[pve-docs.git] / ha-manager.adoc
1 [[chapter-ha-manager]]
2 ifdef::manvolnum[]
3 PVE({manvolnum})
4 ================
5 include::attributes.txt[]
8 ----
10 ha-manager - Proxmox VE HA Manager
13 --------
15 include::ha-manager.1-synopsis.adoc[]
18 -----------
19 endif::manvolnum[]
21 ifndef::manvolnum[]
22 High Availability
23 =================
24 include::attributes.txt[]
25 endif::manvolnum[]
28 Our modern society depends heavily on information provided by
29 computers over the network. Mobile devices amplified that dependency,
30 because people can access the network any time from anywhere. If you
31 provide such services, it is very important that they are available
32 most of the time.
34 We can mathematically define the availability as the ratio of (A) the
35 total time a service is capable of being used during a given interval
36 to (B) the length of the interval. It is normally expressed as a
37 percentage of uptime in a given year.
39 .Availability - Downtime per Year
40 [width="60%",cols="<d,d",options="header"]
41 |===========================================================
42 |Availability % |Downtime per year
43 |99 |3.65 days
44 |99.9 |8.76 hours
45 |99.99 |52.56 minutes
46 |99.999 |5.26 minutes
47 |99.9999 |31.5 seconds
48 |99.99999 |3.15 seconds
49 |===========================================================
51 There are several ways to increase availability. The most elegant
52 solution is to rewrite your software, so that you can run it on
53 several host at the same time. The software itself need to have a way
54 to detect errors and do failover. This is relatively easy if you just
55 want to serve read-only web pages. But in general this is complex, and
56 sometimes impossible because you cannot modify the software
57 yourself. The following solutions works without modifying the
58 software:
60 * Use reliable ``server'' components
62 NOTE: Computer components with same functionality can have varying
63 reliability numbers, depending on the component quality. Most vendors
64 sell components with higher reliability as ``server'' components -
65 usually at higher price.
67 * Eliminate single point of failure (redundant components)
68 ** use an uninterruptible power supply (UPS)
69 ** use redundant power supplies on the main boards
70 ** use ECC-RAM
71 ** use redundant network hardware
72 ** use RAID for local storage
73 ** use distributed, redundant storage for VM data
75 * Reduce downtime
76 ** rapidly accessible administrators (24/7)
77 ** availability of spare parts (other nodes in a {pve} cluster)
78 ** automatic error detection (provided by `ha-manager`)
79 ** automatic failover (provided by `ha-manager`)
81 Virtualization environments like {pve} make it much easier to reach
82 high availability because they remove the ``hardware'' dependency. They
83 also support to setup and use redundant storage and network
84 devices. So if one host fail, you can simply start those services on
85 another host within your cluster.
87 Even better, {pve} provides a software stack called `ha-manager`,
88 which can do that automatically for you. It is able to automatically
89 detect errors and do automatic failover.
91 {pve} `ha-manager` works like an ``automated'' administrator. First, you
92 configure what resources (VMs, containers, ...) it should
93 manage. `ha-manager` then observes correct functionality, and handles
94 service failover to another node in case of errors. `ha-manager` can
95 also handle normal user requests which may start, stop, relocate and
96 migrate a service.
98 But high availability comes at a price. High quality components are
99 more expensive, and making them redundant duplicates the costs at
100 least. Additional spare parts increase costs further. So you should
101 carefully calculate the benefits, and compare with those additional
102 costs.
104 TIP: Increasing availability from 99% to 99.9% is relatively
105 simply. But increasing availability from 99.9999% to 99.99999% is very
106 hard and costly. `ha-manager` has typical error detection and failover
107 times of about 2 minutes, so you can get no more than 99.999%
108 availability.
110 Requirements
111 ------------
113 * at least three cluster nodes (to get reliable quorum)
115 * shared storage for VMs and containers
117 * hardware redundancy (everywhere)
119 * hardware watchdog - if not available we fall back to the
120 linux kernel software watchdog (`softdog`)
122 * optional hardware fencing devices
125 Resources
126 ---------
128 We call the primary management unit handled by `ha-manager` a
129 resource. A resource (also called ``service'') is uniquely
130 identified by a service ID (SID), which consists of the resource type
131 and an type specific ID, e.g.: `vm:100`. That example would be a
132 resource of type `vm` (virtual machine) with the ID 100.
134 For now we have two important resources types - virtual machines and
135 containers. One basic idea here is that we can bundle related software
136 into such VM or container, so there is no need to compose one big
137 service from other services, like it was done with `rgmanager`. In
138 general, a HA enabled resource should not depend on other resources.
141 How It Works
142 ------------
144 This section provides an in detail description of the {PVE} HA-manager
145 internals. It describes how the CRM and the LRM work together.
147 To provide High Availability two daemons run on each node:
149 `pve-ha-lrm`::
151 The local resource manager (LRM), it controls the services running on
152 the local node.
153 It reads the requested states for its services from the current manager
154 status file and executes the respective commands.
156 `pve-ha-crm`::
158 The cluster resource manager (CRM), it controls the cluster wide
159 actions of the services, processes the LRM results and includes the state
160 machine which controls the state of each service.
162 .Locks in the LRM & CRM
163 [NOTE]
164 Locks are provided by our distributed configuration file system (pmxcfs).
165 They are used to guarantee that each LRM is active once and working. As a
166 LRM only executes actions when it holds its lock we can mark a failed node
167 as fenced if we can acquire its lock. This lets us then recover any failed
168 HA services securely without any interference from the now unknown failed node.
169 This all gets supervised by the CRM which holds currently the manager master
170 lock.
172 Local Resource Manager
173 ~~~~~~~~~~~~~~~~~~~~~~
175 The local resource manager (`pve-ha-lrm`) is started as a daemon on
176 boot and waits until the HA cluster is quorate and thus cluster wide
177 locks are working.
179 It can be in three states:
181 *wait for agent lock*::
183 The LRM waits for our exclusive lock. This is also used as idle state if no
184 service is configured.
186 *active*::
188 The LRM holds its exclusive lock and has services configured.
190 *lost agent lock*::
192 The LRM lost its lock, this means a failure happened and quorum was lost.
194 After the LRM gets in the active state it reads the manager status
195 file in `/etc/pve/ha/manager_status` and determines the commands it
196 has to execute for the services it owns.
197 For each command a worker gets started, this workers are running in
198 parallel and are limited to at most 4 by default. This default setting
199 may be changed through the datacenter configuration key `max_worker`.
200 When finished the worker process gets collected and its result saved for
201 the CRM.
203 .Maximum Concurrent Worker Adjustment Tips
204 [NOTE]
205 The default value of at most 4 concurrent workers may be unsuited for
206 a specific setup. For example may 4 live migrations happen at the same
207 time, which can lead to network congestions with slower networks and/or
208 big (memory wise) services. Ensure that also in the worst case no congestion
209 happens and lower the `max_worker` value if needed. In the contrary, if you
210 have a particularly powerful high end setup you may also want to increase it.
212 Each command requested by the CRM is uniquely identifiable by an UID, when
213 the worker finished its result will be processed and written in the LRM
214 status file `/etc/pve/nodes/<nodename>/lrm_status`. There the CRM may collect
215 it and let its state machine - respective the commands output - act on it.
217 The actions on each service between CRM and LRM are normally always synced.
218 This means that the CRM requests a state uniquely marked by an UID, the LRM
219 then executes this action *one time* and writes back the result, also
220 identifiable by the same UID. This is needed so that the LRM does not
221 executes an outdated command.
222 With the exception of the `stop` and the `error` command,
223 those two do not depend on the result produced and are executed
224 always in the case of the stopped state and once in the case of
225 the error state.
227 .Read the Logs
228 [NOTE]
229 The HA Stack logs every action it makes. This helps to understand what
230 and also why something happens in the cluster. Here its important to see
231 what both daemons, the LRM and the CRM, did. You may use
232 `journalctl -u pve-ha-lrm` on the node(s) where the service is and
233 the same command for the pve-ha-crm on the node which is the current master.
235 Cluster Resource Manager
236 ~~~~~~~~~~~~~~~~~~~~~~~~
238 The cluster resource manager (`pve-ha-crm`) starts on each node and
239 waits there for the manager lock, which can only be held by one node
240 at a time. The node which successfully acquires the manager lock gets
241 promoted to the CRM master.
243 It can be in three states:
245 *wait for agent lock*::
247 The LRM waits for our exclusive lock. This is also used as idle state if no
248 service is configured
250 *active*::
252 The LRM holds its exclusive lock and has services configured
254 *lost agent lock*::
256 The LRM lost its lock, this means a failure happened and quorum was lost.
258 It main task is to manage the services which are configured to be highly
259 available and try to always enforce them to the wanted state, e.g.: a
260 enabled service will be started if its not running, if it crashes it will
261 be started again. Thus it dictates the LRM the actions it needs to execute.
263 When an node leaves the cluster quorum, its state changes to unknown.
264 If the current CRM then can secure the failed nodes lock, the services
265 will be 'stolen' and restarted on another node.
267 When a cluster member determines that it is no longer in the cluster
268 quorum, the LRM waits for a new quorum to form. As long as there is no
269 quorum the node cannot reset the watchdog. This will trigger a reboot
270 after the watchdog then times out, this happens after 60 seconds.
272 Configuration
273 -------------
275 The HA stack is well integrated in the Proxmox VE API2. So, for
276 example, HA can be configured via `ha-manager` or the PVE web
277 interface, which both provide an easy to use tool.
279 The resource configuration file can be located at
280 `/etc/pve/ha/resources.cfg` and the group configuration file at
281 `/etc/pve/ha/groups.cfg`. Use the provided tools to make changes,
282 there shouldn't be any need to edit them manually.
284 Node Power Status
285 -----------------
287 If a node needs maintenance you should migrate and or relocate all
288 services which are required to run always on another node first.
289 After that you can stop the LRM and CRM services. But note that the
290 watchdog triggers if you stop it with active services.
292 Package Updates
293 ---------------
295 When updating the ha-manager you should do one node after the other, never
296 all at once for various reasons. First, while we test our software
297 thoughtfully, a bug affecting your specific setup cannot totally be ruled out.
298 Upgrading one node after the other and checking the functionality of each node
299 after finishing the update helps to recover from an eventual problems, while
300 updating all could render you in a broken cluster state and is generally not
301 good practice.
303 Also, the {pve} HA stack uses a request acknowledge protocol to perform
304 actions between the cluster and the local resource manager. For restarting,
305 the LRM makes a request to the CRM to freeze all its services. This prevents
306 that they get touched by the Cluster during the short time the LRM is restarting.
307 After that the LRM may safely close the watchdog during a restart.
308 Such a restart happens on a update and as already stated a active master
309 CRM is needed to acknowledge the requests from the LRM, if this is not the case
310 the update process can be too long which, in the worst case, may result in
311 a watchdog reset.
314 Fencing
315 -------
317 What is Fencing
318 ~~~~~~~~~~~~~~~
320 Fencing secures that on a node failure the dangerous node gets will be rendered
321 unable to do any damage and that no resource runs twice when it gets recovered
322 from the failed node. This is a really important task and one of the base
323 principles to make a system Highly Available.
325 If a node would not get fenced it would be in an unknown state where it may
326 have still access to shared resources, this is really dangerous!
327 Imagine that every network but the storage one broke, now while not
328 reachable from the public network the VM still runs and writes on the shared
329 storage. If we would not fence the node and just start up this VM on another
330 Node we would get dangerous race conditions, atomicity violations the whole VM
331 could be rendered unusable. The recovery could also simply fail if the storage
332 protects from multiple mounts and thus defeat the purpose of HA.
334 How {pve} Fences
335 ~~~~~~~~~~~~~~~~~
337 There are different methods to fence a node, for example fence devices which
338 cut off the power from the node or disable their communication completely.
340 Those are often quite expensive and bring additional critical components in
341 a system, because if they fail you cannot recover any service.
343 We thus wanted to integrate a simpler method in the HA Manager first, namely
344 self fencing with watchdogs.
346 Watchdogs are widely used in critical and dependable systems since the
347 beginning of micro controllers, they are often independent and simple
348 integrated circuit which programs can use to watch them. After opening they need to
349 report periodically. If, for whatever reason, a program becomes unable to do
350 so the watchdogs triggers a reset of the whole server.
352 Server motherboards often already include such hardware watchdogs, these need
353 to be configured. If no watchdog is available or configured we fall back to the
354 Linux Kernel softdog while still reliable it is not independent of the servers
355 Hardware and thus has a lower reliability then a hardware watchdog.
357 Configure Hardware Watchdog
358 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
359 By default all watchdog modules are blocked for security reasons as they are
360 like a loaded gun if not correctly initialized.
361 If you have a hardware watchdog available remove its kernel module from the
362 blacklist, load it with insmod and restart the `watchdog-mux` service or reboot
363 the node.
365 Recover Fenced Services
366 ~~~~~~~~~~~~~~~~~~~~~~~
368 After a node failed and its fencing was successful we start to recover services
369 to other available nodes and restart them there so that they can provide service
370 again.
372 The selection of the node on which the services gets recovered is influenced
373 by the users group settings, the currently active nodes and their respective
374 active service count.
375 First we build a set out of the intersection between user selected nodes and
376 available nodes. Then the subset with the highest priority of those nodes
377 gets chosen as possible nodes for recovery. We select the node with the
378 currently lowest active service count as a new node for the service.
379 That minimizes the possibility of an overload, which else could cause an
380 unresponsive node and as a result a chain reaction of node failures in the
381 cluster.
383 Groups
384 ------
386 A group is a collection of cluster nodes which a service may be bound to.
388 Group Settings
389 ~~~~~~~~~~~~~~
391 nodes::
393 List of group node members where a priority can be given to each node.
394 A service bound to this group will run on the nodes with the highest priority
395 available. If more nodes are in the highest priority class the services will
396 get distributed to those node if not already there. The priorities have a
397 relative meaning only.
399 restricted::
401 Resources bound to this group may only run on nodes defined by the
402 group. If no group node member is available the resource will be
403 placed in the stopped state.
405 nofailback::
407 The resource won't automatically fail back when a more preferred node
408 (re)joins the cluster.
411 Start Failure Policy
412 ---------------------
414 The start failure policy comes in effect if a service failed to start on a
415 node once ore more times. It can be used to configure how often a restart
416 should be triggered on the same node and how often a service should be
417 relocated so that it gets a try to be started on another node.
418 The aim of this policy is to circumvent temporary unavailability of shared
419 resources on a specific node. For example, if a shared storage isn't available
420 on a quorate node anymore, e.g. network problems, but still on other nodes,
421 the relocate policy allows then that the service gets started nonetheless.
423 There are two service start recover policy settings which can be configured
424 specific for each resource.
426 max_restart::
428 Maximum number of tries to restart an failed service on the actual
429 node. The default is set to one.
431 max_relocate::
433 Maximum number of tries to relocate the service to a different node.
434 A relocate only happens after the max_restart value is exceeded on the
435 actual node. The default is set to one.
437 NOTE: The relocate count state will only reset to zero when the
438 service had at least one successful start. That means if a service is
439 re-enabled without fixing the error only the restart policy gets
440 repeated.
442 Error Recovery
443 --------------
445 If after all tries the service state could not be recovered it gets
446 placed in an error state. In this state the service won't get touched
447 by the HA stack anymore. To recover from this state you should follow
448 these steps:
450 * bring the resource back into a safe and consistent state (e.g.,
451 killing its process)
453 * disable the ha resource to place it in an stopped state
455 * fix the error which led to this failures
457 * *after* you fixed all errors you may enable the service again
460 Service Operations
461 ------------------
463 This are how the basic user-initiated service operations (via
464 `ha-manager`) work.
466 enable::
468 The service will be started by the LRM if not already running.
470 disable::
472 The service will be stopped by the LRM if running.
474 migrate/relocate::
476 The service will be relocated (live) to another node.
478 remove::
480 The service will be removed from the HA managed resource list. Its
481 current state will not be touched.
483 start/stop::
485 `start` and `stop` commands can be issued to the resource specific tools
486 (like `qm` or `pct`), they will forward the request to the
487 `ha-manager` which then will execute the action and set the resulting
488 service state (enabled, disabled).
491 Service States
492 --------------
494 stopped::
496 Service is stopped (confirmed by LRM), if detected running it will get stopped
497 again.
499 request_stop::
501 Service should be stopped. Waiting for confirmation from LRM.
503 started::
505 Service is active an LRM should start it ASAP if not already running.
506 If the Service fails and is detected to be not running the LRM restarts it.
508 fence::
510 Wait for node fencing (service node is not inside quorate cluster
511 partition).
512 As soon as node gets fenced successfully the service will be recovered to
513 another node, if possible.
515 freeze::
517 Do not touch the service state. We use this state while we reboot a
518 node, or when we restart the LRM daemon.
520 migrate::
522 Migrate service (live) to other node.
524 error::
526 Service disabled because of LRM errors. Needs manual intervention.
529 ifdef::manvolnum[]
530 include::pve-copyright.adoc[]
531 endif::manvolnum[]