Dzahn has uploaded a new change for review. (
https://gerrit.wikimedia.org/r/368124 )
Change subject: base::monitoring: make it possible to disable monitoring
......................................................................
base::monitoring: make it possible to disable monitoring
Change-Id: Idff6a66567390c438ba2b79dda7e74b8483d3e61
---
M modules/base/manifests/monitoring/host.pp
M modules/profile/manifests/base.pp
2 files changed, 100 insertions(+), 67 deletions(-)
git pull ssh://gerrit.wikimedia.org:29418/operations/puppet
refs/changes/24/368124/1
diff --git a/modules/base/manifests/monitoring/host.pp
b/modules/base/manifests/monitoring/host.pp
index c99aec4..d6943df 100644
--- a/modules/base/manifests/monitoring/host.pp
+++ b/modules/base/manifests/monitoring/host.pp
@@ -29,54 +29,78 @@
$nrpe_check_disk_options = '-w 6% -c 3% -l -e -A -i "/srv/sd[a-b][1-3]"
--exclude-type=tracefs',
$nrpe_check_disk_critical = false,
$raid_write_cache_policy = undef,
+ $enabled = true,
) {
include ::base::puppet::params # In order to be able to use some variables
- # RAID checks
- class { 'raid':
- write_cache_policy => $raid_write_cache_policy,
+ $ensure = $enabled ? {
+ true => 'present',
+ false => 'absent',
}
- ::monitoring::host { $::hostname: }
+ if $enabled {
+ # TODO: FIX ALL THE THINGS in RAID CLASS TOO, SIGH
+ # RAID checks
+ class { 'raid':
+ write_cache_policy => $raid_write_cache_policy,
+ }
+
+
+ }
+
+ ::sudo::user { 'nagios_puppetrun':
+ ensure => $ensure,
+ user => 'nagios',
+ privileges => ['ALL = NOPASSWD:
/usr/local/lib/nagios/plugins/check_puppetrun'],
+ }
+
+ ::monitoring::host { $::hostname:
+ ensure => $ensure,
+ }
::monitoring::service { 'ssh':
+ ensure => $ensure,
description => 'SSH',
check_command => 'check_ssh',
}
- file { '/usr/local/lib/nagios/plugins/check_puppetrun':
- ensure => present,
- owner => 'root',
- group => 'root',
- mode => '0555',
- source => 'puppet:///modules/base/monitoring/check_puppetrun';
+ ::nrpe::monitor_service { 'disk_space':
+ ensure => $ensure,
+ description => 'Disk space',
+ critical => $nrpe_check_disk_critical,
+ nrpe_command => "/usr/lib/nagios/plugins/check_disk
${nrpe_check_disk_options}",
}
- file { '/usr/local/lib/nagios/plugins/check_eth':
- ensure => present,
- owner => 'root',
- group => 'root',
- mode => '0555',
- content => template('base/check_eth.erb'),
- }
- file { '/usr/lib/nagios/plugins/check_sysctl':
- ensure => present,
- owner => 'root',
- group => 'root',
- mode => '0555',
- source => 'puppet:///modules/base/check_sysctl',
+
+ ::nrpe::monitor_service { 'dpkg':
+ ensure => $ensure,
+ description => 'DPKG',
+ nrpe_command => '/usr/local/lib/nagios/plugins/check_dpkg',
}
- file { '/usr/lib/nagios/plugins/check-fresh-files-in-dir.py':
- ensure => present,
- owner => 'root',
- group => 'root',
- mode => '0555',
- source =>
'puppet:///modules/base/monitoring/check-fresh-files-in-dir.py',
+ $warninginterval = $base::puppet::params::freshnessinterval
+ $criticalinterval = $base::puppet::params::freshnessinterval * 2
+ ::nrpe::monitor_service { 'puppet_checkpuppetrun':
+ ensure => $ensure,
+ description => 'puppet last run',
+ nrpe_command => "/usr/bin/sudo
/usr/local/lib/nagios/plugins/check_puppetrun -w ${warninginterval} -c
${criticalinterval}",
}
- ::sudo::user { 'nagios_puppetrun':
- user => 'nagios',
- privileges => ['ALL = NOPASSWD:
/usr/local/lib/nagios/plugins/check_puppetrun'],
+ ::nrpe::monitor_service {'check_eth':
+ ensure => $ensure,
+ description => 'configured eth',
+ nrpe_command => '/usr/local/lib/nagios/plugins/check_eth',
+ }
+
+ ::nrpe::monitor_service { 'check_dhclient':
+ ensure => $ensure,
+ description => 'dhclient process',
+ nrpe_command => '/usr/lib/nagios/plugins/check_procs -w 0:0 -c 0:0 -C
dhclient',
+ }
+
+ ::nrpe::monitor_service { 'check_salt_minion':
+ ensure => $ensure,
+ description => 'salt-minion processes',
+ nrpe_command => "/usr/lib/nagios/plugins/check_procs -w 1: -c 1:5
--ereg-argument-array '/usr/bin/python /usr/bin/salt-minion'",
}
# Check for disk usage on the root partition for labs instances
@@ -85,57 +109,66 @@
# to reflect this check name
if $::realm == 'labs' {
::nrpe::monitor_service { 'root_disk_space':
+ ensure => $ensure,
description => 'Disk space on /',
nrpe_command => '/usr/lib/nagios/plugins/check_disk -w 5% -c 2% -l
-e -p /',
}
}
- ::nrpe::monitor_service { 'disk_space':
- description => 'Disk space',
- critical => $nrpe_check_disk_critical,
- nrpe_command => "/usr/lib/nagios/plugins/check_disk
${nrpe_check_disk_options}",
+ file { '/usr/local/lib/nagios/plugins/check_puppetrun':
+ ensure => $ensure,
+ owner => 'root',
+ group => 'root',
+ mode => '0555',
+ source => 'puppet:///modules/base/monitoring/check_puppetrun';
}
- ::nrpe::monitor_service { 'dpkg':
- description => 'DPKG',
- nrpe_command => '/usr/local/lib/nagios/plugins/check_dpkg',
+ file { '/usr/local/lib/nagios/plugins/check_eth':
+ ensure => $ensure,
+ owner => 'root',
+ group => 'root',
+ mode => '0555',
+ content => template('base/check_eth.erb'),
}
- $warninginterval = $base::puppet::params::freshnessinterval
- $criticalinterval = $base::puppet::params::freshnessinterval * 2
- ::nrpe::monitor_service { 'puppet_checkpuppetrun':
- description => 'puppet last run',
- nrpe_command => "/usr/bin/sudo
/usr/local/lib/nagios/plugins/check_puppetrun -w ${warninginterval} -c
${criticalinterval}",
+
+ file { '/usr/lib/nagios/plugins/check_sysctl':
+ ensure => $ensure,
+ owner => 'root',
+ group => 'root',
+ mode => '0555',
+ source => 'puppet:///modules/base/check_sysctl',
}
- ::nrpe::monitor_service {'check_eth':
- description => 'configured eth',
- nrpe_command => '/usr/local/lib/nagios/plugins/check_eth',
+
+ file { '/usr/lib/nagios/plugins/check-fresh-files-in-dir.py':
+ ensure => $ensure,
+ owner => 'root',
+ group => 'root',
+ mode => '0555',
+ source =>
'puppet:///modules/base/monitoring/check-fresh-files-in-dir.py',
}
- ::nrpe::monitor_service { 'check_dhclient':
- description => 'dhclient process',
- nrpe_command => '/usr/lib/nagios/plugins/check_procs -w 0:0 -c 0:0 -C
dhclient',
- }
- ::nrpe::monitor_service { 'check_salt_minion':
- description => 'salt-minion processes',
- nrpe_command => "/usr/lib/nagios/plugins/check_procs -w 1: -c 1:5
--ereg-argument-array '^/usr/bin/python /usr/bin/salt-minion'",
- }
+
if $::initsystem == 'systemd' {
+
file { '/usr/local/lib/nagios/plugins/check_systemd_state':
- ensure => present,
+ ensure => $ensure,
source => 'puppet:///modules/base/check_systemd_state.py',
owner => 'root',
group => 'root',
mode => '0555',
}
- ::nrpe::monitor_service { 'check_systemd_state':
+
+ ::nrpe::monitor_service { 'check_systemd_state':
+ ensure => $ensure,
description => 'Check systemd state',
nrpe_command =>
'/usr/local/lib/nagios/plugins/check_systemd_state',
}
+
}
if $::productname == 'PowerEdge R320' {
file { '/usr/local/lib/nagios/plugins/check_cpufreq':
- ensure => present,
+ ensure => $ensure,
source => 'puppet:///modules/base/monitoring/check_cpufreq',
owner => 'root',
group => 'root',
@@ -143,8 +176,9 @@
}
::nrpe::monitor_service { 'check_cpufreq':
- description => 'CPU frequency',
- nrpe_command => '/usr/local/lib/nagios/plugins/check_cpufreq 600',
+ ensure => $ensure,
+ description => 'CPU frequency',
+ nrpe_command => '/usr/local/lib/nagios/plugins/check_cpufreq 600',
}
}
}
diff --git a/modules/profile/manifests/base.pp
b/modules/profile/manifests/base.pp
index 7567265..9601239 100644
--- a/modules/profile/manifests/base.pp
+++ b/modules/profile/manifests/base.pp
@@ -91,13 +91,12 @@
}
# unless disabled in Hiera, have Icinga monitoring (T151632)
- if $monitoring {
- class { '::base::monitoring::host':
- contact_group => $group_contact,
- nrpe_check_disk_options => $check_disk_options,
- nrpe_check_disk_critical => $check_disk_critical,
- raid_write_cache_policy => $check_raid_policy,
- }
+ class { '::base::monitoring::host':
+ enabled => $monitoring,
+ contact_group => $group_contact,
+ nrpe_check_disk_options => $check_disk_options,
+ nrpe_check_disk_critical => $check_disk_critical,
+ raid_write_cache_policy => $check_raid_policy,
}
if os_version('ubuntu == trusty') {
--
To view, visit https://gerrit.wikimedia.org/r/368124
To unsubscribe, visit https://gerrit.wikimedia.org/r/settings
Gerrit-MessageType: newchange
Gerrit-Change-Id: Idff6a66567390c438ba2b79dda7e74b8483d3e61
Gerrit-PatchSet: 1
Gerrit-Project: operations/puppet
Gerrit-Branch: production
Gerrit-Owner: Dzahn <dz...@wikimedia.org>
_______________________________________________
MediaWiki-commits mailing list
MediaWiki-commits@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits
