10 ha-manager - Proxmox VE HA Manager
15 include::ha-manager.1-synopsis.adoc[]
26 Our modern society depends heavily on information provided by
27 computers over the network. Mobile devices amplified that dependency,
28 because people can access the network any time from anywhere. If you
29 provide such services, it is very important that they are available
32 We can mathematically define the availability as the ratio of (A) the
33 total time a service is capable of being used during a given interval
34 to (B) the length of the interval. It is normally expressed as a
35 percentage of uptime in a given year.
37 .Availability - Downtime per Year
38 [width="60%",cols="<d,d",options="header"]
39 |===========================================================
40 |Availability % |Downtime per year
45 |99.9999 |31.5 seconds
46 |99.99999 |3.15 seconds
47 |===========================================================
49 There are several ways to increase availability. The most elegant
50 solution is to rewrite your software, so that you can run it on
51 several host at the same time. The software itself need to have a way
52 to detect errors and do failover. This is relatively easy if you just
53 want to serve read-only web pages. But in general this is complex, and
54 sometimes impossible because you cannot modify the software
55 yourself. The following solutions works without modifying the
58 * Use reliable ``server'' components
60 NOTE: Computer components with same functionality can have varying
61 reliability numbers, depending on the component quality. Most vendors
62 sell components with higher reliability as ``server'' components -
63 usually at higher price.
65 * Eliminate single point of failure (redundant components)
66 ** use an uninterruptible power supply (UPS)
67 ** use redundant power supplies on the main boards
69 ** use redundant network hardware
70 ** use RAID for local storage
71 ** use distributed, redundant storage for VM data
74 ** rapidly accessible administrators (24/7)
75 ** availability of spare parts (other nodes in a {pve} cluster)
76 ** automatic error detection (provided by `ha-manager`)
77 ** automatic failover (provided by `ha-manager`)
79 Virtualization environments like {pve} make it much easier to reach
80 high availability because they remove the ``hardware'' dependency. They
81 also support to setup and use redundant storage and network
82 devices. So if one host fail, you can simply start those services on
83 another host within your cluster.
85 Even better, {pve} provides a software stack called `ha-manager`,
86 which can do that automatically for you. It is able to automatically
87 detect errors and do automatic failover.
89 {pve} `ha-manager` works like an ``automated'' administrator. First, you
90 configure what resources (VMs, containers, ...) it should
91 manage. `ha-manager` then observes correct functionality, and handles
92 service failover to another node in case of errors. `ha-manager` can
93 also handle normal user requests which may start, stop, relocate and
96 But high availability comes at a price. High quality components are
97 more expensive, and making them redundant duplicates the costs at
98 least. Additional spare parts increase costs further. So you should
99 carefully calculate the benefits, and compare with those additional
102 TIP: Increasing availability from 99% to 99.9% is relatively
103 simply. But increasing availability from 99.9999% to 99.99999% is very
104 hard and costly. `ha-manager` has typical error detection and failover
105 times of about 2 minutes, so you can get no more than 99.999%
112 You must meet the following requirements before you start with HA:
114 * at least three cluster nodes (to get reliable quorum)
116 * shared storage for VMs and containers
118 * hardware redundancy (everywhere)
120 * use reliable “server” components
122 * hardware watchdog - if not available we fall back to the
123 linux kernel software watchdog (`softdog`)
125 * optional hardware fencing devices
128 [[ha_manager_resources]]
132 We call the primary management unit handled by `ha-manager` a
133 resource. A resource (also called ``service'') is uniquely
134 identified by a service ID (SID), which consists of the resource type
135 and an type specific ID, e.g.: `vm:100`. That example would be a
136 resource of type `vm` (virtual machine) with the ID 100.
138 For now we have two important resources types - virtual machines and
139 containers. One basic idea here is that we can bundle related software
140 into such VM or container, so there is no need to compose one big
141 service from other services, like it was done with `rgmanager`. In
142 general, a HA enabled resource should not depend on other resources.
148 This section provides a detailed description of the {PVE} HA manager
149 internals. It describes all involved daemons and how they work
150 together. To provide HA, two daemons run on each node:
154 The local resource manager (LRM), which controls the services running on
155 the local node. It reads the requested states for its services from
156 the current manager status file and executes the respective commands.
160 The cluster resource manager (CRM), which makes the cluster wide
161 decisions. It sends commands to the LRM, processes the results,
162 and moves resources to other nodes if something fails. The CRM also
163 handles node fencing.
166 .Locks in the LRM & CRM
168 Locks are provided by our distributed configuration file system (pmxcfs).
169 They are used to guarantee that each LRM is active once and working. As a
170 LRM only executes actions when it holds its lock we can mark a failed node
171 as fenced if we can acquire its lock. This lets us then recover any failed
172 HA services securely without any interference from the now unknown failed node.
173 This all gets supervised by the CRM which holds currently the manager master
180 The CRM use a service state enumeration to record the current service
181 state. We display this state on the GUI and you can query it using
182 the `ha-manager` command line tool:
187 master elsa (active, Mon Nov 21 07:23:29 2016)
188 lrm elsa (active, Mon Nov 21 07:23:22 2016)
189 service ct:100 (elsa, stopped)
190 service ct:102 (elsa, started)
191 service vm:501 (elsa, started)
194 Here is the list of possible states:
198 Service is stopped (confirmed by LRM). If the LRM detects a stopped
199 service is still running, it will stop it again.
203 Service should be stopped. The CRM waits for confirmation from the
208 Service is active an LRM should start it ASAP if not already running.
209 If the Service fails and is detected to be not running the LRM
211 (see xref:ha_manager_start_failure_policy[Start Failure Policy]).
215 Wait for node fencing (service node is not inside quorate cluster
216 partition). As soon as node gets fenced successfully the service will
217 be recovered to another node, if possible
218 (see xref:ha_manager_fencing[Fencing]).
222 Do not touch the service state. We use this state while we reboot a
223 node, or when we restart the LRM daemon
224 (see xref:ha_manager_package_updates[Package Updates]).
228 Migrate service (live) to other node.
232 Service is disabled because of LRM errors. Needs manual intervention
233 (see xref:ha_manager_error_recovery[Error Recovery]).
236 Local Resource Manager
237 ~~~~~~~~~~~~~~~~~~~~~~
239 The local resource manager (`pve-ha-lrm`) is started as a daemon on
240 boot and waits until the HA cluster is quorate and thus cluster wide
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.
252 The LRM holds its exclusive lock and has services configured.
256 The LRM lost its lock, this means a failure happened and quorum was lost.
258 After the LRM gets in the active state it reads the manager status
259 file in `/etc/pve/ha/manager_status` and determines the commands it
260 has to execute for the services it owns.
261 For each command a worker gets started, this workers are running in
262 parallel and are limited to at most 4 by default. This default setting
263 may be changed through the datacenter configuration key `max_worker`.
264 When finished the worker process gets collected and its result saved for
267 .Maximum Concurrent Worker Adjustment Tips
269 The default value of at most 4 concurrent workers may be unsuited for
270 a specific setup. For example may 4 live migrations happen at the same
271 time, which can lead to network congestions with slower networks and/or
272 big (memory wise) services. Ensure that also in the worst case no congestion
273 happens and lower the `max_worker` value if needed. In the contrary, if you
274 have a particularly powerful high end setup you may also want to increase it.
276 Each command requested by the CRM is uniquely identifiable by an UID, when
277 the worker finished its result will be processed and written in the LRM
278 status file `/etc/pve/nodes/<nodename>/lrm_status`. There the CRM may collect
279 it and let its state machine - respective the commands output - act on it.
281 The actions on each service between CRM and LRM are normally always synced.
282 This means that the CRM requests a state uniquely marked by an UID, the LRM
283 then executes this action *one time* and writes back the result, also
284 identifiable by the same UID. This is needed so that the LRM does not
285 executes an outdated command.
286 With the exception of the `stop` and the `error` command,
287 those two do not depend on the result produced and are executed
288 always in the case of the stopped state and once in the case of
293 The HA Stack logs every action it makes. This helps to understand what
294 and also why something happens in the cluster. Here its important to see
295 what both daemons, the LRM and the CRM, did. You may use
296 `journalctl -u pve-ha-lrm` on the node(s) where the service is and
297 the same command for the pve-ha-crm on the node which is the current master.
299 Cluster Resource Manager
300 ~~~~~~~~~~~~~~~~~~~~~~~~
302 The cluster resource manager (`pve-ha-crm`) starts on each node and
303 waits there for the manager lock, which can only be held by one node
304 at a time. The node which successfully acquires the manager lock gets
305 promoted to the CRM master.
307 It can be in three states:
309 wait for agent lock::
311 The CRM waits for our exclusive lock. This is also used as idle state if no
312 service is configured
316 The CRM holds its exclusive lock and has services configured
320 The CRM lost its lock, this means a failure happened and quorum was lost.
322 It main task is to manage the services which are configured to be highly
323 available and try to always enforce them to the wanted state, e.g.: a
324 enabled service will be started if its not running, if it crashes it will
325 be started again. Thus it dictates the LRM the actions it needs to execute.
327 When an node leaves the cluster quorum, its state changes to unknown.
328 If the current CRM then can secure the failed nodes lock, the services
329 will be 'stolen' and restarted on another node.
331 When a cluster member determines that it is no longer in the cluster
332 quorum, the LRM waits for a new quorum to form. As long as there is no
333 quorum the node cannot reset the watchdog. This will trigger a reboot
334 after the watchdog then times out, this happens after 60 seconds.
340 The HA stack is well integrated into the {pve} API. So, for example,
341 HA can be configured via the `ha-manager` command line interface, or
342 the {pve} web interface - both interfaces provide an easy way to
343 manage HA. Automation tools can use the API directly.
345 All HA configuration files are within `/etc/pve/ha/`, so they get
346 automatically distributed to the cluster nodes, and all nodes share
347 the same HA configuration.
353 The resource configuration file `/etc/pve/ha/resources.cfg` stores
354 the list of resources managed by `ha-manager`. A resource configuration
355 inside that list look like this:
363 It starts with a resource type followed by a resource specific name,
364 separated with colon. Together this forms the HA resource ID, which is
365 used by all `ha-manager` commands to uniquely identify a resource
366 (example: `vm:100` or `ct:101`). The next lines contain additional
369 include::ha-resources-opts.adoc[]
371 Here is a real world example with one VM and one container. As you see,
372 the syntax of those files is really simple, so it is even posiible to
373 read or edit those files using your favorite editor:
375 .Configuration Example (`/etc/pve/ha/resources.cfg`)
382 # Note: use default settings for everything
385 Above config was generated using the `ha-manager` command line tool:
388 # ha-manager add vm:501 --state started --max_relocate 2
389 # ha-manager add ct:102
393 [[ha_manager_groups]]
397 The HA group configuration file `/etc/pve/ha/groups.cfg` is used to
398 define groups of cluster nodes. A resource can be restricted to run
399 only on the members of such group. A group configuration look like
409 include::ha-groups-opts.adoc[]
411 A commom requirement is that a resource should run on a specific
412 node. Usually the resource is able to run on other nodes, so you can define
413 an unrestricted group with a single member:
416 # ha-manager groupadd prefer_node1 --nodes node1
419 For bigger clusters, it makes sense to define a more detailed failover
420 behavior. For example, you may want to run a set of services on
421 `node1` if possible. If `node1` is not available, you want to run them
422 equally splitted on `node2` and `node3`. If those nodes also fail the
423 services should run on `node4`. To achieve this you could set the node
427 # ha-manager groupadd mygroup1 -nodes "node1:2,node2:1,node3:1,node4"
430 Another use case is if a resource uses other resources only available
431 on specific nodes, lets say `node1` and `node2`. We need to make sure
432 that HA manager does not use other nodes, so we need to create a
433 restricted group with said nodes:
436 # ha-manager groupadd mygroup2 -nodes "node1,node2" -restricted
439 Above commands created the following group configuration fils:
441 .Configuration Example (`/etc/pve/ha/groups.cfg`)
447 nodes node2:1,node4,node1:2,node3:1
455 The `nofailback` options is mostly useful to avoid unwanted resource
456 movements during administartion tasks. For example, if you need to
457 migrate a service to a node which hasn't the highest priority in the
458 group, you need to tell the HA manager to not move this service
459 instantly back by setting the `nofailback` option.
461 Another scenario is when a service was fenced and it got recovered to
462 another node. The admin tries to repair the fenced node and brings it
463 up online again to investigate the failure cause and check if it runs
464 stable again. Setting the `nofailback` flag prevents that the
465 recovered services move straight back to the fenced node.
471 If a node needs maintenance you should migrate and or relocate all
472 services which are required to run always on another node first.
473 After that you can stop the LRM and CRM services. But note that the
474 watchdog triggers if you stop it with active services.
477 [[ha_manager_package_updates]]
481 When updating the ha-manager you should do one node after the other, never
482 all at once for various reasons. First, while we test our software
483 thoughtfully, a bug affecting your specific setup cannot totally be ruled out.
484 Upgrading one node after the other and checking the functionality of each node
485 after finishing the update helps to recover from an eventual problems, while
486 updating all could render you in a broken cluster state and is generally not
489 Also, the {pve} HA stack uses a request acknowledge protocol to perform
490 actions between the cluster and the local resource manager. For restarting,
491 the LRM makes a request to the CRM to freeze all its services. This prevents
492 that they get touched by the Cluster during the short time the LRM is restarting.
493 After that the LRM may safely close the watchdog during a restart.
494 Such a restart happens on a update and as already stated a active master
495 CRM is needed to acknowledge the requests from the LRM, if this is not the case
496 the update process can be too long which, in the worst case, may result in
500 [[ha_manager_fencing]]
507 Fencing secures that on a node failure the dangerous node gets will be rendered
508 unable to do any damage and that no resource runs twice when it gets recovered
509 from the failed node. This is a really important task and one of the base
510 principles to make a system Highly Available.
512 If a node would not get fenced it would be in an unknown state where it may
513 have still access to shared resources, this is really dangerous!
514 Imagine that every network but the storage one broke, now while not
515 reachable from the public network the VM still runs and writes on the shared
516 storage. If we would not fence the node and just start up this VM on another
517 Node we would get dangerous race conditions, atomicity violations the whole VM
518 could be rendered unusable. The recovery could also simply fail if the storage
519 protects from multiple mounts and thus defeat the purpose of HA.
524 There are different methods to fence a node, for example fence devices which
525 cut off the power from the node or disable their communication completely.
527 Those are often quite expensive and bring additional critical components in
528 a system, because if they fail you cannot recover any service.
530 We thus wanted to integrate a simpler method in the HA Manager first, namely
531 self fencing with watchdogs.
533 Watchdogs are widely used in critical and dependable systems since the
534 beginning of micro controllers, they are often independent and simple
535 integrated circuit which programs can use to watch them. After opening they need to
536 report periodically. If, for whatever reason, a program becomes unable to do
537 so the watchdogs triggers a reset of the whole server.
539 Server motherboards often already include such hardware watchdogs, these need
540 to be configured. If no watchdog is available or configured we fall back to the
541 Linux Kernel softdog while still reliable it is not independent of the servers
542 Hardware and thus has a lower reliability then a hardware watchdog.
544 Configure Hardware Watchdog
545 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
546 By default all watchdog modules are blocked for security reasons as they are
547 like a loaded gun if not correctly initialized.
548 If you have a hardware watchdog available remove its kernel module from the
549 blacklist, load it with insmod and restart the `watchdog-mux` service or reboot
552 Recover Fenced Services
553 ~~~~~~~~~~~~~~~~~~~~~~~
555 After a node failed and its fencing was successful we start to recover services
556 to other available nodes and restart them there so that they can provide service
559 The selection of the node on which the services gets recovered is influenced
560 by the users group settings, the currently active nodes and their respective
561 active service count.
562 First we build a set out of the intersection between user selected nodes and
563 available nodes. Then the subset with the highest priority of those nodes
564 gets chosen as possible nodes for recovery. We select the node with the
565 currently lowest active service count as a new node for the service.
566 That minimizes the possibility of an overload, which else could cause an
567 unresponsive node and as a result a chain reaction of node failures in the
571 [[ha_manager_start_failure_policy]]
573 ---------------------
575 The start failure policy comes in effect if a service failed to start on a
576 node once ore more times. It can be used to configure how often a restart
577 should be triggered on the same node and how often a service should be
578 relocated so that it gets a try to be started on another node.
579 The aim of this policy is to circumvent temporary unavailability of shared
580 resources on a specific node. For example, if a shared storage isn't available
581 on a quorate node anymore, e.g. network problems, but still on other nodes,
582 the relocate policy allows then that the service gets started nonetheless.
584 There are two service start recover policy settings which can be configured
585 specific for each resource.
589 Maximum number of tries to restart an failed service on the actual
590 node. The default is set to one.
594 Maximum number of tries to relocate the service to a different node.
595 A relocate only happens after the max_restart value is exceeded on the
596 actual node. The default is set to one.
598 NOTE: The relocate count state will only reset to zero when the
599 service had at least one successful start. That means if a service is
600 re-enabled without fixing the error only the restart policy gets
604 [[ha_manager_error_recovery]]
608 If after all tries the service state could not be recovered it gets
609 placed in an error state. In this state the service won't get touched
610 by the HA stack anymore. To recover from this state you should follow
613 * bring the resource back into a safe and consistent state (e.g.,
616 * disable the ha resource to place it in an stopped state
618 * fix the error which led to this failures
620 * *after* you fixed all errors you may enable the service again
623 [[ha_manager_service_operations]]
627 This are how the basic user-initiated service operations (via
632 The service will be started by the LRM if not already running.
636 The service will be stopped by the LRM if running.
640 The service will be relocated (live) to another node.
644 The service will be removed from the HA managed resource list. Its
645 current state will not be touched.
649 `start` and `stop` commands can be issued to the resource specific tools
650 (like `qm` or `pct`), they will forward the request to the
651 `ha-manager` which then will execute the action and set the resulting
652 service state (enabled, disabled).
656 include::pve-copyright.adoc[]