COROSYNC_CONF

NAME

corosync.conf - corosync executive configuration file

SYNOPSIS

/etc/corosync/corosync.conf

DESCRIPTION

The corosync.conf instructs the corosync executive about various parameters needed to control the corosync executive. Empty lines and lines starting with # character are ignored. The configuration file consists of bracketed top level directives. The possible directive choices are:
totem { }

This top level directive contains configuration options for the totem protocol.

logging { }

This top level directive contains configuration options for logging.

quorum { }

This top level directive contains configuration options for quorum.

nodelist { }

This top level directive contains configuration options for nodes in cluster.

system { }

This top level directive contains configuration options related to system.

resources { }

This top level directive contains configuration options for resources.

nozzle { }

This top level directive contains configuration options for a libnozzle device.

The interface sub-directive of totem is optional for UDP and knet transports.

For knet, multiple interface subsections define parameters for each knet link on the system.

For UDPU an interface section is not needed and it is recommended that the nodelist is used to define cluster nodes.
linknumber

This specifies the link number for the interface. When using the knet protocol, each interface should specify separate link numbers to uniquely identify to the membership protocol which interface to use for which link. The linknumber must start at 0. For UDP the only supported linknumber is 0.

knet_link_priority

This specifies the priority for the link when knet is used in ’passive’ mode. (see link_mode below)

knet_ping_interval

This specifies the interval between knet link pings. knet_ping_interval and knet_ping_timeout are a pair, if one is specified the other should be too, otherwise one will be calculated from the token timeout and one will be taken from the config file. (default is token timeout / (knet_pong_count*2))

knet_ping_timeout

If no ping is received within this time, the knet link is declared dead. knet_ping_interval and knet_ping_timeout are a pair, if one is specified the other should be too, otherwise one will be calculated from the token timeout and one will be taken from the config file. (default is token timeout / knet_pong_count)

knet_ping_precision

How many values of latency are used to calculate the average link latency. (default 2048 samples)

knet_pong_count

How many valid ping/pongs before a link is marked UP. (default 2)

knet_transport

Which IP transport knet should use. valid values are "sctp" or "udp". (default: udp)

bindnetaddr (udp only)

This specifies the network address the corosync executive should bind to when using udp.

bindnetaddr (udp only) should be an IP address configured on the system, or a network address.

For example, if the local interface is 192.168.5.92 with netmask 255.255.255.0, you should set bindnetaddr to 192.168.5.92 or 192.168.5.0. If the local interface is 192.168.5.92 with netmask 255.255.255.192, set bindnetaddr to 192.168.5.92 or 192.168.5.64, and so forth.

This may also be an IPV6 address, in which case IPV6 networking will be used. In this case, the exact address must be specified and there is no automatic selection of the network interface within a specific subnet as with IPv4.

If IPv6 networking is used, the nodeid field in nodelist must be specified.

broadcast (udp only)

This is optional and can be set to yes. If it is set to yes, the broadcast address will be used for communication. If this option is set, mcastaddr should not be set.

mcastaddr (udp only)

This is the multicast address used by corosync executive. The default should work for most networks, but the network administrator should be queried about a multicast address to use. Avoid 224.x.x.x because this is a "config" multicast address.

This may also be an IPV6 multicast address, in which case IPV6 networking will be used. If IPv6 networking is used, the nodeid field in nodelist must be specified.

It’s not necessary to use this option if cluster_name option is used. If both options are used, mcastaddr has higher priority.

mcastport (udp only)

This specifies the UDP port number. It is possible to use the same multicast address on a network with the corosync services configured for different UDP ports. Please note corosync uses two UDP ports mcastport (for mcast receives) and mcastport - 1 (for mcast sends). If you have multiple clusters on the same network using the same mcastaddr please configure the mcastports with a gap.

ttl (udp only)

This specifies the Time To Live (TTL). If you run your cluster on a routed network then the default of "1" will be too small. This option provides a way to increase this up to 255. The valid range is 0..255.

Within the totem directive, there are seven configuration options of which one is required, five are optional, and one is required when IPV6 is configured in the interface subdirective. The required directive controls the version of the totem configuration. The optional option unless using IPV6 directive controls identification of the processor. The optional options control secrecy and authentication, the network mode of operation and maximum network MTU field.
version

This specifies the version of the configuration file. Currently the only valid version for this directive is 2.

clear_node_high_bit

This configuration option is optional and is only relevant when no nodeid is specified. Some corosync clients require a signed 32 bit nodeid that is greater than zero however by default corosync uses all 32 bits of the IPv4 address space when generating a nodeid. Set this option to yes to force the high bit to be zero and therefore ensure the nodeid is a positive signed 32 bit integer.

WARNING: Cluster behavior is undefined if this option is enabled on only a subset of the cluster (for example during a rolling upgrade).

crypto_model

This specifies which cryptographic library should be used by knet. Supported values depend on the libknet build and on the installed cryptography libraries. Typically nss and openssl will be available but gcrypt and others could also be allowed.

The default is nss.

crypto_hash

This specifies which HMAC authentication should be used to authenticate all messages. Valid values are none (no authentication), md5, sha1, sha256, sha384 and sha512. Encrypted transmission is only supported for the knet transport.

The default is none.

crypto_cipher

This specifies which cipher should be used to encrypt all messages. Valid values are none (no encryption), aes256, aes192 and aes128. Enabling crypto_cipher, requires also enabling of crypto_hash. Encrypted transmission is only supported for the knet transport.

The default is none.

secauth

This implies crypto_cipher=aes256 and crypto_hash=sha256, unless those options are explicitly set. Encrypted transmission is only supported for the knet transport.

The default is off.

keyfile

This specifies the fully qualified path to the shared key used to authenticate and encrypt data used within the Totem protocol.

The default is /etc/corosync/authkey.

link_mode

This specifies the Kronosnet mode, which may be passive, active, or rr (round-robin). passive: the active link with the highest priority (highest number) will be used. If one or more links share the same priority the one with the lowest link ID will be used. active: All active links will be used simultaneously to send traffic. link priority is ignored. rr: Round-Robin policy. Each packet will be sent to the next active link in order.

If only one interface directive is specified, passive is automatically chosen.

The maximum number of interface directives that is allowed with Kronosnet is 8. For other transports it is 1.

For UDPU (and UDP), setting this value beyond 1500, the regular frame MTU, requires ethernet devices that support large, or also called jumbo, frames. If any device in the network doesn’t support large frames, the protocol will not operate properly. The hosts must also have their mtu size set from 1500 to whatever frame size is specified here.

Please note while some NICs or switches claim large frame support, they support 9000 MTU as the maximum frame size including the IP header. Setting the netmtu and host MTUs to 9000 will cause totem to use the full 9000 bytes of the frame. Then Linux will add a 18 byte header moving the full frame size to 9018. As a result some hardware will not operate properly with this size of data. A netmtu of 8982 seems to work for the few large frame devices that have been tested. Some manufacturers claim large frame support when in fact they support frame sizes of 4500 bytes.

When sending multicast traffic, if the network frequently reconfigures, chances are that some device in the network doesn’t support large frames.

Choose hardware carefully if intending to use large frame support.

The default is 1500 for UDPU (and UDP) and 65536 for the KNET transport.

transport

This directive controls the transport mechanism used. The default is knet. The transport type can also be set to udpu or udp. Only knet allows crypto or multiple interfaces per node.

cluster_name

This specifies the name of cluster and it’s used for automatic generating of multicast address.

config_version

This specifies version of config file. This is converted to unsigned 64-bit int. By default it’s 0. Option is used to prevent joining old nodes with not up-to-date configuration. If value is not 0, and node is going for first time (only for first time, join after split doesn’t follow this rules) from single-node membership to multiple nodes membership, other nodes config_versions are collected. If current node config_version is not equal to highest of collected versions, corosync is terminated.

ip_version

This specifies version of IP to ask DNS resolver for. The value can be one of ipv4 (look only for an IPv4 address) , ipv6 (check only IPv6 address) , ipv4-6 (look for all address families and use first IPv4 address found in the list if there is such address, otherwise use first IPv6 address) and ipv6-4 (look for all address families and use first IPv6 address found in the list if there is such address, otherwise use first IPv4 address).

Default (if unspecified) is ipv6-4 for knet and udpu transports and ipv4 for udp.

The knet transport supports IPv4 and IPv6 addresses concurrently, provided they are consistent on each link.

Within the totem directive, there are several configuration options which are used to control the operation of the protocol. It is generally not recommended to change any of these values without proper guidance and sufficient testing. Some networks may require larger values if suffering from frequent reconfigurations. Some applications may require faster failure detection times which can be achieved by reducing the token timeout.

For real token timeout used by totem it’s possible to read cmap value of runtime.config.totem.token key.

Be careful to use the same timeout values on each of the nodes in the cluster or unpredictable results may occur.

The default is 3000 milliseconds.

token_warning

Specifies the interval between warnings that the token has not been received. The value is a percentage of the token timeout and can be set to 0 to disable warnings.

The default is 75%.

token_coefficient

This value is used only when nodelist section is specified and contains at least 3 nodes. If so, real token timeout is then computed as token + (number_of_nodes - 2) * token_coefficient. This allows cluster to scale without manually changing token timeout every time new node is added. This value can be set to 0 resulting in effective removal of this feature.

The default is 650 milliseconds.

token_retransmit

This timeout specifies in milliseconds after how long before receiving a token the token is retransmitted. This will be automatically calculated if token is modified. It is not recommended to alter this value without guidance from the corosync community.

The minimum is 30 milliseconds. If not set and error occur, make sure token / (token_retransmits_before_loss_const + 0.2) is more than 30.

The default is 238 milliseconds for two nodes cluster. Three or more nodes reference token_coefficient.

knet_compression_model

Type of compression used by Kronosnet. Supported values depend on the libknet build and on the installed compression libraries. Typically zlib and lz4 will be available but bzip2 and others could also be allowed. The default is ’none’.

knet_compression_threshold

Tells knet to NOT compress any packets that are smaller than the value indicated. Default 100 bytes.

Set to 0 to reset to the default. Set to 1 to compress everything.

knet_compression_level

Many compression libraries allow tuning of compression parameters. For example 0 or 1 ... 9 are commonly used to determine the level of compression. This value is passed unmodified to the compression library so it is recommended to consult the library’s documentation for more detailed information.

The default is 180 milliseconds.

token_retransmits_before_loss_const

This value identifies how many token retransmits should be attempted before forming a new configuration. It is also used for token_retransmit and hold calculations.

The default is 4 retransmissions.

The default is 50 milliseconds.

send_join

This timeout specifies in milliseconds an upper range between 0 and send_join to wait before sending a join message. For configurations with less than 32 nodes, this parameter is not necessary. For larger rings, this parameter is necessary to ensure the NIC is not overflowed with join messages on formation of a new ring. A reasonable value for large rings (128 nodes) would be 80msec. Other timer values must also change if this value is changed. Seek advice from the corosync mailing list if trying to run larger configurations.

The default is 0 milliseconds.

consensus

This timeout specifies in milliseconds how long to wait for consensus to be achieved before starting a new round of membership configuration. The minimum value for consensus must be 1.2 * token. This value will be automatically calculated at 1.2 * token if the user doesn’t specify a consensus value.

For two node clusters, a consensus larger than the join timeout but less than token is safe. For three node or larger clusters, consensus should be larger than token. There is an increasing risk of odd membership changes, which still guarantee virtual synchrony, as node count grows if consensus is less than token.

The default is 3600 milliseconds.

The default is 200 milliseconds.

downcheck

This timeout specifies in milliseconds how long to wait before checking that a network interface is back up after it has been downed.

The default is 1000 milliseconds.

fail_recv_const

This constant specifies how many rotations of the token without receiving any of the messages when messages should be received may occur before a new configuration is formed.

The default is 2500 failures to receive a message.

seqno_unchanged_const

This constant specifies how many rotations of the token without any multicast traffic should occur before the hold timer is started.

The default is 30 rotations.

heartbeat_failures_allowed

[HeartBeating mechanism] Configures the optional HeartBeating mechanism for faster failure detection. Keep in mind that engaging this mechanism in lossy networks could cause faulty loss declaration as the mechanism relies on the network for heartbeating.

So as a rule of thumb use this mechanism if you require improved failure in low to medium utilized networks.

This constant specifies the number of heartbeat failures the system should tolerate before declaring heartbeat failure e.g 3. Also if this value is not set or is 0 then the heartbeat mechanism is not engaged in the system and token rotation is the method of failure detection

The default is 0 (disabled).

max_network_delay

[HeartBeating mechanism] This constant specifies in milliseconds the approximate delay that your network takes to transport one packet from one machine to another. This value is to be set by system engineers and please don’t change if not sure as this effects the failure detection mechanism using heartbeat.

The default is 50 milliseconds.

window_size

This constant specifies the maximum number of messages that may be sent on one token rotation. If all processors perform equally well, this value could be large (300), which would introduce higher latency from origination to delivery for very large rings. To reduce latency in large rings(16+), the defaults are a safe compromise. If 1 or more slow processor(s) are present among fast processors, window_size should be no larger than 256000 / netmtu to avoid overflow of the kernel receive buffers. The user is notified of this by the display of a retransmit list in the notification logs. There is no loss of data, but performance is reduced when these errors occur.

The default is 50 messages.

max_messages

This constant specifies the maximum number of messages that may be sent by one processor on receipt of the token. The max_messages parameter is limited to 256000 / netmtu to prevent overflow of the kernel transmit buffers.

The default is 17 messages.

miss_count_const

This constant defines the maximum number of times on receipt of a token a message is checked for retransmission before a retransmission occurs. This parameter is useful to modify for switches that delay multicast packets compared to unicast packets. The default setting works well for nearly all modern switches.

The default is 5 messages.

knet_pmtud_interval

How often the knet PMTUd runs to look for network MTU changes. Value in seconds, default: 30

knet_mtu

Switch between manual and automatic MTU discovery. A value of 0 means automatic, other values set a manual MTU. In a setup with multiple interfaces, please specify the lowest MTU of the selected interfaces.

The default value is 0.

block_unlisted_ips

Allow UDPU and KNET to drop packets from IP addresses that are not known (nodes which don’t exist in the nodelist) to corosync. Value is yes or no.

This feature is mainly to protect against the joining of nodes with outdated configurations after a cluster split. Another use case is to allow the atomic merge of two independent clusters.

Changing the default value is not recommended, the overhead is tiny and an existing cluster may fail if corosync is started on an unlisted node with an old configuration.

The default value is yes.

cancel_token_hold_on_retransmit

Allows Corosync to hold token by representative when there is too much retransmit messages. This allows network to process increased load without overloading it. Used mechanism is same as described for hold directive.

Some deployments may prefer to never hold token when there is retransmit messages. If so, option should be set to yes.

The default value is no.

Within the logging directive, there are several configuration options which are all optional.

The following 3 options are valid only for the top level logging directive:
timestamp

This specifies that a timestamp is placed on all log messages. It can be one of off (no timestamp), on (second precision timestamp) or hires (millisecond precision timestamp - only when supported by LibQB).

The default is hires (or on if hires is not supported).

fileline

This specifies that file and line should be printed.

The default is off.

function_name

This specifies that the code function name should be printed.

The default is off.

blackbox

This specifies that blackbox functionality should be enabled.

The default is on.

The following options are valid both for top level logging directive and they can be overridden in logger_subsys entries.
to_stderr
to_logfile
to_syslog

These specify the destination of logging output. Any combination of these options may be specified. Valid options are yes and no.

The default is syslog and stderr.

Please note, if you are using to_logfile and want to rotate the file, use logrotate(8) with the option copytruncate. eg.
/var/log/corosync.log {

}

logfile

If the to_logfile directive is set to yes , this option specifies the pathname of the log file.

No default.

logfile_priority

This specifies the logfile priority for this particular subsystem. Ignored if debug is on. Possible values are: alert, crit, debug (same as debug = on), emerg, err, info, notice, warning.

The default is: info.

syslog_facility

This specifies the syslog facility type that will be used for any messages sent to syslog. options are daemon, local0, local1, local2, local3, local4, local5, local6 & local7.

The default is daemon.

syslog_priority

This specifies the syslog level for this particular subsystem. Ignored if debug is on. Possible values are: alert, crit, debug (same as debug = on), emerg, err, info, notice, warning.

The default is: info.

The default is off.

Within the logging directive, logger_subsys directives are optional.

Within the logger_subsys sub-directive, all of the above logging configuration options are valid and can be used to override the default settings. The subsys entry, described below, is mandatory to identify the subsystem.

Within the quorum directive it is possible to specify the quorum algorithm to use with the
provider

directive. At the time of writing only corosync_votequorum is supported. See votequorum(5) for configuration options.

Within the nodelist directive it is possible to specify specific information about nodes in cluster. Directive can contain only node sub-directive, which specifies every node that should be a member of the membership, and where non-default options are needed. Every node must have at least ring0_addr field filled.

Every node that should be a member of the membership must be specified.

Possible options are:
ringX_addr

This specifies IP or network hostname address of the particular node. X is a link number.

1.

Within the system directive it is possible to specify system options.

Possible options are:
qb_ipc_type

This specifies type of IPC to use. Can be one of native (default), shm and socket. Native means one of shm or socket, depending on what is supported by OS. On systems with support for both, SHM is selected. SHM is generally faster, but need to allocate ring buffer file in /dev/shm.

sched_rr

Should be set to yes (default) if corosync should try to set round robin realtime scheduling with maximal priority to itself. When setting of scheduler fails, fallback to set maximal priority.

priority

Set priority of corosync process. Valid only when sched_rr is set to no. Can be ether numeric value with similar meaning as nice(1) or max / min meaning maximal / minimal priority (so minimal / maximal nice value).

move_to_root_cgroup

Can be one of yes (Corosync always moves itself to root cgroup), no (Corosync never tries to move itself to root cgroup) or auto (Corosync first checks if sched_rr is enabled, and if so, it tries to set round robin realtime scheduling with maximal priority to itself. If setting of priority fails, corosync tries to move itself to root cgroup and retries setting of priority).

This feature is available only for systems with cgroups v1 with RT sched enabled (Linux with CONFIG_RT_GROUP_SCHED kernel option) and cgroups v2.

It’s worth noting that currently (May 3 2021) cgroup2 doesnât yet support control of realtime processes and the cpu controller can only be enabled when all RT processes are in the root cgroup (applies only for kernel with CONFIG_RT_GROUP_SCHED enabled). So when move_to_root_cgroup is disabled, kernel is compiled with CONFIG_RT_GROUP_SCHED and systemd is used, it may be impossible to make systemd options like CPUQuota working correctly until corosync is stopped.

Also when moving to root cgroup is enforced and used together with cgroup2 and systemd it makes impossible (most of the time) for journald to add systemd specific metadata (most importantly _SYSTEMD_UNIT) properly, because corosync is moved out of cgroup created by systemd. This means it is not possible to filter corosync logged messages based on these metadata (for example using -u or _SYSTEMD_UNIT=UNIT pattern) and also running systemctl status doesn’t display (all) corosync log messages. The problem is even worse because journald caches pid for some time (approx. 5 sec) so initial corosync messages have correct metadata.

allow_knet_handle_fallback

If knet handle creation fails using privileged operations, allow fallback to creating knet handle using unprivileged operations. Defaults to no, meaning if privileged knet handle creation fails, corosync will refuse to start.

The knet handle will always be created using privileged operations if possible, setting this to yes only allows fallback to unprivileged operations. This fallback may result in performance issues, but if running in an unprivileged environment, e.g. as a normal user or in unprivileged container, this may be required.

state_dir

Existing directory where corosync should chdir into. Corosync stores important state files and blackboxes there.

The default is /var/lib/corosync.

Within the resources directive it is possible to specify options for resources.

Possible option is:
watchdog_device

(Valid only if Corosync was compiled with watchdog support.)
Watchdog device to use, for example /dev/watchdog. If unset, empty or "off", no watchdog is used.

In a cluster with properly configured power fencing a watchdog provides no additional value. On the other hand, slow watchdog communication may incur multi-second delays in the Corosync main loop, potentially breaking down membership. IPMI watchdogs are particularly notorious in this regard: read about kipmid_max_busy_us in IPMI.txt in the Linux kernel documentation.

Within the nozzle directive it is possible to specify options for a libnozzle device. This is a pseudo ethernet device that routes network traffic through a channel on the corosync knet network (NOT cpg or any corosync internal service) to other nodes in the cluster. This allows applications to take advantage of knet features such as multipathing, automatic failover, link switching etc. Note that libnozzle is not a reliable transport, but you can tunnel TCP through it for reliable communications.
libnozzle also supports optional interface up/down scripts that are kept under a /etc/corosync/updown.d/ directory. See the knet documentation for more information.
Only one nozzle device is allowed.
The nozzle stanza takes several options:

ipprefix

specifies the IP address prefix for the nozzle device (see above)

macaddr

Specifies the MAC address prefix for the nozzle device. As for the IP address, the bottom part of the MAC address will be filled in with the node id. In this case no prefix applies, the bottom two bytes of the MAC address will always be overwritten with the node id. So specifying macaddr: 54:54:12:24:12:12 on nodeid 1 will result in it having a MAC address of 54:54:12:24:00:01

TO ADD A NEW NODE TO THE CLUSTER

For example to add a node with address 10.24.38.108 with nodeid 3. The node has the name NEW (in DNS or /etc/hosts) and is not currently running corosync. The current corosync.conf nodelist looks like this:

nodelist {
node {
nodeid: 1
ring0_addr: 10.24.38.101
name: node1
}
node {
nodeid: 2
ring0_addr: 10.24.38.102
name: node2

}
}

Add a new entry for the node below the existing nodes. Node entries don’t have to be in nodeid order, but it will help keep you sane. So the nodelist now looks like this:

nodelist {
node {
nodeid: 1
ring0_addr: 10.24.38.101
name: node1
}
node {
nodeid: 2
ring0_addr: 10.24.38.102
name: node2

}
node {
nodeid: 3
ring0_addr: 10.24.38.108
name: NEW

}
}

This file must then be copied onto all three nodes - the existing two nodes, and the new one. On one of the existing corosync nodes, tell corosync to re-read the updated config file into memory:

corosync-cfgtool -R

This command only needs to be run on one node in the cluster. You may then start corosync on the NEW node and it should join the cluster. If this doesn’t work as expected then check the communications between all three nodes is working, and check the syslog files on all nodes for more information. It’s important to note that the key bit of information about a node failing to join might be on a different node than you expect.

TO REMOVE A NODE FROM THE CLUSTER

This is the reverse procedure to ’Adding a node’ above. First you need to shut down the node you will be removing from the cluster.

corosync-cfgtool -H

Then delete the nodelist stanza from corosync.conf and finally update corosync on the remaining nodes by running

corosync-cfgtool -R

on one of them.

ADDRESS RESOLUTION

corosync resolves ringX_addr names/IP addresses using the getaddrinfo(3) call with respect of totem.ip_version setting.

getaddrinfo() function uses a sophisticated algorithm to sort node addresses into a preferred order and corosync always chooses the first address in that list of the required family. As such it is essential that your DNS or /etc/hosts files are correctly configured so that all addresses for ringX appear on the same network (or are reachable with minimal hops) and over the same IP protocol. If this is not the case then some nodes might not be able to join the cluster. It is possible to override the search order used by getaddrinfo() using the configuration file /etc/gai.conf(5) if necessary, but this is not recommended.

If there is any doubt about the order of addresses returned from getaddrinfo() then it might be simpler to use IP addresses (v4 or v6) in the ringX_addr field.

FILES

/etc/corosync/corosync.conf

The corosync executive configuration file.

SEE ALSO

corosync_overview(7), votequorum(5), corosync-qdevice(8), logrotate(8) getaddrinfo(3) gai.conf(5)