100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
|
# File 'manifests/profile/base/pacemaker.pp', line 100
class tripleo::profile::base::pacemaker (
$step = Integer(lookup('step')),
$pcs_tries = lookup('pcs_tries', undef, undef, 20),
$pcs_user = 'hacluster',
$pcs_password = lookup('hacluster_pwd', undef, undef, undef),
$remote_short_node_names = lookup('pacemaker_remote_short_node_names', undef, undef, []),
$remote_node_ips = lookup('pacemaker_remote_node_ips', undef, undef, []),
$remote_authkey = undef,
$remote_reconnect_interval = lookup('pacemaker_remote_reconnect_interval', undef, undef, 60),
$remote_monitor_interval = lookup('pacemaker_remote_monitor_interval', undef, undef, 20),
$remote_tries = lookup('pacemaker_remote_tries', undef, undef, 5),
$remote_try_sleep = lookup('pacemaker_remote_try_sleep', undef, undef, 60),
$cluster_recheck_interval = lookup('pacemaker_cluster_recheck_interval', undef, undef, undef),
$encryption = true,
$resource_op_defaults = undef,
$enable_instanceha = lookup('tripleo::instanceha', undef, undef, false),
$pcsd_bind_addr = undef,
$tls_priorities = lookup('tripleo::pacemaker::tls_priorities', undef, undef, undef),
$cluster_properties = {},
) {
if count($remote_short_node_names) != count($remote_node_ips) {
fail("Count of ${remote_short_node_names} is not equal to count of ${remote_node_ips}")
}
validate_legacy(Hash, 'validate_hash', $cluster_properties)
if $pcs_password == undef {
fail('The $pcs_password param is undefined, did you forget to include tripleo::profile::base::pacemaker in your role?')
}
# During FFU when override keys are set we need to use the old authkey style
# This should be kept until FFU from CentOS 7->8 is being supported
if count(lookup('pacemaker_node_ips_override', undef, undef, [])) > 0 {
$force_old_style_remotes_real = true
} else {
$force_old_style_remotes_real = false
}
Pcmk_resource <| |> {
tries => 10,
try_sleep => 3,
}
if $::hostname == downcase(lookup('pacemaker_short_bootstrap_node_name')) {
$pacemaker_master = true
} else {
$pacemaker_master = false
}
# enable_fencing guides the enablement of the stonith-enabled cluster-wide property
# enable_stonith_resources drives the creation of the stonith resources themselves and happens at
# step2. The reason for step2 is the following:
# During step1 the cluster is created (and also the pcmk remote resources in case of IHA)
# Since stonith resources are created on each node separately we need to have the guarantee that
# all cluster nodes + remote exist before creating stonith resources for them
$enable_fencing = str2bool(lookup('enable_fencing', undef, undef, false)) and $step >= 5
$enable_stonith_resources = str2bool(lookup('enable_fencing', undef, undef, false)) and $step >= 2
if $step >= 1 {
if (lookup('pacemaker_short_node_names_override', undef, undef, undef)) {
$pacemaker_short_node_names = join(lookup('pacemaker_short_node_names_override'), ',')
} else {
$pacemaker_short_node_names = join(lookup('pacemaker_short_node_names'), ',')
}
$pacemaker_cluster_members = downcase(regsubst($pacemaker_short_node_names, ',', ' ', 'G'))
$cluster_setup_extras = {
"totem token=${lookup('corosync_token_timeout', undef, undef, 1000)}" => '',
}
# If pacemaker_node_ips is not empty we want to create the array
# for puppet pacemaker to use as addresses list which is an array of arrays.
$pacemaker_node_ips = lookup('pacemaker_node_ips_override', undef, undef,
lookup('pacemaker_node_ips', undef, undef, []))
if count($pacemaker_node_ips) > 0 {
$pacemaker_node_ips_real = $pacemaker_node_ips.map |$x| { Array([$x]) }
} else {
$pacemaker_node_ips_real = []
}
class { 'pacemaker':
hacluster_pwd => $pcs_password,
}
-> class { 'pacemaker::corosync':
cluster_members => $pacemaker_cluster_members,
setup_cluster => $pacemaker_master,
cluster_setup_extras => $cluster_setup_extras,
remote_authkey => $remote_authkey,
force_authkey => $force_old_style_remotes_real,
cluster_members_addr => $pacemaker_node_ips_real,
pcsd_bind_addr => $pcsd_bind_addr,
tls_priorities => $tls_priorities,
}
if $pacemaker_master {
class { 'pacemaker::stonith':
disable => !$enable_fencing,
tries => $pcs_tries,
}
Class['pacemaker::stonith'] -> Exec<|tag == 'pacemaker-scaleup'|>
}
if $enable_stonith_resources {
include tripleo::fencing
# enable stonith after all Pacemaker resources have been created
Pcmk_resource<||> -> Class['tripleo::fencing']
Pcmk_constraint<||> -> Class['tripleo::fencing']
Exec <| tag == 'pacemaker_constraint' |> -> Class['tripleo::fencing']
# enable stonith after all fencing devices have been created
Class['tripleo::fencing'] -> Pcmk_property<|title == 'Enable STONITH'|>
}
# We have pacemaker remote nodes configured so let's add them as resources
# We do this during step 1 right after wait-for-settle, because during step 2
# resources might already be created on pacemaker remote nodes and we need
# a guarantee that remote nodes are already up
if $pacemaker_master and count($remote_short_node_names) > 0 {
# Creates a { "node" => "ip_address", ...} hash
$remotes_hash = hash(zip($remote_short_node_names, $remote_node_ips))
$remote_short_node_names.each |String $remote_short_node| {
pacemaker::resource::remote { $remote_short_node:
remote_address => $remotes_hash[$remote_short_node],
reconnect_interval => $remote_reconnect_interval,
op_params => "monitor interval=${remote_monitor_interval}",
tries => $remote_tries,
try_sleep => $remote_try_sleep,
pcs_user => $pcs_user,
pcs_password => $pcs_password,
# When we force the use of old style remotes we must also use --force
force => $force_old_style_remotes_real,
force_oldstyle => $force_old_style_remotes_real,
before => Exec["exec-wait-for-${remote_short_node}"],
notify => Exec["exec-wait-for-${remote_short_node}"],
}
$check_command = "pcs status | grep -q -e \"${remote_short_node}[[:blank:]].*Started\""
exec { "exec-wait-for-${remote_short_node}":
path => '/usr/sbin:/usr/bin:/sbin:/bin',
command => $check_command,
unless => $check_command,
timeout => 30,
tries => 180,
try_sleep => 10,
tag => 'remote_ready',
}
}
}
}
if $enable_instanceha and $pacemaker_master {
include tripleo::profile::base::pacemaker::instance_ha
}
if ($step >= 2 and $pacemaker_master) {
create_resources('pacemaker::property', $cluster_properties)
if ! $enable_instanceha {
include pacemaker::resource_defaults
}
# Set pacemaker resource op defaults only when specified
if $resource_op_defaults != undef {
class { 'pacemaker::resource_op_defaults':
tries => $pcs_tries,
defaults => $resource_op_defaults,
}
}
# When we have a non-zero number of pacemaker remote nodes we
# want to set the cluster-recheck-interval property to something
# lower (unless the operator has explicitly set a value)
if count($remote_short_node_names) > 0 and $cluster_recheck_interval == undef {
pacemaker::property{ 'cluster-recheck-interval-property':
property => 'cluster-recheck-interval',
value => '60s',
tries => $pcs_tries,
}
} elsif $cluster_recheck_interval != undef {
pacemaker::property{ 'cluster-recheck-interval-property':
property => 'cluster-recheck-interval',
value => $cluster_recheck_interval,
tries => $pcs_tries,
}
}
# See LP#1948032. We want to ensure that we set the default resource time out
# before we create any bundles. Because the default timeout affects how long
# we wait for podman, and on a starved system the default of 20s might bite us
# too early on
Pcmk_resource_default<||> -> Pcmk_bundle<||>
}
}
|