This module manages pacemaker on Linux distros with the pcs tool.
Apache 2.0
Please log tickets and issues at https://github.com/radez/puppet-pacemaker
This module uses the fact osfamily which is supported by Facter 1.6.1+. If you do not have facter 1.6.1 in your environment, the following manifests will provide the same functionality in site.pp (before declaring any node):
if ! $::osfamily {
case $::operatingsystem {
'RedHat', 'Fedora', 'CentOS', 'Scientific', 'SLC', 'Ascendos', 'CloudLinux', 'PSBM', 'OracleLinux', 'OVS', 'OEL': {
$osfamily = 'RedHat'
}
'ubuntu', 'debian': {
$osfamily = 'Debian'
}
'SLES', 'SLED', 'OpenSuSE', 'SuSE': {
$osfamily = 'Suse'
}
'Solaris', 'Nexenta': {
$osfamily = 'Solaris'
}
default: {
$osfamily = $::operatingsystem
}
}
}
This module is based on work by Dan Radez
Installs Pacemaker and corosync and creates a cluster
class {"pacemaker::corosync":
cluster_name => "cluster_name",
cluster_members => "clu-host-1 clu-host-2 clu-host-3",
}
The pacemaker::corosync resource must be executed on each node
If one wants to set non default settings for corosync the cluster_setup_extras variable can be used
class {"pacemaker::corosync":
cluster_name => "cluster_name",
cluster_members => "clu-host-1 clu-host-2 clu-host-3",
cluster_setup_extras => {'--totem' => '10000', '--ipv6' => ''},
}
class {"pacemaker::stonith":
disable => true,
}
class {"pacemaker::stonith::ipmilan":
address => "10.10.10.100",
username => "admin",
password => "admin",
}
pacemaker::constraint::base{ "$name":
constraint_type => (order|location|colocation),
first_resource => $first_resource,
second_resource => $second_resource,
first_action => $first_action,
second_action => $second_action,
location => $location,
score => $score,
ensure => $ensure}
}
A few different types are provided for several pacemaker resource types. Howerver, all of these types end up wrapping the pcmk_resource provider. The pcmk_resource provider itself essentially wraps the "pcs resource" command. Params such as "resource_params," "group_params" and "clone_params" map transparently to the command line for pcs. I.e., if $clone_params is set to "interleave=true", the pcs command to create the resource looks like:
pcs resource create <res_name> <res_type> ... --clone interleave=true
If $clone_params is undef, --clone is omitted from the "pcs resource create command". Likewise for $group_params and $master_params. Use empty strings for parameters where --clone or --master is desired on the command line without extra parameters.
$meta_params and $op_params behave somewhat similarly, but without the distinction between empty strings and undef. I.e., either "meta " should be present in the pcs resource create command or not at all.
Finally, $resource_params are simply params that show up as options in the command immediately after the resource type without any additional keywords.
Properties can be set clusterwide or per-node using the pacemaker::property function. In the former case, when a property name is not recognized by pacemaker the force parameter needs to be true.
See the pcs man page for documentation for about the "pcs resource create" subcommand and clone, group, meta, op options, etc. Note that older verions of pcs may not be compatible with recent ones in terms of where arguments and options are expected.
The pcmk_resource provider (and all pacmaker::resource:: types) will not update an already existing resource. I.e., if the resource myservice already exists without $clone_params, declaring "pcmk_resource {'myservice': ..." with $clone_params will have no effect. The resource would have to be deleted first (either manually on the command line with "pcs resource delete" or via puppet with ensure => 'absent') -- then the declaration to create 'myservice' with $clone_params succeeds as expected.
## Note that pacemaker::resource::service results in a systemd:
## resource being created if systemd is supported on the OS.
## Otherwise it's an lsb: resource. These examples assume the OS
## supports systemd.
pacemaker::resource::ip {"ip-192.168.201.59":
ip_address => '192.168.201.59',
nic => 'eth3',
cidr_netmask => '',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
#results in Debug: /usr/sbin/pcs resource create ip-192.168.201.59 IPaddr2 ip=192.168.201.59 nic=eth3
# another IP
pacemaker::resource::ip {"ip-192.168.201.58":
ip_address => '192.168.201.58',
tries => 1,
try_sleep => 1,
post_success_sleep => 0,
}
#results in Debug: /usr/sbin/pcs resource create ip-192.168.201.58 IPaddr2 ip=192.168.201.58 cidr_netmask=32
pacemaker::resource::filesystem{ "fs-varlibglanceimages":
device => '192.168.200.100:/mnt/glance',
directory => '/var/lib/glance/images/',
fsoptions => '',
fstype => 'nfs',
clone_params => '',
op_params => 'monitor interval=30s',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
# results in Debug: /usr/sbin/pcs resource create fs-varlibglanceimages Filesystem device=192.168.200.100:/mnt/glance directory=/var/lib/glance/images/ fstype=nfs op monitor interval=30s --clone
pacemaker::resource::service {"lb-haproxy":
service_name => 'haproxy',
op_params => 'monitor start-delay=10s',
clone_params => '',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
# results in Debug: /usr/sbin/pcs resource create lb-haproxy systemd:haproxy op monitor start-delay=10s --clone
pacemaker::resource::service {"cinder-api":
service_name => 'openstack-cinder-api',
clone_params => 'interleave=true',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
# results in Debug: /usr/sbin/pcs resource create cinder-api systemd:openstack-cinder-api --clone interleave=true
pacemaker::resource::ocf {"neutron-scale":
ocf_agent_name => 'neutron:NeutronScale',
clone_params => 'globally-unique=true clone-max=3 interleave=true',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
# results in Debug: /usr/sbin/pcs resource create neutron-scale ocf:neutron:NeutronScale --clone globally-unique=true clone-max=3 interleave=true
pacemaker::property { 'global-bar':
property => 'bar',
value => 'baz',
force => true,
tries => 1,
try_sleep => 1,
}
# results in Debug: /usr/sbin/pcs property set --force bar=baz
pcmk_resource { "galera":
resource_type => "galera",
resource_params => 'enable_creation=true wsrep_cluster_address="gcomm://pcmk-c1a1,pcmk-c1a2,pcmk-c1a3"',
meta_params => "master-max=3 ordered=true",
op_params => 'promote timeout=300s on-fail=block',
master_params => '',
post_success_sleep => 0,
tries => 1,
try_sleep => 1,
}
# Debug results in : /usr/sbin/pcs resource create galera galera enable_creation=true wsrep_cluster_address="gcomm://pcmk-c1a1,pcmk-c1a2,pcmk-c1a3" meta master-max=3 ordered=true op promote timeout=300s on-fail=block --master