WarlockFish/fedora-infra_ansible
1
0
Fork 0
mirror of https://pagure.io/fedora-infra/ansible.git synced 2026-05-11 10:32:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

Nagios: remove first batch of services

Browse Source 
This removes the known-good things we've had in Zabbix for a while -
RAID, disk space, processes, and mail queue. It also removes swap which
we've decided we don't need.

Also includes some FS overrides on the Zabbix side so the relevant
NFS mounts get monitored on the OCI, and pkgs hosts, as per Nagios had.

Signed-off-by: Greg Sutcliffe <fedora@emeraldreverie.org>
This commit is contained in:
Greg Sutcliffe
2026-01-09 16:27:45 +00:00
committed by gwmngilfen
parent 3cb16462ea
commit 93ed0457e0
10 changed files with 8 additions and 176 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
inventory/group_vars/oci_registry
View File

@@ -8,3 +8,5 @@ ipa_host_group_desc: OCI Registry service
nfs_mount_opts: "rw,hard,bg,intr,noatime,nodev,nosuid,sec=sys,nfsvers=3"
primary_auth_source: ipa
tcp_ports: [5000]
zabbix_macros:
'VFS.FS.FSTYPE.MATCHES': '^(btrfs|ext2|ext3|ext4|reiser|xfs|ffs|ufs|jfs|jfs2|vxfs|hfs|apfs|refs|ntfs|fat32|zfs|nfs)$'

2
inventory/group_vars/oci_registry_stg
View File

@@ -7,3 +7,5 @@ ipa_host_group: oci-registry
ipa_host_group_desc: OCI Registry service
nfs_mount_opts: "rw,hard,bg,intr,noatime,nodev,nosuid,sec=sys,nfsvers=3"
tcp_ports: [5000]
zabbix_macros:
'VFS.FS.FSTYPE.MATCHES': '^(btrfs|ext2|ext3|ext4|reiser|xfs|ffs|ufs|jfs|jfs2|vxfs|hfs|apfs|refs|ntfs|fat32|zfs|nfs)$'

2
inventory/group_vars/pkgs
View File

@@ -46,3 +46,5 @@ tcp_ports: [80, 443]
# There vars are used to configure mod_wsgi
wsgi_procs: 20
wsgi_threads: 5
zabbix_macros:
'VFS.FS.FSTYPE.MATCHES': '^(btrfs|ext2|ext3|ext4|reiser|xfs|ffs|ufs|jfs|jfs2|vxfs|hfs|apfs|refs|ntfs|fat32|zfs|nfs)$'

2
inventory/group_vars/pkgs_stg
View File

@@ -41,3 +41,5 @@ tcp_ports: [80, 443, 8444, 8443, 8445]
# There vars are used to configure mod_wsgi
wsgi_procs: 4
wsgi_threads: 4
zabbix_macros:
'VFS.FS.FSTYPE.MATCHES': '^(btrfs|ext2|ext3|ext4|reiser|xfs|ffs|ufs|jfs|jfs2|vxfs|hfs|apfs|refs|ntfs|fat32|zfs|nfs)$'

28
roles/nagios_server/files/nagios/services/copr.cfg
View File

@@ -1,31 +1,3 @@
define service {
hostgroup_name copr_back_aws, copr_back_dev_aws
service_description Disk space on Copr Backend used for build results
check_command check_by_nrpe!check_disk_/var/lib/copr/public_html
use disktemplate
}
define service {
hostgroup_name copr_dist_git_aws, copr_dist_git_dev_aws
service_description Disk space on Copr DistGit used for imported packages
check_command check_by_nrpe!check_disk_/var/lib/dist-git
use disktemplate
}
define service {
hostgroup_name copr_front_aws, copr_front_dev_aws
service_description Disk space on Copr Frontend used for the database
check_command check_by_nrpe!check_disk_/srv/copr-fe
use disktemplate
}
define service {
hostgroup_name copr_keygen_aws, copr_keygen_dev_aws
service_description Disk space on Copr Keygen used for GPG keys
check_command check_by_nrpe!check_disk_/var/lib/copr-keygen
use disktemplate
}
define service {
hostgroup_name copr_back_aws, copr_back_dev_aws
service_description The copr-ping package builds

39
roles/nagios_server/files/nagios/services/procs.cfg
View File

@@ -1,25 +1,3 @@
define service {
hostgroup virtservers
service_description Total Processes
check_command check_by_nrpe!check_total_procs
use defaulttemplate
}
define service {
hostgroup virtservers
service_description Zombie Processes
check_command check_by_nrpe!check_zombie_procs
use defaulttemplate
}
define service {
hostgroup virtservers
service_description Cron Daemon
check_command check_by_nrpe!check_cron
use defaulttemplate
}
define service {
hostgroup proxies
service_description Varnish Process
@@ -27,15 +5,6 @@ define service {
use defaulttemplate
}
# TODO: uncomment when retrace is moved
#define service {
# hostgroup retrace
# service_description Total Processes
# check_command check_by_nrpe!check_total_procs
# use retracetemplate
#}
define service {
hostgroup sign_bridge
service_description Sigul bridge Process
@@ -43,17 +12,9 @@ define service {
use defaulttemplate
}
define service {
hostgroup all, !mincheckgrp
service_description Rsyslogd Process
check_command check_by_nrpe!check_rsyslogd_proc
use defaulttemplate
}
define service {
hostgroup pagure
service_description Systemd Units
check_command check_by_nrpe!check_systemd_units
use defaulttemplate
}

16
roles/nagios_server/files/nagios/services/raid.cfg
View File

@@ -1,16 +0,0 @@
# Special-case this so it can use retracetemplate.
# TODO: uncomment when retrace is moved
#define service {
# hostgroup_name retrace
# service_description Check_Raid
# check_command check_by_nrpe!check_raid
# use retracetemplate
#}
# Everything else uses this group.
define service {
hostgroup_name CheckRaid
service_description Check_Raid
check_command check_by_nrpe!check_raid
use lighttemplate
}

71
roles/nagios_server/files/nagios/services/rdu3_internal/disk.cfg
View File

@@ -1,77 +1,6 @@
define service {
hostgroup_name all, !mincheckgrp, !logging_rdu3
service_description Disk_Space_/
check_command check_by_nrpe!check_disk_/
use disktemplate
}
define service {
hostgroup_name all, !mincheckgrp
service_description Disk Space /boot
check_command check_by_nrpe!check_disk_/boot
use disktemplate
}
# TODO: uncomment when qahardware is back online
#define service {
# hostgroup_name qahardware
# service_description Disk Space /srv
# check_command check_by_nrpe!check_disk_/srv
# use disktemplate
#}
define service {
host_name log01.rdu3.fedoraproject.org
service_description Disk space /var/log
check_command check_by_nrpe!check_disk_/var/log
use disktemplate
}
define service {
hostgroup_name pkgs
service_description Check read-only filesystem
check_command check_by_nrpe!check_readonly_fs
use disktemplate
}
define service {
hostgroup_name pkgs
service_description Disk space /srv/cache/lookaside
check_command check_by_nrpe!check_disk_/srv/cache/lookaside
use disktemplate
}
define service {
hostgroup_name koji
service_description Disk space /
check_command check_by_nrpe!check_disk_/
use ppc-secondarytemplate
}
define service {
hostgroup_name retrace
service_description Disk space /
check_command check_by_nrpe!check_disk_/
use retracetemplate
}
define service {
hostgroup_name retrace
service_description Disk Space for huge /srv
check_command check_by_nrpe!check_disk_huge_/srv
use disktemplate
}
define service {
hostgroup_name people
service_description Disk space /project
check_command check_by_nrpe!check_disk_/project/
use disktemplate
}
define service {
hostgroup_name oci_registry
service_description Disk space /srv/registry
check_command check_by_nrpe!check_disk_/srv/registry
use disktemplate
}

14
roles/nagios_server/files/nagios/services/swap.cfg
View File

@@ -1,14 +0,0 @@
define service {
hostgroup CheckSwap
service_description Swap-Is-Low
check_command check_by_nrpe!check_swap
use criticaltemplate
}
# TODO: uncomment once retrace exists again
#define service {
# hostgroup retrace
# service_description Swap
# check_command check_by_nrpe!check_swap
# use retracetemplate
#}

8
roles/nagios_server/templates/nagios/services/mail_queue.cfg.j2
View File

@@ -1,12 +1,4 @@
{% if nagios_location == "rdu3_internal" %}
define service {
hostgroup nomail
service_description mail_queue
check_command check_by_nrpe!check_postfix_queue
max_check_attempts 7
use defaulttemplate
}
define service {
host_name bastion01.{{datacenter}}.fedoraproject.org
service_description mail_queue_redhat