10 pvecm - Proxmox VE Cluster Manager
15 include::pvecm.1-synopsis.adoc[]
27 The {pve} cluster manager `pvecm` is a tool to create a group of
28 physical servers. Such a group is called a *cluster*. We use the
29 http://www.corosync.org[Corosync Cluster Engine] for reliable group
30 communication. There's no explicit limit for the number of nodes in a cluster.
31 In practice, the actual possible node count may be limited by the host and
32 network performance. Currently (2021), there are reports of clusters (using
33 high-end enterprise hardware) with over 50 nodes in production.
35 `pvecm` can be used to create a new cluster, join nodes to a cluster,
36 leave the cluster, get status information, and do various other cluster-related
37 tasks. The **P**rox**m**o**x** **C**luster **F**ile **S**ystem (``pmxcfs'')
38 is used to transparently distribute the cluster configuration to all cluster
41 Grouping nodes into a cluster has the following advantages:
43 * Centralized, web-based management
45 * Multi-master clusters: each node can do all management tasks
47 * Use of `pmxcfs`, a database-driven file system, for storing configuration
48 files, replicated in real-time on all nodes using `corosync`
50 * Easy migration of virtual machines and containers between physical
55 * Cluster-wide services like firewall and HA
61 * All nodes must be able to connect to each other via UDP ports 5404 and 5405
64 * Date and time must be synchronized.
66 * An SSH tunnel on TCP port 22 between nodes is required.
68 * If you are interested in High Availability, you need to have at
69 least three nodes for reliable quorum. All nodes should have the
72 * We recommend a dedicated NIC for the cluster traffic, especially if
73 you use shared storage.
75 * The root password of a cluster node is required for adding nodes.
77 NOTE: It is not possible to mix {pve} 3.x and earlier with {pve} 4.X cluster
80 NOTE: While it's possible to mix {pve} 4.4 and {pve} 5.0 nodes, doing so is
81 not supported as a production configuration and should only be done temporarily,
82 during an upgrade of the whole cluster from one major version to another.
84 NOTE: Running a cluster of {pve} 6.x with earlier versions is not possible. The
85 cluster protocol (corosync) between {pve} 6.x and earlier versions changed
86 fundamentally. The corosync 3 packages for {pve} 5.4 are only intended for the
87 upgrade procedure to {pve} 6.0.
93 First, install {pve} on all nodes. Make sure that each node is
94 installed with the final hostname and IP configuration. Changing the
95 hostname and IP is not possible after cluster creation.
97 While it's common to reference all node names and their IPs in `/etc/hosts` (or
98 make their names resolvable through other means), this is not necessary for a
99 cluster to work. It may be useful however, as you can then connect from one node
100 to another via SSH, using the easier to remember node name (see also
101 xref:pvecm_corosync_addresses[Link Address Types]). Note that we always
102 recommend referencing nodes by their IP addresses in the cluster configuration.
105 [[pvecm_create_cluster]]
109 You can either create a cluster on the console (login via `ssh`), or through
110 the API using the {pve} web interface (__Datacenter -> Cluster__).
112 NOTE: Use a unique name for your cluster. This name cannot be changed later.
113 The cluster name follows the same rules as node names.
115 [[pvecm_cluster_create_via_gui]]
119 [thumbnail="screenshot/gui-cluster-create.png"]
121 Under __Datacenter -> Cluster__, click on *Create Cluster*. Enter the cluster
122 name and select a network connection from the drop-down list to serve as the
123 main cluster network (Link 0). It defaults to the IP resolved via the node's
126 As of {pve} 6.2, up to 8 fallback links can be added to a cluster. To add a
127 redundant link, click the 'Add' button and select a link number and IP address
128 from the respective fields. Prior to {pve} 6.2, to add a second link as
129 fallback, you can select the 'Advanced' checkbox and choose an additional
130 network interface (Link 1, see also xref:pvecm_redundancy[Corosync Redundancy]).
132 NOTE: Ensure that the network selected for cluster communication is not used for
133 any high traffic purposes, like network storage or live-migration.
134 While the cluster network itself produces small amounts of data, it is very
135 sensitive to latency. Check out full
136 xref:pvecm_cluster_network_requirements[cluster network requirements].
138 [[pvecm_cluster_create_via_cli]]
139 Create via the Command Line
140 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
142 Login via `ssh` to the first {pve} node and run the following command:
145 hp1# pvecm create CLUSTERNAME
148 To check the state of the new cluster use:
154 Multiple Clusters in the Same Network
155 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
157 It is possible to create multiple clusters in the same physical or logical
158 network. In this case, each cluster must have a unique name to avoid possible
159 clashes in the cluster communication stack. Furthermore, this helps avoid human
160 confusion by making clusters clearly distinguishable.
162 While the bandwidth requirement of a corosync cluster is relatively low, the
163 latency of packages and the package per second (PPS) rate is the limiting
164 factor. Different clusters in the same network can compete with each other for
165 these resources, so it may still make sense to use separate physical network
166 infrastructure for bigger clusters.
168 [[pvecm_join_node_to_cluster]]
169 Adding Nodes to the Cluster
170 ---------------------------
172 CAUTION: A node that is about to be added to the cluster cannot hold any guests.
173 All existing configuration in `/etc/pve` is overwritten when joining a cluster,
174 since guest IDs could otherwise conflict. As a workaround, you can create a
175 backup of the guest (`vzdump`) and restore it under a different ID, after the
176 node has been added to the cluster.
178 Join Node to Cluster via GUI
179 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
181 [thumbnail="screenshot/gui-cluster-join-information.png"]
183 Log in to the web interface on an existing cluster node. Under __Datacenter ->
184 Cluster__, click the *Join Information* button at the top. Then, click on the
185 button *Copy Information*. Alternatively, copy the string from the 'Information'
188 [thumbnail="screenshot/gui-cluster-join.png"]
190 Next, log in to the web interface on the node you want to add.
191 Under __Datacenter -> Cluster__, click on *Join Cluster*. Fill in the
192 'Information' field with the 'Join Information' text you copied earlier.
193 Most settings required for joining the cluster will be filled out
194 automatically. For security reasons, the cluster password has to be entered
197 NOTE: To enter all required data manually, you can disable the 'Assisted Join'
200 After clicking the *Join* button, the cluster join process will start
201 immediately. After the node has joined the cluster, its current node certificate
202 will be replaced by one signed from the cluster certificate authority (CA).
203 This means that the current session will stop working after a few seconds. You
204 then might need to force-reload the web interface and log in again with the
207 Now your node should be visible under __Datacenter -> Cluster__.
209 Join Node to Cluster via Command Line
210 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
212 Log in to the node you want to join into an existing cluster via `ssh`.
215 # pvecm add IP-ADDRESS-CLUSTER
218 For `IP-ADDRESS-CLUSTER`, use the IP or hostname of an existing cluster node.
219 An IP address is recommended (see xref:pvecm_corosync_addresses[Link Address Types]).
222 To check the state of the cluster use:
228 .Cluster status after adding 4 nodes
240 Date: Tue Sep 14 11:06:47 2021
241 Quorum provider: corosync_votequorum
247 Votequorum information
248 ~~~~~~~~~~~~~~~~~~~~~~
255 Membership information
256 ~~~~~~~~~~~~~~~~~~~~~~
258 0x00000001 1 192.168.15.91
259 0x00000002 1 192.168.15.92 (local)
260 0x00000003 1 192.168.15.93
261 0x00000004 1 192.168.15.94
264 If you only want a list of all nodes, use:
270 .List nodes in a cluster
274 Membership information
275 ~~~~~~~~~~~~~~~~~~~~~~
283 [[pvecm_adding_nodes_with_separated_cluster_network]]
284 Adding Nodes with Separated Cluster Network
285 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
287 When adding a node to a cluster with a separated cluster network, you need to
288 use the 'link0' parameter to set the nodes address on that network:
292 pvecm add IP-ADDRESS-CLUSTER -link0 LOCAL-IP-ADDRESS-LINK0
295 If you want to use the built-in xref:pvecm_redundancy[redundancy] of the
296 Kronosnet transport layer, also use the 'link1' parameter.
298 Using the GUI, you can select the correct interface from the corresponding
299 'Link X' fields in the *Cluster Join* dialog.
301 Remove a Cluster Node
302 ---------------------
304 CAUTION: Read the procedure carefully before proceeding, as it may
305 not be what you want or need.
307 Move all virtual machines from the node. Ensure that you have made copies of any
308 local data or backups that you want to keep. In addition, make sure to remove
309 any scheduled replication jobs to the node to be removed.
311 CAUTION: Failure to remove replication jobs to a node before removing said node
312 will result in the replication job becoming irremovable. Especially note that
313 replication automatically switches direction if a replicated VM is migrated, so
314 by migrating a replicated VM from a node to be deleted, replication jobs will be
315 set up to that node automatically.
317 In the following example, we will remove the node hp4 from the cluster.
319 Log in to a *different* cluster node (not hp4), and issue a `pvecm nodes`
320 command to identify the node ID to remove:
325 Membership information
326 ~~~~~~~~~~~~~~~~~~~~~~
335 At this point, you must power off hp4 and ensure that it will not power on
336 again (in the network) with its current configuration.
338 IMPORTANT: As mentioned above, it is critical to power off the node
339 *before* removal, and make sure that it will *not* power on again
340 (in the existing cluster network) with its current configuration.
341 If you power on the node as it is, the cluster could end up broken,
342 and it could be difficult to restore it to a functioning state.
344 After powering off the node hp4, we can safely remove it from the cluster.
347 hp1# pvecm delnode hp4
351 NOTE: At this point, it is possible that you will receive an error message
352 stating `Could not kill node (error = CS_ERR_NOT_EXIST)`. This does not
353 signify an actual failure in the deletion of the node, but rather a failure in
354 corosync trying to kill an offline node. Thus, it can be safely ignored.
356 Use `pvecm nodes` or `pvecm status` to check the node list again. It should
364 Votequorum information
365 ~~~~~~~~~~~~~~~~~~~~~~
372 Membership information
373 ~~~~~~~~~~~~~~~~~~~~~~
375 0x00000001 1 192.168.15.90 (local)
376 0x00000002 1 192.168.15.91
377 0x00000003 1 192.168.15.92
380 If, for whatever reason, you want this server to join the same cluster again,
383 * do a fresh install of {pve} on it,
385 * then join it, as explained in the previous section.
387 NOTE: After removal of the node, its SSH fingerprint will still reside in the
388 'known_hosts' of the other nodes. If you receive an SSH error after rejoining
389 a node with the same IP or hostname, run `pvecm updatecerts` once on the
390 re-added node to update its fingerprint cluster wide.
392 [[pvecm_separate_node_without_reinstall]]
393 Separate a Node Without Reinstalling
394 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
396 CAUTION: This is *not* the recommended method, proceed with caution. Use the
397 previous method if you're unsure.
399 You can also separate a node from a cluster without reinstalling it from
400 scratch. But after removing the node from the cluster, it will still have
401 access to any shared storage. This must be resolved before you start removing
402 the node from the cluster. A {pve} cluster cannot share the exact same
403 storage with another cluster, as storage locking doesn't work over the cluster
404 boundary. Furthermore, it may also lead to VMID conflicts.
406 It's suggested that you create a new storage, where only the node which you want
407 to separate has access. This can be a new export on your NFS or a new Ceph
408 pool, to name a few examples. It's just important that the exact same storage
409 does not get accessed by multiple clusters. After setting up this storage, move
410 all data and VMs from the node to it. Then you are ready to separate the
411 node from the cluster.
413 WARNING: Ensure that all shared resources are cleanly separated! Otherwise you
414 will run into conflicts and problems.
416 First, stop the corosync and pve-cluster services on the node:
419 systemctl stop pve-cluster
420 systemctl stop corosync
423 Start the cluster file system again in local mode:
429 Delete the corosync configuration files:
432 rm /etc/pve/corosync.conf
433 rm -r /etc/corosync/*
436 You can now start the file system again as a normal service:
440 systemctl start pve-cluster
443 The node is now separated from the cluster. You can deleted it from any
444 remaining node of the cluster with:
447 pvecm delnode oldnode
450 If the command fails due to a loss of quorum in the remaining node, you can set
451 the expected votes to 1 as a workaround:
457 And then repeat the 'pvecm delnode' command.
459 Now switch back to the separated node and delete all the remaining cluster
460 files on it. This ensures that the node can be added to another cluster again
465 rm /var/lib/corosync/*
468 As the configuration files from the other nodes are still in the cluster
469 file system, you may want to clean those up too. After making absolutely sure
470 that you have the correct node name, you can simply remove the entire
471 directory recursively from '/etc/pve/nodes/NODENAME'.
473 CAUTION: The node's SSH keys will remain in the 'authorized_key' file. This
474 means that the nodes can still connect to each other with public key
475 authentication. You should fix this by removing the respective keys from the
476 '/etc/pve/priv/authorized_keys' file.
482 {pve} use a quorum-based technique to provide a consistent state among
485 [quote, from Wikipedia, Quorum (distributed computing)]
487 A quorum is the minimum number of votes that a distributed transaction
488 has to obtain in order to be allowed to perform an operation in a
492 In case of network partitioning, state changes requires that a
493 majority of nodes are online. The cluster switches to read-only mode
496 NOTE: {pve} assigns a single vote to each node by default.
502 The cluster network is the core of a cluster. All messages sent over it have to
503 be delivered reliably to all nodes in their respective order. In {pve} this
504 part is done by corosync, an implementation of a high performance, low overhead,
505 high availability development toolkit. It serves our decentralized configuration
506 file system (`pmxcfs`).
508 [[pvecm_cluster_network_requirements]]
511 This needs a reliable network with latencies under 2 milliseconds (LAN
512 performance) to work properly. The network should not be used heavily by other
513 members; ideally corosync runs on its own network. Do not use a shared network
514 for corosync and storage (except as a potential low-priority fallback in a
515 xref:pvecm_redundancy[redundant] configuration).
517 Before setting up a cluster, it is good practice to check if the network is fit
518 for that purpose. To ensure that the nodes can connect to each other on the
519 cluster network, you can test the connectivity between them with the `ping`
522 If the {pve} firewall is enabled, ACCEPT rules for corosync will automatically
523 be generated - no manual action is required.
525 NOTE: Corosync used Multicast before version 3.0 (introduced in {pve} 6.0).
526 Modern versions rely on https://kronosnet.org/[Kronosnet] for cluster
527 communication, which, for now, only supports regular UDP unicast.
529 CAUTION: You can still enable Multicast or legacy unicast by setting your
530 transport to `udp` or `udpu` in your xref:pvecm_edit_corosync_conf[corosync.conf],
531 but keep in mind that this will disable all cryptography and redundancy support.
532 This is therefore not recommended.
534 Separate Cluster Network
535 ~~~~~~~~~~~~~~~~~~~~~~~~
537 When creating a cluster without any parameters, the corosync cluster network is
538 generally shared with the web interface and the VMs' network. Depending on
539 your setup, even storage traffic may get sent over the same network. It's
540 recommended to change that, as corosync is a time-critical, real-time
543 Setting Up a New Network
544 ^^^^^^^^^^^^^^^^^^^^^^^^
546 First, you have to set up a new network interface. It should be on a physically
547 separate network. Ensure that your network fulfills the
548 xref:pvecm_cluster_network_requirements[cluster network requirements].
550 Separate On Cluster Creation
551 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
553 This is possible via the 'linkX' parameters of the 'pvecm create'
554 command, used for creating a new cluster.
556 If you have set up an additional NIC with a static address on 10.10.10.1/25,
557 and want to send and receive all cluster communication over this interface,
562 pvecm create test --link0 10.10.10.1
565 To check if everything is working properly, execute:
568 systemctl status corosync
571 Afterwards, proceed as described above to
572 xref:pvecm_adding_nodes_with_separated_cluster_network[add nodes with a separated cluster network].
574 [[pvecm_separate_cluster_net_after_creation]]
575 Separate After Cluster Creation
576 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
578 You can do this if you have already created a cluster and want to switch
579 its communication to another network, without rebuilding the whole cluster.
580 This change may lead to short periods of quorum loss in the cluster, as nodes
581 have to restart corosync and come up one after the other on the new network.
583 Check how to xref:pvecm_edit_corosync_conf[edit the corosync.conf file] first.
584 Then, open it and you should see a file similar to:
618 provider: corosync_votequorum
622 cluster_name: testcluster
634 NOTE: `ringX_addr` actually specifies a corosync *link address*. The name "ring"
635 is a remnant of older corosync versions that is kept for backwards
638 The first thing you want to do is add the 'name' properties in the node entries,
639 if you do not see them already. Those *must* match the node name.
641 Then replace all addresses from the 'ring0_addr' properties of all nodes with
642 the new addresses. You may use plain IP addresses or hostnames here. If you use
643 hostnames, ensure that they are resolvable from all nodes (see also
644 xref:pvecm_corosync_addresses[Link Address Types]).
646 In this example, we want to switch cluster communication to the
647 10.10.10.1/25 network, so we change the 'ring0_addr' of each node respectively.
649 NOTE: The exact same procedure can be used to change other 'ringX_addr' values
650 as well. However, we recommend only changing one link address at a time, so
651 that it's easier to recover if something goes wrong.
653 After we increase the 'config_version' property, the new configuration file
668 ring0_addr: 10.10.10.2
675 ring0_addr: 10.10.10.3
682 ring0_addr: 10.10.10.1
688 provider: corosync_votequorum
692 cluster_name: testcluster
704 Then, after a final check to see that all changed information is correct, we
705 save it and once again follow the
706 xref:pvecm_edit_corosync_conf[edit corosync.conf file] section to bring it into
709 The changes will be applied live, so restarting corosync is not strictly
710 necessary. If you changed other settings as well, or notice corosync
711 complaining, you can optionally trigger a restart.
713 On a single node execute:
717 systemctl restart corosync
720 Now check if everything is okay:
724 systemctl status corosync
727 If corosync begins to work again, restart it on all other nodes too.
728 They will then join the cluster membership one by one on the new network.
730 [[pvecm_corosync_addresses]]
734 A corosync link address (for backwards compatibility denoted by 'ringX_addr' in
735 `corosync.conf`) can be specified in two ways:
737 * **IPv4/v6 addresses** can be used directly. They are recommended, since they
738 are static and usually not changed carelessly.
740 * **Hostnames** will be resolved using `getaddrinfo`, which means that by
741 default, IPv6 addresses will be used first, if available (see also
742 `man gai.conf`). Keep this in mind, especially when upgrading an existing
745 CAUTION: Hostnames should be used with care, since the addresses they
746 resolve to can be changed without touching corosync or the node it runs on -
747 which may lead to a situation where an address is changed without thinking
748 about implications for corosync.
750 A separate, static hostname specifically for corosync is recommended, if
751 hostnames are preferred. Also, make sure that every node in the cluster can
752 resolve all hostnames correctly.
754 Since {pve} 5.1, while supported, hostnames will be resolved at the time of
755 entry. Only the resolved IP is saved to the configuration.
757 Nodes that joined the cluster on earlier versions likely still use their
758 unresolved hostname in `corosync.conf`. It might be a good idea to replace
759 them with IPs or a separate hostname, as mentioned above.
766 Corosync supports redundant networking via its integrated Kronosnet layer by
767 default (it is not supported on the legacy udp/udpu transports). It can be
768 enabled by specifying more than one link address, either via the '--linkX'
769 parameters of `pvecm`, in the GUI as **Link 1** (while creating a cluster or
770 adding a new node) or by specifying more than one 'ringX_addr' in
773 NOTE: To provide useful failover, every link should be on its own
774 physical network connection.
776 Links are used according to a priority setting. You can configure this priority
777 by setting 'knet_link_priority' in the corresponding interface section in
778 `corosync.conf`, or, preferably, using the 'priority' parameter when creating
779 your cluster with `pvecm`:
782 # pvecm create CLUSTERNAME --link0 10.10.10.1,priority=15 --link1 10.20.20.1,priority=20
785 This would cause 'link1' to be used first, since it has the higher priority.
787 If no priorities are configured manually (or two links have the same priority),
788 links will be used in order of their number, with the lower number having higher
791 Even if all links are working, only the one with the highest priority will see
792 corosync traffic. Link priorities cannot be mixed, meaning that links with
793 different priorities will not be able to communicate with each other.
795 Since lower priority links will not see traffic unless all higher priorities
796 have failed, it becomes a useful strategy to specify networks used for
797 other tasks (VMs, storage, etc.) as low-priority links. If worst comes to
798 worst, a higher latency or more congested connection might be better than no
801 Adding Redundant Links To An Existing Cluster
802 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
804 To add a new link to a running configuration, first check how to
805 xref:pvecm_edit_corosync_conf[edit the corosync.conf file].
807 Then, add a new 'ringX_addr' to every node in the `nodelist` section. Make
808 sure that your 'X' is the same for every node you add it to, and that it is
809 unique for each node.
811 Lastly, add a new 'interface', as shown below, to your `totem`
812 section, replacing 'X' with the link number chosen above.
814 Assuming you added a link with number 1, the new configuration file could look
829 ring0_addr: 10.10.10.2
830 ring1_addr: 10.20.20.2
837 ring0_addr: 10.10.10.3
838 ring1_addr: 10.20.20.3
845 ring0_addr: 10.10.10.1
846 ring1_addr: 10.20.20.1
852 provider: corosync_votequorum
856 cluster_name: testcluster
870 The new link will be enabled as soon as you follow the last steps to
871 xref:pvecm_edit_corosync_conf[edit the corosync.conf file]. A restart should not
872 be necessary. You can check that corosync loaded the new link using:
875 journalctl -b -u corosync
878 It might be a good idea to test the new link by temporarily disconnecting the
879 old link on one node and making sure that its status remains online while
886 If you see a healthy cluster state, it means that your new link is being used.
889 Role of SSH in {pve} Clusters
890 -----------------------------
892 {pve} utilizes SSH tunnels for various features.
894 * Proxying console/shell sessions (node and guests)
896 When using the shell for node B while being connected to node A, connects to a
897 terminal proxy on node A, which is in turn connected to the login shell on node
898 B via a non-interactive SSH tunnel.
900 * VM and CT memory and local-storage migration in 'secure' mode.
902 During the migration, one or more SSH tunnel(s) are established between the
903 source and target nodes, in order to exchange migration information and
904 transfer memory and disk contents.
906 * Storage replication
908 .Pitfalls due to automatic execution of `.bashrc` and siblings
911 In case you have a custom `.bashrc`, or similar files that get executed on
912 login by the configured shell, `ssh` will automatically run it once the session
913 is established successfully. This can cause some unexpected behavior, as those
914 commands may be executed with root permissions on any of the operations
915 described above. This can cause possible problematic side-effects!
917 In order to avoid such complications, it's recommended to add a check in
918 `/root/.bashrc` to make sure the session is interactive, and only then run
921 You can add this snippet at the beginning of your `.bashrc` file:
924 # Early exit if not running interactively to avoid side-effects!
933 Corosync External Vote Support
934 ------------------------------
936 This section describes a way to deploy an external voter in a {pve} cluster.
937 When configured, the cluster can sustain more node failures without
938 violating safety properties of the cluster communication.
940 For this to work, there are two services involved:
942 * A QDevice daemon which runs on each {pve} node
944 * An external vote daemon which runs on an independent server
946 As a result, you can achieve higher availability, even in smaller setups (for
949 QDevice Technical Overview
950 ~~~~~~~~~~~~~~~~~~~~~~~~~~
952 The Corosync Quorum Device (QDevice) is a daemon which runs on each cluster
953 node. It provides a configured number of votes to the cluster's quorum
954 subsystem, based on an externally running third-party arbitrator's decision.
955 Its primary use is to allow a cluster to sustain more node failures than
956 standard quorum rules allow. This can be done safely as the external device
957 can see all nodes and thus choose only one set of nodes to give its vote.
958 This will only be done if said set of nodes can have quorum (again) after
959 receiving the third-party vote.
961 Currently, only 'QDevice Net' is supported as a third-party arbitrator. This is
962 a daemon which provides a vote to a cluster partition, if it can reach the
963 partition members over the network. It will only give votes to one partition
964 of a cluster at any time.
965 It's designed to support multiple clusters and is almost configuration and
966 state free. New clusters are handled dynamically and no configuration file
967 is needed on the host running a QDevice.
969 The only requirements for the external host are that it needs network access to
970 the cluster and to have a corosync-qnetd package available. We provide a package
971 for Debian based hosts, and other Linux distributions should also have a package
972 available through their respective package manager.
974 NOTE: In contrast to corosync itself, a QDevice connects to the cluster over
975 TCP/IP. The daemon may even run outside of the cluster's LAN and can have longer
981 We support QDevices for clusters with an even number of nodes and recommend
982 it for 2 node clusters, if they should provide higher availability.
983 For clusters with an odd node count, we currently discourage the use of
984 QDevices. The reason for this is the difference in the votes which the QDevice
985 provides for each cluster type. Even numbered clusters get a single additional
986 vote, which only increases availability, because if the QDevice
987 itself fails, you are in the same position as with no QDevice at all.
989 On the other hand, with an odd numbered cluster size, the QDevice provides
990 '(N-1)' votes -- where 'N' corresponds to the cluster node count. This
991 alternative behavior makes sense; if it had only one additional vote, the
992 cluster could get into a split-brain situation. This algorithm allows for all
993 nodes but one (and naturally the QDevice itself) to fail. However, there are two
996 * If the QNet daemon itself fails, no other node may fail or the cluster
997 immediately loses quorum. For example, in a cluster with 15 nodes, 7
998 could fail before the cluster becomes inquorate. But, if a QDevice is
999 configured here and it itself fails, **no single node** of the 15 may fail.
1000 The QDevice acts almost as a single point of failure in this case.
1002 * The fact that all but one node plus QDevice may fail sounds promising at
1003 first, but this may result in a mass recovery of HA services, which could
1004 overload the single remaining node. Furthermore, a Ceph server will stop
1005 providing services if only '((N-1)/2)' nodes or less remain online.
1007 If you understand the drawbacks and implications, you can decide yourself if
1008 you want to use this technology in an odd numbered cluster setup.
1013 We recommend running any daemon which provides votes to corosync-qdevice as an
1014 unprivileged user. {pve} and Debian provide a package which is already
1015 configured to do so.
1016 The traffic between the daemon and the cluster must be encrypted to ensure a
1017 safe and secure integration of the QDevice in {pve}.
1019 First, install the 'corosync-qnetd' package on your external server
1022 external# apt install corosync-qnetd
1025 and the 'corosync-qdevice' package on all cluster nodes
1028 pve# apt install corosync-qdevice
1031 After doing this, ensure that all the nodes in the cluster are online.
1033 You can now set up your QDevice by running the following command on one
1037 pve# pvecm qdevice setup <QDEVICE-IP>
1040 The SSH key from the cluster will be automatically copied to the QDevice.
1042 NOTE: Make sure that the SSH configuration on your external server allows root
1043 login via password, if you are asked for a password during this step.
1045 After you enter the password and all the steps have successfully completed, you
1046 will see "Done". You can verify that the QDevice has been set up with:
1053 Votequorum information
1054 ~~~~~~~~~~~~~~~~~~~~~
1059 Flags: Quorate Qdevice
1061 Membership information
1062 ~~~~~~~~~~~~~~~~~~~~~~
1063 Nodeid Votes Qdevice Name
1064 0x00000001 1 A,V,NMW 192.168.22.180 (local)
1065 0x00000002 1 A,V,NMW 192.168.22.181
1066 0x00000000 1 Qdevice
1071 Frequently Asked Questions
1072 ~~~~~~~~~~~~~~~~~~~~~~~~~~
1077 In case of a tie, where two same-sized cluster partitions cannot see each other
1078 but can see the QDevice, the QDevice chooses one of those partitions randomly
1079 and provides a vote to it.
1081 Possible Negative Implications
1082 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1084 For clusters with an even node count, there are no negative implications when
1085 using a QDevice. If it fails to work, it is the same as not having a QDevice
1088 Adding/Deleting Nodes After QDevice Setup
1089 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1091 If you want to add a new node or remove an existing one from a cluster with a
1092 QDevice setup, you need to remove the QDevice first. After that, you can add or
1093 remove nodes normally. Once you have a cluster with an even node count again,
1094 you can set up the QDevice again as described previously.
1096 Removing the QDevice
1097 ^^^^^^^^^^^^^^^^^^^^
1099 If you used the official `pvecm` tool to add the QDevice, you can remove it
1103 pve# pvecm qdevice remove
1108 //There is still stuff to add here
1111 Corosync Configuration
1112 ----------------------
1114 The `/etc/pve/corosync.conf` file plays a central role in a {pve} cluster. It
1115 controls the cluster membership and its network.
1116 For further information about it, check the corosync.conf man page:
1122 For node membership, you should always use the `pvecm` tool provided by {pve}.
1123 You may have to edit the configuration file manually for other changes.
1124 Here are a few best practice tips for doing this.
1126 [[pvecm_edit_corosync_conf]]
1130 Editing the corosync.conf file is not always very straightforward. There are
1131 two on each cluster node, one in `/etc/pve/corosync.conf` and the other in
1132 `/etc/corosync/corosync.conf`. Editing the one in our cluster file system will
1133 propagate the changes to the local one, but not vice versa.
1135 The configuration will get updated automatically, as soon as the file changes.
1136 This means that changes which can be integrated in a running corosync will take
1137 effect immediately. Thus, you should always make a copy and edit that instead,
1138 to avoid triggering unintended changes when saving the file while editing.
1142 cp /etc/pve/corosync.conf /etc/pve/corosync.conf.new
1145 Then, open the config file with your favorite editor, such as `nano` or
1146 `vim.tiny`, which come pre-installed on every {pve} node.
1148 NOTE: Always increment the 'config_version' number after configuration changes;
1149 omitting this can lead to problems.
1151 After making the necessary changes, create another copy of the current working
1152 configuration file. This serves as a backup if the new configuration fails to
1153 apply or causes other issues.
1157 cp /etc/pve/corosync.conf /etc/pve/corosync.conf.bak
1160 Then replace the old configuration file with the new one:
1163 mv /etc/pve/corosync.conf.new /etc/pve/corosync.conf
1166 You can check if the changes could be applied automatically, using the following
1170 systemctl status corosync
1171 journalctl -b -u corosync
1174 If the changes could not be applied automatically, you may have to restart the
1175 corosync service via:
1178 systemctl restart corosync
1181 On errors, check the troubleshooting section below.
1186 Issue: 'quorum.expected_votes must be configured'
1187 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1189 When corosync starts to fail and you get the following message in the system log:
1193 corosync[1647]: [QUORUM] Quorum provider: corosync_votequorum failed to initialize.
1194 corosync[1647]: [SERV ] Service engine 'corosync_quorum' failed to load for reason
1195 'configuration error: nodelist or quorum.expected_votes must be configured!'
1199 It means that the hostname you set for a corosync 'ringX_addr' in the
1200 configuration could not be resolved.
1202 Write Configuration When Not Quorate
1203 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1205 If you need to change '/etc/pve/corosync.conf' on a node with no quorum, and you
1206 understand what you are doing, use:
1212 This sets the expected vote count to 1 and makes the cluster quorate. You can
1213 then fix your configuration, or revert it back to the last working backup.
1215 This is not enough if corosync cannot start anymore. In that case, it is best to
1216 edit the local copy of the corosync configuration in
1217 '/etc/corosync/corosync.conf', so that corosync can start again. Ensure that on
1218 all nodes, this configuration has the same content to avoid split-brain
1222 [[pvecm_corosync_conf_glossary]]
1223 Corosync Configuration Glossary
1224 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1227 This names the different link addresses for the Kronosnet connections between
1234 It is obvious that a cluster is not quorate when all nodes are
1235 offline. This is a common case after a power failure.
1237 NOTE: It is always a good idea to use an uninterruptible power supply
1238 (``UPS'', also called ``battery backup'') to avoid this state, especially if
1241 On node startup, the `pve-guests` service is started and waits for
1242 quorum. Once quorate, it starts all guests which have the `onboot`
1245 When you turn on nodes, or when power comes back after power failure,
1246 it is likely that some nodes will boot faster than others. Please keep in
1247 mind that guest startup is delayed until you reach quorum.
1253 Migrating virtual guests to other nodes is a useful feature in a
1254 cluster. There are settings to control the behavior of such
1255 migrations. This can be done via the configuration file
1256 `datacenter.cfg` or for a specific migration via API or command line
1259 It makes a difference if a guest is online or offline, or if it has
1260 local resources (like a local disk).
1262 For details about virtual machine migration, see the
1263 xref:qm_migration[QEMU/KVM Migration Chapter].
1265 For details about container migration, see the
1266 xref:pct_migration[Container Migration Chapter].
1271 The migration type defines if the migration data should be sent over an
1272 encrypted (`secure`) channel or an unencrypted (`insecure`) one.
1273 Setting the migration type to insecure means that the RAM content of a
1274 virtual guest is also transferred unencrypted, which can lead to
1275 information disclosure of critical data from inside the guest (for
1276 example, passwords or encryption keys).
1278 Therefore, we strongly recommend using the secure channel if you do
1279 not have full control over the network and can not guarantee that no
1280 one is eavesdropping on it.
1282 NOTE: Storage migration does not follow this setting. Currently, it
1283 always sends the storage content over a secure channel.
1285 Encryption requires a lot of computing power, so this setting is often
1286 changed to "unsafe" to achieve better performance. The impact on
1287 modern systems is lower because they implement AES encryption in
1288 hardware. The performance impact is particularly evident in fast
1289 networks, where you can transfer 10 Gbps or more.
1294 By default, {pve} uses the network in which cluster communication
1295 takes place to send the migration traffic. This is not optimal both because
1296 sensitive cluster traffic can be disrupted and this network may not
1297 have the best bandwidth available on the node.
1299 Setting the migration network parameter allows the use of a dedicated
1300 network for all migration traffic. In addition to the memory,
1301 this also affects the storage traffic for offline migrations.
1303 The migration network is set as a network using CIDR notation. This
1304 has the advantage that you don't have to set individual IP addresses
1305 for each node. {pve} can determine the real address on the
1306 destination node from the network specified in the CIDR form. To
1307 enable this, the network must be specified so that each node has exactly one
1308 IP in the respective network.
1313 We assume that we have a three-node setup, with three separate
1314 networks. One for public communication with the Internet, one for
1315 cluster communication, and a very fast one, which we want to use as a
1316 dedicated network for migration.
1318 A network configuration for such a setup might look as follows:
1321 iface eno1 inet manual
1325 iface vmbr0 inet static
1326 address 192.X.Y.57/24
1334 iface eno2 inet static
1339 iface eno3 inet static
1343 Here, we will use the network 10.1.2.0/24 as a migration network. For
1344 a single migration, you can do this using the `migration_network`
1345 parameter of the command line tool:
1348 # qm migrate 106 tre --online --migration_network 10.1.2.0/24
1351 To configure this as the default network for all migrations in the
1352 cluster, set the `migration` property of the `/etc/pve/datacenter.cfg`
1356 # use dedicated migration network
1357 migration: secure,network=10.1.2.0/24
1360 NOTE: The migration type must always be set when the migration network
1361 is set in `/etc/pve/datacenter.cfg`.
1365 include::pve-copyright.adoc[]