5 include::attributes.txt[]
10 ha-manager - Proxmox VE HA Manager
15 include::ha-manager.1-synopsis.adoc[]
24 include::attributes.txt[]
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
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
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 erors 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
60 * Use reliable "server" components
62 NOTE: Computer components with same functionality can have varying
63 reliability numbers, depending on the component quality. Most verdors
64 sell components with higher reliability as "server" components -
65 usually at higher price.
67 * Eliminate single point of failure (redundant components)
69 - use an uniteruptable power supply (UPS)
70 - use redundant power supplies on the main boards
72 - use redundant network hardware
73 - use RAID for local storage
74 - use distributed, redundant storage for VM data
78 - rapidly accessible adminstrators (24/7)
79 - availability of spare parts (other nodes is a {pve} cluster)
80 - automatic error detection ('ha-manager')
81 - automatic failover ('ha-manager')
83 Virtualization environments like {pve} makes it much easier to reach
84 high availability because they remove the "hardware" dependency. They
85 also support to setup and use redundant storage and network
86 devices. So if one host fail, you can simply start those services on
87 another host within your cluster. Even better, 'ha-manager' can do
88 that automatically for you. It is able to automatically detect errors
89 and do automatic failover.
91 But high availability comes at a price. High quality components are
92 more expensive, and making them redundant duplicates the costs at
93 least. Additional spare parts increase costs further. So you should
94 carefully calculate the benefits, and compare with those additional
97 TIP: Increasing availability from 99% to 99.9% is relatively
98 simply. But increasing availability from 99.9999% to 99.99999% is very
101 'ha-manager' handles management of user-defined cluster services. This
102 includes handling of user requests which may start, stop, relocate,
104 The cluster resource manager daemon also handles restarting and relocating
105 services to another node in the event of failures.
107 A service (also called resource) is uniquely identified by a service ID
108 (SID) which consists of the service type and an type specific id, e.g.:
109 'vm:100'. That example would be a service of type vm (Virtual machine)
115 * at least three nodes
119 * hardware redundancy
121 * hardware watchdog - if not available we fall back to the
122 linux kernel soft dog
127 This section provides an in detail description of the {PVE} HA-manager
128 internals. It describes how the CRM and the LRM work together.
130 To provide High Availability two daemons run on each node:
134 The local resource manager (LRM), it controls the services running on
136 It reads the requested states for its services from the current manager
137 status file and executes the respective commands.
141 The cluster resource manager (CRM), it controls the cluster wide
142 actions of the services, processes the LRM result includes the state
143 machine which controls the state of each service.
145 .Locks in the LRM & CRM
147 Locks are provided by our distributed configuration file system (pmxcfs).
148 They are used to guarantee that each LRM is active and working as a
149 LRM only executes actions when he has its lock we can mark a failed node
150 as fenced if we get its lock. This lets us then recover the failed HA services
151 securely without the failed (but maybe still running) LRM interfering.
152 This all gets supervised by the CRM which holds currently the manager master
155 Local Resource Manager
156 ~~~~~~~~~~~~~~~~~~~~~~
158 The local resource manager ('pve-ha-lrm') is started as a daemon on
159 boot and waits until the HA cluster is quorate and thus cluster wide
162 It can be in three states:
164 * *wait for agent lock*: the LRM waits for our exclusive lock. This is
165 also used as idle sate if no service is configured
166 * *active*: the LRM holds its exclusive lock and has services configured
167 * *lost agent lock*: the LRM lost its lock, this means a failure happened
170 After the LRM gets in the active state it reads the manager status
171 file in '/etc/pve/ha/manager_status' and determines the commands it
172 has to execute for the service it owns.
173 For each command a worker gets started, this workers are running in
174 parallel and are limited to maximal 4 by default. This default setting
175 may be changed through the datacenter configuration key "max_worker".
177 .Maximal Concurrent Worker Adjustment Tips
179 The default value of 4 maximal concurrent Workers may be unsuited for
180 a specific setup. For example may 4 live migrations happen at the same
181 time, which can lead to network congestions with slower networks and/or
182 big (memory wise) services. Ensure that also in the worst case no congestion
183 happens and lower the "max_worker" value if needed. In the contrary, if you
184 have a particularly powerful high end setup you may also want to increase it.
186 Each command requested by the CRM is uniquely identifiable by an UID, when
187 the worker finished its result will be processed and written in the LRM
188 status file '/etc/pve/nodes/<nodename>/lrm_status'. There the CRM may collect
189 it and let its state machine - respective the commands output - act on it.
191 The actions on each service between CRM and LRM are normally always synced.
192 This means that the CRM requests a state uniquely marked by an UID, the LRM
193 then executes this action *one time* and writes back the result, also
194 identifiable by the same UID. This is needed so that the LRM does not
195 executes an outdated command.
196 With the exception of the 'stop' and the 'error' command,
197 those two do not depend on the result produce and are executed
198 always in the case of the stopped state and once in the case of
203 The HA Stack logs every action it makes. This helps to understand what
204 and also why something happens in the cluster. Here its important to see
205 what both daemons, the LRM and the CRM, did. You may use
206 `journalctl -u pve-ha-lrm` on the node(s) where the service is and
207 the same command for the pve-ha-crm on the node which is the current master.
209 Cluster Resource Manager
210 ~~~~~~~~~~~~~~~~~~~~~~~~
212 The cluster resource manager ('pve-ha-crm') starts on each node and
213 waits there for the manager lock, which can only be held by one node
214 at a time. The node which successfully acquires the manager lock gets
215 promoted to the CRM master.
217 It can be in three states: TODO
219 * *wait for agent lock*: the LRM waits for our exclusive lock. This is
220 also used as idle sate if no service is configured
221 * *active*: the LRM holds its exclusive lock and has services configured
222 * *lost agent lock*: the LRM lost its lock, this means a failure happened
225 It main task is to manage the services which are configured to be highly
226 available and try to get always bring them in the wanted state, e.g.: a
227 enabled service will be started if its not running, if it crashes it will
228 be started again. Thus it dictates the LRM the wanted actions.
230 When an node leaves the cluster quorum, its state changes to unknown.
231 If the current CRM then can secure the failed nodes lock, the services
232 will be 'stolen' and restarted on another node.
234 When a cluster member determines that it is no longer in the cluster
235 quorum, the LRM waits for a new quorum to form. As long as there is no
236 quorum the node cannot reset the watchdog. This will trigger a reboot
242 The HA stack is well integrated int the Proxmox VE API2. So, for
243 example, HA can be configured via 'ha-manager' or the PVE web
244 interface, which both provide an easy to use tool.
246 The resource configuration file can be located at
247 '/etc/pve/ha/resources.cfg' and the group configuration file at
248 '/etc/pve/ha/groups.cfg'. Use the provided tools to make changes,
249 there shouldn't be any need to edit them manually.
254 If a node needs maintenance you should migrate and or relocate all
255 services which are required to run always on another node first.
256 After that you can stop the LRM and CRM services. But note that the
257 watchdog triggers if you stop it with active services.
265 Fencing secures that on a node failure the dangerous node gets will be rendered
266 unable to do any damage and that no resource runs twice when it gets recovered
267 from the failed node.
269 Configure Hardware Watchdog
270 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
271 By default all watchdog modules are blocked for security reasons as they are
272 like a loaded gun if not correctly initialized.
273 If you have a hardware watchdog available remove its module from the blacklist
274 and restart 'the watchdog-mux' service.
277 Resource/Service Agents
278 -------------------------
280 A resource or also called service can be managed by the
281 ha-manager. Currently we support virtual machines and container.
286 A group is a collection of cluster nodes which a service may be bound to.
293 list of group node members
297 resources bound to this group may only run on nodes defined by the
298 group. If no group node member is available the resource will be
299 placed in the stopped state.
303 the resource won't automatically fail back when a more preferred node
304 (re)joins the cluster.
310 There are two service recover policy settings which can be configured
311 specific for each resource.
315 maximal number of tries to restart an failed service on the actual
316 node. The default is set to one.
320 maximal number of tries to relocate the service to a different node.
321 A relocate only happens after the max_restart value is exceeded on the
322 actual node. The default is set to one.
324 Note that the relocate count state will only reset to zero when the
325 service had at least one successful start. That means if a service is
326 re-enabled without fixing the error only the restart policy gets
332 If after all tries the service state could not be recovered it gets
333 placed in an error state. In this state the service won't get touched
334 by the HA stack anymore. To recover from this state you should follow
337 * bring the resource back into an safe and consistent state (e.g:
340 * disable the ha resource to place it in an stopped state
342 * fix the error which led to this failures
344 * *after* you fixed all errors you may enable the service again
350 This are how the basic user-initiated service operations (via
355 the service will be started by the LRM if not already running.
359 the service will be stopped by the LRM if running.
363 the service will be relocated (live) to another node.
367 the service will be removed from the HA managed resource list. Its
368 current state will not be touched.
372 start and stop commands can be issued to the resource specific tools
373 (like 'qm' or 'pct'), they will forward the request to the
374 'ha-manager' which then will execute the action and set the resulting
375 service state (enabled, disabled).
383 Service is stopped (confirmed by LRM)
387 Service should be stopped. Waiting for confirmation from LRM.
391 Service is active an LRM should start it ASAP if not already running.
395 Wait for node fencing (service node is not inside quorate cluster
400 Do not touch the service state. We use this state while we reboot a
401 node, or when we restart the LRM daemon.
405 Migrate service (live) to other node.
409 Service disabled because of LRM errors. Needs manual intervention.
413 include::pve-copyright.adoc[]