemon can successfully run basic tasks such as
320 a powercycle request even when the system disk is heavily damaged and
321 reading/writing to disk fails constantly.
330 Current State and shortcomings
331 ++++++++++++++++++++++++++++++
333 In Ganeti 2.0 all nodes are equal, but some are more equal than others.
334 In particular they are divided between "master", "master candidates" and
335 "normal". (Moreover they can be offline or drained, but this is not
336 important for the current discussion). In general the whole
337 configuration is only replicated to master candidates, and some partial
338 information is spread to all nodes via ssconf.
340 This change was done so that the most frequent Ganeti operations didn't
341 need to contact all nodes, and so clusters could become bigger. If we
342 want more information to be available on all nodes, we need to add more
343 ssconf values, which is counter-balancing the change, or to talk with
344 the master node, which is not designed to happen now, and requires its
347 Information such as the instance->primary_node mapping will be needed on
348 all nodes, and we also want to make sure services external to the
349 cluster can query this information as well. This information must be
350 available at all times, so we can't query it through RAPI, which would
351 be a single point of failure, as it's only available on the master.
357 In order to allow fast and highly available access read-only to some
358 configuration values, we'll create a new ganeti-confd daemon, which will
359 run on master candidates. This daemon will talk via UDP, and
360 authenticate messages using HMAC with a cluster-wide shared key. This
361 key will be generated at cluster init time, and stored on the clusters
362 alongside the ganeti SSL keys, and readable only by root.
364 An interested client can query a value by making a request to a subset
365 of the cluster master candidates. It will then wait to get a few
366 responses, and use the one with the highest configuration serial number.
367 Since the configuration serial number is increased each time the ganeti
368 config is updated, and the serial number is included in all answers,
369 this can be used to make sure to use the most recent answer, in case
370 some master candidates are stale or in the middle of a configuration
373 In order to prevent replay attacks queries will contain the current unix
374 timestamp according to the client, and the server will verify that its
375 timestamp is in the same 5 minutes range (this requires synchronized
376 clocks, which is a good idea anyway). Queries will also contain a "salt"
377 which they expect the answers to be sent with, and clients are supposed
378 to accept only answers which contain salt generated by them.
380 The configuration daemon will be able to answer simple queries such as:
382 - master candidates list
386 - instance primary nodes
391 A confd query will look like this, on the wire::
394 "msg": "{\"type\": 1,
395 \"rsalt\": \"9aa6ce92-8336-11de-af38-001d093e835f\",
397 \"query\": \"node1.example.com\"}\n",
398 "salt": "1249637704",
399 "hmac": "4a4139b2c3c5921f7e439469a0a45ad200aead0f"
402 "plj0" is a fourcc that details the message content. It stands for plain
403 json 0, and can be changed as we move on to different type of protocols
404 (for example protocol buffers, or encrypted json). What follows is a
405 json encoded string, with the following fields:
407 - 'msg' contains a JSON-encoded query, its fields are:
409 - 'protocol', integer, is the confd protocol version (initially just
410 constants.CONFD_PROTOCOL_VERSION, with a value of 1)
411 - 'type', integer, is the query type. For example "node role by name"
412 or "node primary ip by instance ip". Constants will be provided for
413 the actual available query types.
414 - 'query', string, is the search key. For example an ip, or a node
416 - 'rsalt', string, is the required response salt. The client must use
417 it to recognize which answer it's getting.
419 - 'salt' must be the current unix timestamp, according to the client.
420 Servers can refuse messages which have a wrong timing, according to
421 their configuration and clock.
422 - 'hmac' is an hmac signature of salt+msg, with the cluster hmac key
424 If an answer comes back (which is optional, since confd works over UDP)
425 it will be in this format::
428 "msg": "{\"status\": 0,
432 "salt": "9aa6ce92-8336-11de-af38-001d093e835f",
433 "hmac": "aaeccc0dff9328fdf7967cb600b6a80a6a9332af"
438 - 'plj0' the message type magic fourcc, as discussed above
439 - 'msg' contains a JSON-encoded answer, its fields are:
441 - 'protocol', integer, is the confd protocol version (initially just
442 constants.CONFD_PROTOCOL_VERSION, with a value of 1)
443 - 'status', integer, is the error code. Initially just 0 for 'ok' or
444 '1' for 'error' (in which case answer contains an error detail,
445 rather than an answer), but in the future it may be expanded to have
446 more meanings (eg: 2, the answer is compressed)
447 - 'answer', is the actual answer. Its type and meaning is query
448 specific. For example for "node primary ip by instance ip" queries
449 it will be a string containing an IP address, for "node role by
450 name" queries it will be an integer which encodes the role (master,
451 candidate, drained, offline) according to constants.
453 - 'salt' is the requested salt from the query. A client can use it to
454 recognize what query the answer is answering.
455 - 'hmac' is an hmac signature of salt+msg, with the cluster hmac key
461 Current State and shortcomings
462 ++++++++++++++++++++++++++++++
464 Currently LURedistributeConfig triggers a copy of the updated
465 configuration file to all master candidates and of the ssconf files to
466 all nodes. There are other files which are maintained manually but which
467 are important to keep in sync. These are:
469 - rapi SSL key certificate file (rapi.pem) (on master candidates)
470 - rapi user/password file rapi_users (on master candidates)
472 Furthermore there are some files which are hypervisor specific but we
473 may want to keep in sync:
475 - the xen-hvm hypervisor uses one shared file for all vnc passwords, and
476 copies the file once, during node add. This design is subject to
477 revision to be able to have different passwords for different groups
478 of instances via the use of hypervisor parameters, and to allow
479 xen-hvm and kvm to use an equal system to provide password-protected
480 vnc sessions. In general, though, it would be useful if the vnc
481 password files were copied as well, to avoid unwanted vnc password
482 changes on instance failover/migrate.
484 Optionally the admin may want to also ship files such as the global
485 xend.conf file, and the network scripts to all nodes.
490 RedistributeConfig will be changed to copy also the rapi files, and to
491 call every enabled hypervisor asking for a list of additional files to
492 copy. Users will have the possibility to populate a file containing a
493 list of files to be distributed; this file will be propagated as well.
494 Such solution is really simple to implement and it's easily usable by
497 This code will be also shared (via tasklets or by other means, if
498 tasklets are not ready for 2.1) with the AddNode and SetNodeParams LUs
499 (so that the relevant files will be automatically shipped to new master
500 candidates as they are set).
505 Current State and shortcomings
506 ++++++++++++++++++++++++++++++
508 Currently just the xen-hvm hypervisor supports setting a password to
509 connect the the instances' VNC console, and has one common password
512 This doesn't allow different passwords for different instances/groups of
513 instances, and makes it necessary to remember to copy the file around
514 the cluster when the password changes.
519 We'll change the VNC password file to a vnc_password_file hypervisor
520 parameter. This way it can have a cluster default, but also a different
521 value for each instance. The VNC enabled hypervisors (xen and kvm) will
522 publish all the password files in use through the cluster so that a
523 redistribute-config will ship them to all nodes (see the Redistribute
524 Config proposed changes above).
526 The current VNC_PASSWORD_FILE constant will be removed, but its value
527 will be used as the default HV_VNC_PASSWORD_FILE value, thus retaining
528 backwards compatibility with 2.0.
530 The code to export the list of VNC password files from the hypervisors
531 to RedistributeConfig will be shared between the KVM and xen-hvm
537 Current State and shortcomings
538 ++++++++++++++++++++++++++++++
540 Currently disks and network interfaces have a few tweakable options and
541 all the rest is left to a default we chose. We're finding that we need
542 more and more to tweak some of these parameters, for example to disable
543 barriers for DRBD devices, or allow striping for the LVM volumes.
545 Moreover for many of these parameters it will be nice to have
546 cluster-wide defaults, and then be able to change them per
552 We will add new cluster level diskparams and netparams, which will
553 contain all the tweakable parameters. All values which have a sensible
554 cluster-wide default will go into this new structure while parameters
555 which have unique values will not.
557 Example of network parameters:
559 - link: for mode "bridge" the bridge to connect to, for mode route it
560 can contain the routing table, or the destination interface
562 Example of disk parameters:
563 - stripe: lvm stripes
564 - stripe_size: lvm stripe size
565 - meta_flushes: drbd, enable/disable metadata "barriers"
566 - data_flushes: drbd, enable/disable data "barriers"
568 Some parameters are bound to be disk-type specific (drbd, vs lvm, vs
569 files) or hypervisor specific (nic models for example), but for now they
570 will all live in the same structure. Each component is supposed to
571 validate only the parameters it knows about, and ganeti itself will make
572 sure that no "globally unknown" parameters are added, and that no
573 parameters have overridden meanings for different components.
575 The parameters will be kept, as for the BEPARAMS into a "default"
576 category, which will allow us to expand on by creating instance
577 "classes" in the future. Instance classes is not a feature we plan
578 implementing in 2.1, though.
581 Global hypervisor parameters
582 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
584 Current State and shortcomings
585 ++++++++++++++++++++++++++++++
587 Currently all hypervisor parameters are modifiable both globally
588 (cluster level) and at instance level. However, there is no other
589 framework to held hypervisor-specific parameters, so if we want to add
590 a new class of hypervisor parameters that only makes sense on a global
591 level, we have to change the hvparams framework.
596 We add a new (global, not per-hypervisor) list of parameters which are
597 not changeable on a per-instance level. The create, modify and query
598 instance operations are changed to not allow/show these parameters.
600 Furthermore, to allow transition of parameters to the global list, and
601 to allow cleanup of inadverdently-customised parameters, the
602 ``UpgradeConfig()`` method of instances will drop any such parameters
603 from their list of hvparams, such that a restart of the master daemon
604 is all that is needed for cleaning these up.
606 Also, the framework is simple enough that if we need to replicate it
607 at beparams level we can do so easily.
610 Non bridged instances support
611 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
613 Current State and shortcomings
614 ++++++++++++++++++++++++++++++
616 Currently each instance NIC must be connected to a bridge, and if the
617 bridge is not specified the default cluster one is used. This makes it
618 impossible to use the vif-route xen network scripts, or other
619 alternative mechanisms that don't need a bridge to work.
624 The new "mode" network parameter will distinguish between bridged
625 interfaces and routed ones.
627 When mode is "bridge" the "link" parameter will contain the bridge the
628 instance should be connected to, effectively making things as today. The
629 value has been migrated from a nic field to a parameter to allow for an
630 easier manipulation of the cluster default.
632 When mode is "route" the ip field of the interface will become
633 mandatory, to allow for a route to be set. In the future we may want
634 also to accept multiple IPs or IP/mask values for this purpose. We will
635 evaluate possible meanings of the link parameter to signify a routing
636 table to be used, which would allow for insulation between instance
637 groups (as today happens for different bridges).
639 For now we won't add a parameter to specify which network script gets
640 called for which instance, so in a mixed cluster the network script must
641 be able to handle both cases. The default kvm vif script will be changed
642 to do so. (Xen doesn't have a ganeti provided script, so nothing will be
643 done for that hypervisor)
645 Introducing persistent UUIDs
646 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
648 Current state and shortcomings
649 ++++++++++++++++++++++++++++++
651 Some objects in the Ganeti configurations are tracked by their name
652 while also supporting renames. This creates an extra difficulty,
653 because neither Ganeti nor external management tools can then track
654 the actual entity, and due to the name change it behaves like a new
657 Proposed changes part 1
658 +++++++++++++++++++++++
660 We will change Ganeti to use UUIDs for entity tracking, but in a
661 staggered way. In 2.1, we will simply add an “uuid” attribute to each
662 of the instances, nodes and cluster itself. This will be reported on
663 instance creation for nodes, and on node adds for the nodes. It will
664 be of course avaiblable for querying via the OpQueryNodes/Instance and
665 cluster information, and via RAPI as well.
667 Note that Ganeti will not provide any way to change this attribute.
669 Upgrading from Ganeti 2.0 will automatically add an ‘uuid’ attribute
670 to all entities missing it.
673 Proposed changes part 2
674 +++++++++++++++++++++++
676 In the next release (e.g. 2.2), the tracking of objects will change
677 from the name to the UUID internally, and externally Ganeti will
678 accept both forms of identification; e.g. an RAPI call would be made
679 either against ``/2/instances/foo.bar`` or against
680 ``/2/instances/bb3b2e42…``. Since an FQDN must have at least a dot,
681 and dots are not valid characters in UUIDs, we will not have namespace
684 Another change here is that node identification (during cluster
685 operations/queries like master startup, “am I the master?” and
686 similar) could be done via UUIDs which is more stable than the current
687 hostname-based scheme.
689 Internal tracking refers to the way the configuration is stored; a
690 DRBD disk of an instance refers to the node name (so that IPs can be
691 changed easily), but this is still a problem for name changes; thus
692 these will be changed to point to the node UUID to ease renames.
694 The advantages of this change (after the second round of changes), is
695 that node rename becomes trivial, whereas today node rename would
696 require a complete lock of all instances.
699 Automated disk repairs infrastructure
700 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
702 Replacing defective disks in an automated fashion is quite difficult
703 with the current version of Ganeti. These changes will introduce
704 additional functionality and interfaces to simplify automating disk
705 replacements on a Ganeti node.
707 Fix node volume group
708 +++++++++++++++++++++
710 This is the most difficult addition, as it can lead to dataloss if it's
711 not properly safeguarded.
713 The operation must be done only when all the other nodes that have
714 instances in common with the target node are fine, i.e. this is the only
715 node with problems, and also we have to double-check that all instances
716 on this node have at least a good copy of the data.
718 This might mean that we have to enhance the GetMirrorStatus calls, and
719 introduce and a smarter version that can tell us more about the status
722 Stop allocation on a given PV
723 +++++++++++++++++++++++++++++
725 This is somewhat simple. First we need a "list PVs" opcode (and its
726 associated logical unit) and then a set PV status opcode/LU. These in
727 combination should allow both checking and changing the disk/PV status.
732 This new opcode or opcode change must list the instance-disk-index and
733 node combinations of the instance together with their status. This will
734 allow determining what part of the instance is broken (if any).
739 This new opcode/LU/RAPI call will run ``replace-disks -p`` as needed, in
740 order to fix the instance status. It only affects primary instances;
741 secondaries can just be moved away.
746 This new opcode/LU/RAPI call will take over the current ``gnt-node
747 migrate`` code and run migrate for all instances on the node.
752 This new opcode/LU/RAPI call will take over the current ``gnt-node
753 evacuate`` code and run replace-secondary with an iallocator script for
754 all instances on the node.
760 In order to allow running different processes under unique user-ids
761 on a node, we introduce the user-id pool concept.
763 The user-id pool is a cluster-wide configuration parameter.
764 It is a list of user-ids and/or user-id ranges that are reserved
765 for running Ganeti processes (including KVM instances).
766 The code guarantees that on a given node a given user-id is only
767 handed out if there is no other process running with that user-id.
769 Please note, that this can only be guaranteed if all processes in
770 the system - that run under a user-id belonging to the pool - are
771 started by reserving a user-id first. That can be accomplished
772 either by using the RequestUnusedUid() function to get an unused
773 user-id or by implementing the same locking mechanism.
778 The functions that are specific to the user-id pool feature are located
779 in a separate module: ``lib/uidpool.py``.
784 The user-id pool is a single cluster parameter. It is stored in the
785 *Cluster* object under the ``uid_pool`` name as a list of integer
786 tuples. These tuples represent the boundaries of user-id ranges.
787 For single user-ids, the boundaries are equal.
789 The internal user-id pool representation is converted into a
790 string: a newline separated list of user-ids or user-id ranges.
791 This string representation is distributed to all the nodes via the
792 *ssconf* mechanism. This means that the user-id pool can be
793 accessed in a read-only way on any node without consulting the master
794 node or master candidate nodes.
799 The value of the user-id pool cluster parameter can be initialized
800 at cluster initialization time using the
802 ``gnt-cluster init --uid-pool <uid-pool definition> ...``
806 As there is no sensible default value for the user-id pool parameter,
807 it is initialized to an empty list if no ``--uid-pool`` option is
808 supplied at cluster init time.
810 If the user-id pool is empty, the user-id pool feature is considered
816 The user-id pool cluster parameter can be modified from the
817 command-line with the following commands:
819 - ``gnt-cluster modify --uid-pool <uid-pool definition>``
820 - ``gnt-cluster modify --add-uids <uid-pool definition>``
821 - ``gnt-cluster modify --remove-uids <uid-pool definition>``
823 The ``--uid-pool`` option overwrites the current setting with the
824 supplied ``<uid-pool definition>``, while
825 ``--add-uids``/``--remove-uids`` adds/removes the listed uids
826 or uid-ranges from the pool.
828 The ``<uid-pool definition>`` should be a comma-separated list of
829 user-ids or user-id ranges. A range should be defined by a lower and
830 a higher boundary. The boundaries should be separated with a dash.
831 The boundaries are inclusive.
833 The ``<uid-pool definition>`` is parsed into the internal
834 representation, sanity-checked and stored in the ``uid_pool``
835 attribute of the *Cluster* object.
837 It is also immediately converted into a string (formatted in the
838 input format) and distributed to all nodes via the *ssconf* mechanism.
843 The current value of the user-id pool cluster parameter is printed
844 by the ``gnt-cluster info`` command.
846 The output format is accepted by the ``gnt-cluster modify --uid-pool``
852 The ``uidpool.py`` module provides a function (``RequestUnusedUid``)
853 for requesting an unused user-id from the pool.
855 This will try to find a random user-id that is not currently in use.
856 The algorithm is the following:
858 1) Randomize the list of user-ids in the user-id pool
859 2) Iterate over this randomized UID list
860 3) Create a lock file (it doesn't matter if it already exists)
861 4) Acquire an exclusive POSIX lock on the file, to provide mutual
862 exclusion for the following non-atomic operations
863 5) Check if there is a process in the system with the given UID
864 6) If there isn't, return the UID, otherwise unlock the file and
865 continue the iteration over the user-ids
867 The user can than start a new process with this user-id.
868 Once a process is successfully started, the exclusive POSIX lock can
869 be released, but the lock file will remain in the filesystem.
870 The presence of such a lock file means that the given user-id is most
871 probably in use. The lack of a uid lock file does not guarantee that
872 there are no processes with that user-id.
874 After acquiring the exclusive POSIX lock, ``RequestUnusedUid``
875 always performs a check to see if there is a process running with the
878 A user-id can be returned to the pool, by calling the
879 ``ReleaseUid`` function. This will remove the corresponding lock file.
880 Note, that it doesn't check if there is any process still running
881 with that user-id. The removal of the lock file only means that there
882 are most probably no processes with the given user-id. This helps
883 in speeding up the process of finding a user-id that is guaranteed to
886 There is a convenience function, called ``ExecWithUnusedUid`` that
887 wraps the execution of a function (or any callable) that requires a
888 unique user-id. ``ExecWithUnusedUid`` takes care of requesting an
889 unused user-id and unlocking the lock file. It also automatically
890 returns the user-id to the pool if the callable raises an exception.
895 Requesting a user-id from the pool:
899 from ganeti import ssconf
900 from ganeti import uidpool
902 # Get list of all user-ids in the uid-pool from ssconf
903 ss = ssconf.SimpleStore()
904 uid_pool = uidpool.ParseUidPool(ss.GetUidPool(), separator="\n")
905 all_uids = set(uidpool.ExpandUidPool(uid_pool))
907 uid = uidpool.RequestUnusedUid(all_uids)
909 <start a process with the UID>
910 # Once the process is started, we can release the file lock
913 # Return the UID to the pool
914 uidpool.ReleaseUid(uid)
921 from ganeti import uidpool
923 uid = <get the UID the process is running under>
925 uidpool.ReleaseUid(uid)
928 External interface changes
929 --------------------------
934 The OS API of Ganeti 2.0 has been built with extensibility in mind.
935 Since we pass everything as environment variables it's a lot easier to
936 send new information to the OSes without breaking retrocompatibility.
937 This section of the design outlines the proposed extensions to the API
938 and their implementation.
940 API Version Compatibility Handling
941 ++++++++++++++++++++++++++++++++++
943 In 2.1 there will be a new OS API version (eg. 15), which should be
944 mostly compatible with api 10, except for some new added variables.
945 Since it's easy not to pass some variables we'll be able to handle
946 Ganeti 2.0 OSes by just filtering out the newly added piece of
947 information. We will still encourage OSes to declare support for the new
948 API after checking that the new variables don't provide any conflict for
949 them, and we will drop api 10 support after ganeti 2.1 has released.
951 New Environment variables
952 +++++++++++++++++++++++++
954 Some variables have never been added to the OS api but would definitely
955 be useful for the OSes. We plan to add an INSTANCE_HYPERVISOR variable
956 to allow the OS to make changes relevant to the virtualization the
957 instance is going to use. Since this field is immutable for each
958 instance, the os can tight the install without caring of making sure the
959 instance can run under any virtualization technology.
961 We also want the OS to know the particular hypervisor parameters, to be
962 able to customize the install even more. Since the parameters can
963 change, though, we will pass them only as an "FYI": if an OS ties some
964 instance functionality to the value of a particular hypervisor parameter
965 manual changes or a reinstall may be needed to adapt the instance to the
966 new environment. This is not a regression as of today, because even if
967 the OSes are left blind about this information, sometimes they still
968 need to make compromises and cannot satisfy all possible parameter
974 Currently we are assisting to some degree of "os proliferation" just to
975 change a simple installation behavior. This means that the same OS gets
976 installed on the cluster multiple times, with different names, to
977 customize just one installation behavior. Usually such OSes try to share
978 as much as possible through symlinks, but this still causes
979 complications on the user side, especially when multiple parameters must
982 For example today if you want to install debian etch, lenny or squeeze
983 you probably need to install the debootstrap OS multiple times, changing
984 its configuration file, and calling it debootstrap-etch,
985 debootstrap-lenny or debootstrap-squeeze. Furthermore if you have for
986 example a "server" and a "development" environment which installs
987 different packages/configuration files and must be available for all
988 installs you'll probably end up with deboostrap-etch-server,
989 debootstrap-etch-dev, debootrap-lenny-server, debootstrap-lenny-dev,
990 etc. Crossing more than two parameters quickly becomes not manageable.
992 In order to avoid this we plan to make OSes more customizable, by
993 allowing each OS to declare a list of variants which can be used to
994 customize it. The variants list is mandatory and must be written, one
995 variant per line, in the new "variants.list" file inside the main os
996 dir. At least one supported variant must be supported. When choosing the
997 OS exactly one variant will have to be specified, and will be encoded in
998 the os name as <OS-name>+<variant>. As for today it will be possible to
999 change an instance's OS at creation or install time.
1001 The 2.1 OS list will be the combination of each OS, plus its supported
1002 variants. This will cause the name name proliferation to remain, but at
1003 least the internal OS code will be simplified to just parsing the passed
1004 variant, without the need for symlinks or code duplication.
1006 Also we expect the OSes to declare only "interesting" variants, but to
1007 accept some non-declared ones which a user will be able to pass in by
1008 overriding the checks ganeti does. This will be useful for allowing some
1009 variations to be used without polluting the OS list (per-OS
1010 documentation should list all supported variants). If a variant which is
1011 not internally supported is forced through, the OS scripts should abort.
1013 In the future (post 2.1) we may want to move to full fledged parameters
1014 all orthogonal to each other (for example "architecture" (i386, amd64),
1015 "suite" (lenny, squeeze, ...), etc). (As opposed to the variant, which
1016 is a single parameter, and you need a different variant for all the set
1017 of combinations you want to support). In this case we envision the
1018 variants to be moved inside of Ganeti and be associated with lists
1019 parameter->values associations, which will then be passed to the OS.
1025 Current State and shortcomings
1026 ++++++++++++++++++++++++++++++
1028 The iallocator interface allows creation of instances without manually
1029 specifying nodes, but instead by specifying plugins which will do the
1030 required computations and produce a valid node list.
1032 However, the interface is quite akward to use:
1034 - one cannot set a 'default' iallocator script
1035 - one cannot use it to easily test if allocation would succeed
1036 - some new functionality, such as rebalancing clusters and calculating
1037 capacity estimates is needed
1042 There are two area of improvements proposed:
1044 - improving the use of the current interface
1045 - extending the IAllocator API to cover more automation
1048 Default iallocator names
1049 ^^^^^^^^^^^^^^^^^^^^^^^^
1051 The cluster will hold, for each type of iallocator, a (possibly empty)
1052 list of modules that will be used automatically.
1054 If the list is empty, the behaviour will remain the same.
1056 If the list has one entry, then ganeti will behave as if
1057 '--iallocator' was specifyed on the command line. I.e. use this
1058 allocator by default. If the user however passed nodes, those will be
1061 If the list has multiple entries, they will be tried in order until
1062 one gives a successful answer.
1067 The create instance LU will get a new 'dry-run' option that will just
1068 simulate the placement, and return the chosen node-lists after running
1069 all the usual checks.
1074 Instance add/removals/moves can create a situation where load on the
1075 nodes is not spread equally. For this, a new iallocator mode will be
1076 implemented called ``balance`` in which the plugin, given the current
1077 cluster state, and a maximum number of operations, will need to
1078 compute the instance relocations needed in order to achieve a "better"
1079 (for whatever the script believes it's better) cluster.
1081 Cluster capacity calculation
1082 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1084 In this mode, called ``capacity``, given an instance specification and
1085 the current cluster state (similar to the ``allocate`` mode), the
1086 plugin needs to return:
1088 - how many instances can be allocated on the cluster with that
1090 - on which nodes these will be allocated (in order)
1092 .. vim: set textwidth=72 :