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, and such clusters can consist of up to 32 physical nodes
31 (probably more, dependent on network latency).
33 `pvecm` can be used to create a new cluster, join nodes to a cluster,
34 leave the cluster, get status information and do various other cluster
35 related tasks. The **P**rox**m**o**x** **C**luster **F**ile **S**ystem (``pmxcfs'')
36 is used to transparently distribute the cluster configuration to all cluster
39 Grouping nodes into a cluster has the following advantages:
41 * Centralized, web based management
43 * Multi-master clusters: each node can do all management task
45 * `pmxcfs`: database-driven file system for storing configuration files,
46 replicated in real-time on all nodes using `corosync`.
48 * Easy migration of virtual machines and containers between physical
53 * Cluster-wide services like firewall and HA
59 * All nodes must be in the same network as `corosync` uses IP Multicast
60 to communicate between nodes (also see
61 http://www.corosync.org[Corosync Cluster Engine]). Corosync uses UDP
62 ports 5404 and 5405 for cluster communication.
64 NOTE: Some switches do not support IP multicast by default and must be
65 manually enabled first.
67 * Date and time have to be synchronized.
69 * SSH tunnel on TCP port 22 between nodes is used.
71 * If you are interested in High Availability, you need to have at
72 least three nodes for reliable quorum. All nodes should have the
75 * We recommend a dedicated NIC for the cluster traffic, especially if
76 you use shared storage.
78 * Root password of a cluster node is required for adding nodes.
80 NOTE: It is not possible to mix Proxmox VE 3.x and earlier with
81 Proxmox VE 4.0 cluster nodes.
87 First, install {PVE} on all nodes. Make sure that each node is
88 installed with the final hostname and IP configuration. Changing the
89 hostname and IP is not possible after cluster creation.
91 Currently the cluster creation can either be done on the console (login via
92 `ssh`) or the API, which we have a GUI implementation for (__Datacenter ->
95 While it's often common use to reference all other nodenames in `/etc/hosts`
96 with their IP this is not strictly necessary for a cluster, which normally uses
97 multicast, to work. It maybe useful as you then can connect from one node to
98 the other with SSH through the easier to remember node name.
100 [[pvecm_create_cluster]]
104 Login via `ssh` to the first {pve} node. Use a unique name for your cluster.
105 This name cannot be changed later. The cluster name follows the same rules as
109 hp1# pvecm create CLUSTERNAME
112 CAUTION: The cluster name is used to compute the default multicast address.
113 Please use unique cluster names if you run more than one cluster inside your
114 network. To avoid human confusion, it is also recommended to choose different
115 names even if clusters do not share the cluster network.
117 To check the state of your cluster use:
123 Multiple Clusters In Same Network
124 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
126 It is possible to create multiple clusters in the same physical or logical
127 network. Each cluster must have a unique name, which is used to generate the
128 cluster's multicast group address. As long as no duplicate cluster names are
129 configured in one network segment, the different clusters won't interfere with
132 If multiple clusters operate in a single network it may be beneficial to setup
133 an IGMP querier and enable IGMP Snooping in said network. This may reduce the
134 load of the network significantly because multicast packets are only delivered
135 to endpoints of the respective member nodes.
138 [[pvecm_join_node_to_cluster]]
139 Adding Nodes to the Cluster
140 ---------------------------
142 Login via `ssh` to the node you want to add.
145 hp2# pvecm add IP-ADDRESS-CLUSTER
148 For `IP-ADDRESS-CLUSTER` use the IP from an existing cluster node.
150 CAUTION: A new node cannot hold any VMs, because you would get
151 conflicts about identical VM IDs. Also, all existing configuration in
152 `/etc/pve` is overwritten when you join a new node to the cluster. To
153 workaround, use `vzdump` to backup and restore to a different VMID after
154 adding the node to the cluster.
156 To check the state of cluster:
162 .Cluster status after adding 4 nodes
167 Date: Mon Apr 20 12:30:13 2015
168 Quorum provider: corosync_votequorum
174 Votequorum information
175 ~~~~~~~~~~~~~~~~~~~~~~
182 Membership information
183 ~~~~~~~~~~~~~~~~~~~~~~
185 0x00000001 1 192.168.15.91
186 0x00000002 1 192.168.15.92 (local)
187 0x00000003 1 192.168.15.93
188 0x00000004 1 192.168.15.94
191 If you only want the list of all nodes use:
197 .List nodes in a cluster
201 Membership information
202 ~~~~~~~~~~~~~~~~~~~~~~
210 [[adding-nodes-with-separated-cluster-network]]
211 Adding Nodes With Separated Cluster Network
212 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
214 When adding a node to a cluster with a separated cluster network you need to
215 use the 'ringX_addr' parameters to set the nodes address on those networks:
219 pvecm add IP-ADDRESS-CLUSTER -ring0_addr IP-ADDRESS-RING0
222 If you want to use the Redundant Ring Protocol you will also want to pass the
223 'ring1_addr' parameter.
226 Remove a Cluster Node
227 ---------------------
229 CAUTION: Read carefully the procedure before proceeding, as it could
230 not be what you want or need.
232 Move all virtual machines from the node. Make sure you have no local
233 data or backups you want to keep, or save them accordingly.
234 In the following example we will remove the node hp4 from the cluster.
236 Log in to a *different* cluster node (not hp4), and issue a `pvecm nodes`
237 command to identify the node ID to remove:
242 Membership information
243 ~~~~~~~~~~~~~~~~~~~~~~
252 At this point you must power off hp4 and
253 make sure that it will not power on again (in the network) as it
256 IMPORTANT: As said above, it is critical to power off the node
257 *before* removal, and make sure that it will *never* power on again
258 (in the existing cluster network) as it is.
259 If you power on the node as it is, your cluster will be screwed up and
260 it could be difficult to restore a clean cluster state.
262 After powering off the node hp4, we can safely remove it from the cluster.
265 hp1# pvecm delnode hp4
268 If the operation succeeds no output is returned, just check the node
269 list again with `pvecm nodes` or `pvecm status`. You should see
277 Date: Mon Apr 20 12:44:28 2015
278 Quorum provider: corosync_votequorum
284 Votequorum information
285 ~~~~~~~~~~~~~~~~~~~~~~
292 Membership information
293 ~~~~~~~~~~~~~~~~~~~~~~
295 0x00000001 1 192.168.15.90 (local)
296 0x00000002 1 192.168.15.91
297 0x00000003 1 192.168.15.92
300 If, for whatever reason, you want that this server joins the same
301 cluster again, you have to
303 * reinstall {pve} on it from scratch
305 * then join it, as explained in the previous section.
307 [[pvecm_separate_node_without_reinstall]]
308 Separate A Node Without Reinstalling
309 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
311 CAUTION: This is *not* the recommended method, proceed with caution. Use the
312 above mentioned method if you're unsure.
314 You can also separate a node from a cluster without reinstalling it from
315 scratch. But after removing the node from the cluster it will still have
316 access to the shared storages! This must be resolved before you start removing
317 the node from the cluster. A {pve} cluster cannot share the exact same
318 storage with another cluster, as storage locking doesn't work over cluster
319 boundary. Further, it may also lead to VMID conflicts.
321 Its suggested that you create a new storage where only the node which you want
322 to separate has access. This can be an new export on your NFS or a new Ceph
323 pool, to name a few examples. Its just important that the exact same storage
324 does not gets accessed by multiple clusters. After setting this storage up move
325 all data from the node and its VMs to it. Then you are ready to separate the
326 node from the cluster.
328 WARNING: Ensure all shared resources are cleanly separated! You will run into
329 conflicts and problems else.
331 First stop the corosync and the pve-cluster services on the node:
334 systemctl stop pve-cluster
335 systemctl stop corosync
338 Start the cluster filesystem again in local mode:
344 Delete the corosync configuration files:
347 rm /etc/pve/corosync.conf
351 You can now start the filesystem again as normal service:
355 systemctl start pve-cluster
358 The node is now separated from the cluster. You can deleted it from a remaining
359 node of the cluster with:
362 pvecm delnode oldnode
365 If the command failed, because the remaining node in the cluster lost quorum
366 when the now separate node exited, you may set the expected votes to 1 as a workaround:
372 And the repeat the 'pvecm delnode' command.
374 Now switch back to the separated node, here delete all remaining files left
375 from the old cluster. This ensures that the node can be added to another
376 cluster again without problems.
380 rm /var/lib/corosync/*
383 As the configuration files from the other nodes are still in the cluster
384 filesystem you may want to clean those up too. Remove simply the whole
385 directory recursive from '/etc/pve/nodes/NODENAME', but check three times that
386 you used the correct one before deleting it.
388 CAUTION: The nodes SSH keys are still in the 'authorized_key' file, this means
389 the nodes can still connect to each other with public key authentication. This
390 should be fixed by removing the respective keys from the
391 '/etc/pve/priv/authorized_keys' file.
396 {pve} use a quorum-based technique to provide a consistent state among
399 [quote, from Wikipedia, Quorum (distributed computing)]
401 A quorum is the minimum number of votes that a distributed transaction
402 has to obtain in order to be allowed to perform an operation in a
406 In case of network partitioning, state changes requires that a
407 majority of nodes are online. The cluster switches to read-only mode
410 NOTE: {pve} assigns a single vote to each node by default.
415 The cluster network is the core of a cluster. All messages sent over it have to
416 be delivered reliable to all nodes in their respective order. In {pve} this
417 part is done by corosync, an implementation of a high performance low overhead
418 high availability development toolkit. It serves our decentralized
419 configuration file system (`pmxcfs`).
421 [[cluster-network-requirements]]
424 This needs a reliable network with latencies under 2 milliseconds (LAN
425 performance) to work properly. While corosync can also use unicast for
426 communication between nodes its **highly recommended** to have a multicast
427 capable network. The network should not be used heavily by other members,
428 ideally corosync runs on its own network.
429 *never* share it with network where storage communicates too.
431 Before setting up a cluster it is good practice to check if the network is fit
434 * Ensure that all nodes are in the same subnet. This must only be true for the
435 network interfaces used for cluster communication (corosync).
437 * Ensure all nodes can reach each other over those interfaces, using `ping` is
438 enough for a basic test.
440 * Ensure that multicast works in general and a high package rates. This can be
441 done with the `omping` tool. The final "%loss" number should be < 1%.
445 omping -c 10000 -i 0.001 -F -q NODE1-IP NODE2-IP ...
448 * Ensure that multicast communication works over an extended period of time.
449 This uncovers problems where IGMP snooping is activated on the network but
450 no multicast querier is active. This test has a duration of around 10
455 omping -c 600 -i 1 -q NODE1-IP NODE2-IP ...
458 Your network is not ready for clustering if any of these test fails. Recheck
459 your network configuration. Especially switches are notorious for having
460 multicast disabled by default or IGMP snooping enabled with no IGMP querier
463 In smaller cluster its also an option to use unicast if you really cannot get
466 Separate Cluster Network
467 ~~~~~~~~~~~~~~~~~~~~~~~~
469 When creating a cluster without any parameters the cluster network is generally
470 shared with the Web UI and the VMs and its traffic. Depending on your setup
471 even storage traffic may get sent over the same network. Its recommended to
472 change that, as corosync is a time critical real time application.
474 Setting Up A New Network
475 ^^^^^^^^^^^^^^^^^^^^^^^^
477 First you have to setup a new network interface. It should be on a physical
478 separate network. Ensure that your network fulfills the
479 <<cluster-network-requirements,cluster network requirements>>.
481 Separate On Cluster Creation
482 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
484 This is possible through the 'ring0_addr' and 'bindnet0_addr' parameter of
485 the 'pvecm create' command used for creating a new cluster.
487 If you have setup an additional NIC with a static address on 10.10.10.1/25
488 and want to send and receive all cluster communication over this interface
493 pvecm create test --ring0_addr 10.10.10.1 --bindnet0_addr 10.10.10.0
496 To check if everything is working properly execute:
499 systemctl status corosync
502 Afterwards, proceed as descripted in the section to
503 <<adding-nodes-with-separated-cluster-network,add nodes with a separated cluster network>>.
505 [[separate-cluster-net-after-creation]]
506 Separate After Cluster Creation
507 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
509 You can do this also if you have already created a cluster and want to switch
510 its communication to another network, without rebuilding the whole cluster.
511 This change may lead to short durations of quorum loss in the cluster, as nodes
512 have to restart corosync and come up one after the other on the new network.
514 Check how to <<edit-corosync-conf,edit the corosync.conf file>> first.
515 The open it and you should see a file similar to:
549 provider: corosync_votequorum
553 cluster_name: thomas-testcluster
559 bindnetaddr: 192.168.30.50
566 The first you want to do is add the 'name' properties in the node entries if
567 you do not see them already. Those *must* match the node name.
569 Then replace the address from the 'ring0_addr' properties with the new
570 addresses. You may use plain IP addresses or also hostnames here. If you use
571 hostnames ensure that they are resolvable from all nodes.
573 In my example I want to switch my cluster communication to the 10.10.10.1/25
574 network. So I replace all 'ring0_addr' respectively. I also set the bindnetaddr
575 in the totem section of the config to an address of the new network. It can be
576 any address from the subnet configured on the new network interface.
578 After you increased the 'config_version' property the new configuration file
594 ring0_addr: 10.10.10.2
601 ring0_addr: 10.10.10.3
608 ring0_addr: 10.10.10.1
614 provider: corosync_votequorum
618 cluster_name: thomas-testcluster
624 bindnetaddr: 10.10.10.1
631 Now after a final check whether all changed information is correct we save it
632 and see again the <<edit-corosync-conf,edit corosync.conf file>> section to
633 learn how to bring it in effect.
635 As our change cannot be enforced live from corosync we have to do an restart.
637 On a single node execute:
640 systemctl restart corosync
643 Now check if everything is fine:
647 systemctl status corosync
650 If corosync runs again correct restart corosync also on all other nodes.
651 They will then join the cluster membership one by one on the new network.
654 Redundant Ring Protocol
655 ~~~~~~~~~~~~~~~~~~~~~~~
656 To avoid a single point of failure you should implement counter measurements.
657 This can be on the hardware and operating system level through network bonding.
659 Corosync itself offers also a possibility to add redundancy through the so
660 called 'Redundant Ring Protocol'. This protocol allows running a second totem
661 ring on another network, this network should be physically separated from the
662 other rings network to actually increase availability.
664 RRP On Cluster Creation
665 ~~~~~~~~~~~~~~~~~~~~~~~
667 The 'pvecm create' command provides the additional parameters 'bindnetX_addr',
668 'ringX_addr' and 'rrp_mode', can be used for RRP configuration.
670 NOTE: See the <<corosync-conf-glossary,glossary>> if you do not know what each parameter means.
672 So if you have two networks, one on the 10.10.10.1/24 and the other on the
673 10.10.20.1/24 subnet you would execute:
677 pvecm create CLUSTERNAME -bindnet0_addr 10.10.10.1 -ring0_addr 10.10.10.1 \
678 -bindnet1_addr 10.10.20.1 -ring1_addr 10.10.20.1
681 RRP On Existing Clusters
682 ~~~~~~~~~~~~~~~~~~~~~~~~
684 You will take similar steps as described in
685 <<separate-cluster-net-after-creation,separating the cluster network>> to
686 enable RRP on an already running cluster. The single difference is, that you
687 will add `ring1` and use it instead of `ring0`.
689 First add a new `interface` subsection in the `totem` section, set its
690 `ringnumber` property to `1`. Set the interfaces `bindnetaddr` property to an
691 address of the subnet you have configured for your new ring.
692 Further set the `rrp_mode` to `passive`, this is the only stable mode.
694 Then add to each node entry in the `nodelist` section its new `ring1_addr`
695 property with the nodes additional ring address.
697 So if you have two networks, one on the 10.10.10.1/24 and the other on the
698 10.10.20.1/24 subnet, the final configuration file should look like:
709 bindnetaddr: 10.10.10.1
713 bindnetaddr: 10.10.20.1
723 ring0_addr: 10.10.10.1
724 ring1_addr: 10.10.20.1
731 ring0_addr: 10.10.10.2
732 ring1_addr: 10.10.20.2
735 [...] # other cluster nodes here
738 [...] # other remaining config sections here
742 Bring it in effect like described in the
743 <<edit-corosync-conf,edit the corosync.conf file>> section.
745 This is a change which cannot take live in effect and needs at least a restart
746 of corosync. Recommended is a restart of the whole cluster.
748 If you cannot reboot the whole cluster ensure no High Availability services are
749 configured and the stop the corosync service on all nodes. After corosync is
750 stopped on all nodes start it one after the other again.
752 Corosync Configuration
753 ----------------------
755 The `/etc/pve/corosync.conf` file plays a central role in {pve} cluster. It
756 controls the cluster member ship and its network.
757 For reading more about it check the corosync.conf man page:
763 For node membership you should always use the `pvecm` tool provided by {pve}.
764 You may have to edit the configuration file manually for other changes.
765 Here are a few best practice tips for doing this.
767 [[edit-corosync-conf]]
771 Editing the corosync.conf file can be not always straight forward. There are
772 two on each cluster, one in `/etc/pve/corosync.conf` and the other in
773 `/etc/corosync/corosync.conf`. Editing the one in our cluster file system will
774 propagate the changes to the local one, but not vice versa.
776 The configuration will get updated automatically as soon as the file changes.
777 This means changes which can be integrated in a running corosync will take
778 instantly effect. So you should always make a copy and edit that instead, to
779 avoid triggering some unwanted changes by an in between safe.
783 cp /etc/pve/corosync.conf /etc/pve/corosync.conf.new
786 Then open the Config file with your favorite editor, `nano` and `vim.tiny` are
787 preinstalled on {pve} for example.
789 NOTE: Always increment the 'config_version' number on configuration changes,
790 omitting this can lead to problems.
792 After making the necessary changes create another copy of the current working
793 configuration file. This serves as a backup if the new configuration fails to
794 apply or makes problems in other ways.
798 cp /etc/pve/corosync.conf /etc/pve/corosync.conf.bak
801 Then move the new configuration file over the old one:
804 mv /etc/pve/corosync.conf.new /etc/pve/corosync.conf
807 You may check with the commands
810 systemctl status corosync
811 journalctl -b -u corosync
814 If the change could applied automatically. If not you may have to restart the
815 corosync service via:
818 systemctl restart corosync
821 On errors check the troubleshooting section below.
826 Issue: 'quorum.expected_votes must be configured'
827 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
829 When corosync starts to fail and you get the following message in the system log:
833 corosync[1647]: [QUORUM] Quorum provider: corosync_votequorum failed to initialize.
834 corosync[1647]: [SERV ] Service engine 'corosync_quorum' failed to load for reason
835 'configuration error: nodelist or quorum.expected_votes must be configured!'
839 It means that the hostname you set for corosync 'ringX_addr' in the
840 configuration could not be resolved.
843 Write Configuration When Not Quorate
844 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
846 If you need to change '/etc/pve/corosync.conf' on an node with no quorum, and you
847 know what you do, use:
853 This sets the expected vote count to 1 and makes the cluster quorate. You can
854 now fix your configuration, or revert it back to the last working backup.
856 This is not enough if corosync cannot start anymore. Here its best to edit the
857 local copy of the corosync configuration in '/etc/corosync/corosync.conf' so
858 that corosync can start again. Ensure that on all nodes this configuration has
859 the same content to avoid split brains. If you are not sure what went wrong
860 it's best to ask the Proxmox Community to help you.
863 [[corosync-conf-glossary]]
864 Corosync Configuration Glossary
865 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
868 This names the different ring addresses for the corosync totem rings used for
869 the cluster communication.
872 Defines to which interface the ring should bind to. It may be any address of
873 the subnet configured on the interface we want to use. In general its the
874 recommended to just use an address a node uses on this interface.
877 Specifies the mode of the redundant ring protocol and may be passive, active or
878 none. Note that use of active is highly experimental and not official
879 supported. Passive is the preferred mode, it may double the cluster
880 communication throughput and increases availability.
886 It is obvious that a cluster is not quorate when all nodes are
887 offline. This is a common case after a power failure.
889 NOTE: It is always a good idea to use an uninterruptible power supply
890 (``UPS'', also called ``battery backup'') to avoid this state, especially if
893 On node startup, the `pve-guests` service is started and waits for
894 quorum. Once quorate, it starts all guests which have the `onboot`
897 When you turn on nodes, or when power comes back after power failure,
898 it is likely that some nodes boots faster than others. Please keep in
899 mind that guest startup is delayed until you reach quorum.
905 Migrating virtual guests to other nodes is a useful feature in a
906 cluster. There are settings to control the behavior of such
907 migrations. This can be done via the configuration file
908 `datacenter.cfg` or for a specific migration via API or command line
911 It makes a difference if a Guest is online or offline, or if it has
912 local resources (like a local disk).
914 For Details about Virtual Machine Migration see the
915 xref:qm_migration[QEMU/KVM Migration Chapter]
917 For Details about Container Migration see the
918 xref:pct_migration[Container Migration Chapter]
923 The migration type defines if the migration data should be sent over an
924 encrypted (`secure`) channel or an unencrypted (`insecure`) one.
925 Setting the migration type to insecure means that the RAM content of a
926 virtual guest gets also transferred unencrypted, which can lead to
927 information disclosure of critical data from inside the guest (for
928 example passwords or encryption keys).
930 Therefore, we strongly recommend using the secure channel if you do
931 not have full control over the network and can not guarantee that no
932 one is eavesdropping to it.
934 NOTE: Storage migration does not follow this setting. Currently, it
935 always sends the storage content over a secure channel.
937 Encryption requires a lot of computing power, so this setting is often
938 changed to "unsafe" to achieve better performance. The impact on
939 modern systems is lower because they implement AES encryption in
940 hardware. The performance impact is particularly evident in fast
941 networks where you can transfer 10 Gbps or more.
947 By default, {pve} uses the network in which cluster communication
948 takes place to send the migration traffic. This is not optimal because
949 sensitive cluster traffic can be disrupted and this network may not
950 have the best bandwidth available on the node.
952 Setting the migration network parameter allows the use of a dedicated
953 network for the entire migration traffic. In addition to the memory,
954 this also affects the storage traffic for offline migrations.
956 The migration network is set as a network in the CIDR notation. This
957 has the advantage that you do not have to set individual IP addresses
958 for each node. {pve} can determine the real address on the
959 destination node from the network specified in the CIDR form. To
960 enable this, the network must be specified so that each node has one,
961 but only one IP in the respective network.
967 We assume that we have a three-node setup with three separate
968 networks. One for public communication with the Internet, one for
969 cluster communication and a very fast one, which we want to use as a
970 dedicated network for migration.
972 A network configuration for such a setup might look as follows:
975 iface eno1 inet manual
979 iface vmbr0 inet static
981 netmask 255.255.250.0
989 iface eno2 inet static
991 netmask 255.255.255.0
995 iface eno3 inet static
997 netmask 255.255.255.0
1000 Here, we will use the network 10.1.2.0/24 as a migration network. For
1001 a single migration, you can do this using the `migration_network`
1002 parameter of the command line tool:
1005 # qm migrate 106 tre --online --migration_network 10.1.2.0/24
1008 To configure this as the default network for all migrations in the
1009 cluster, set the `migration` property of the `/etc/pve/datacenter.cfg`
1013 # use dedicated migration network
1014 migration: secure,network=10.1.2.0/24
1017 NOTE: The migration type must always be set when the migration network
1018 gets set in `/etc/pve/datacenter.cfg`.
1022 include::pve-copyright.adoc[]