path: root/roles/os_zabbix/vars/template_os_linux.yml

---
g_template_os_linux:
  name: Template OS Linux
  zitems:
  - key: kernel.uname.sysname
    applications:
    - Kernel
    value_type: string

  - key: kernel.all.cpu.wait.total
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.cpu.irq.hard
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.cpu.idle
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.uname.distro
    applications:
    - Kernel
    value_type: string

  - key: kernel.uname.nodename
    applications:
    - Kernel
    value_type: string

  - key: kernel.all.cpu.irq.soft
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.load.15_minute
    applications:
    - Kernel
    value_type: float

  - key: kernel.all.cpu.sys
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.load.5_minute
    applications:
    - Kernel
    value_type: float

  - key: kernel.all.cpu.nice
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.load.1_minute
    applications:
    - Kernel
    value_type: float

  - key: kernel.uname.version
    applications:
    - Kernel
    value_type: string

  - key: kernel.all.uptime
    applications:
    - Kernel
    value_type: int

  - key: kernel.all.cpu.user
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.uname.machine
    applications:
    - Kernel
    value_type: string

  - key: hinv.ncpu
    applications:
    - Kernel
    value_type: int

  - key: kernel.all.cpu.steal
    applications:
    - Kernel
    value_type: float
    units: '%'

  - key: kernel.all.pswitch
    applications:
    - Kernel
    value_type: int

  - key: kernel.uname.release
    applications:
    - Kernel
    value_type: string

  - key: proc.nprocs
    applications:
    - Kernel
    value_type: int

  # Memory Items
  - key: mem.freemem
    applications:
    - Memory
    value_type: int
    description: "PCP: free system memory metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: mem.util.bufmem
    applications:
    - Memory
    value_type: int
    description: "PCP: Memory allocated for buffer_heads.; I/O buffers metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: swap.used
    applications:
    - Memory
    value_type: int
    description: "PCP: swap used metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: swap.length
    applications:
    - Memory
    value_type: int
    description: "PCP: total swap available metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: mem.physmem
    applications:
    - Memory
    value_type: int
    description: "PCP: The value of this metric corresponds to the \"MemTotal\" field reported by /proc/meminfo. Note that this does not necessarily correspond to actual installed physical memory - there may be areas of the physical address space mapped as ROM in various peripheral devices and the bios may be mirroring certain ROMs in RAM."
    multiplier: 1024
    units: B

  - key: swap.free
    applications:
    - Memory
    value_type: int
    description: "PCP: swap free metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: mem.util.available
    applications:
    - Memory
    value_type: int
    description: "PCP: The amount of memory that is available for a new workload, without pushing the system into swap. Estimated from MemFree, Active(file), Inactive(file), and SReclaimable, as well as the \"low\" watermarks from /proc/zoneinfo.; available memory from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: mem.util.used
    applications:
    - Memory
    value_type: int
    description: "PCP: Used memory is the difference between mem.physmem and mem.freemem; used memory metric from /proc/meminfo"
    multiplier: 1024
    units: B

  - key: mem.util.cached
    applications:
    - Memory
    value_type: int
    description: "PCP: Memory used by the page cache, including buffered file data.  This is in-memory cache for files read from the disk (the pagecache) but doesn't include SwapCached.; page cache metric from /proc/meminfo"
    multiplier: 1024
    units: B

  zdiscoveryrules:
  - name: disc.filesys
    key: disc.filesys
    lifetime: 1
    description: "Dynamically register the filesystems"

  - name: disc.disk
    key: disc.disk
    lifetime: 1
    description: "Dynamically register disks on a node"

  - name: disc.network
    key: disc.network
    lifetime: 1
    description: "Dynamically register network interfaces on a node"

  zitemprototypes:
  - discoveryrule_key: disc.filesys
    name: "disc.filesys.full.{#OSO_FILESYS}"
    key: "disc.filesys.full[{#OSO_FILESYS}]"
    value_type: float
    description: "PCP filesys.full option.  This is the percent full returned from pcp filesys.full"
    applications:
    - Disk

  - discoveryrule_key: disc.filesys
    name: "Percentage of used inodes on {#OSO_FILESYS}"
    key: "disc.filesys.inodes.pused[{#OSO_FILESYS}]"
    value_type: float
    description: "PCP derived value of percentage of used inodes on a filesystem."
    applications:
    - Disk

  - discoveryrule_key: disc.disk
    name: "TPS (IOPS) for disk {#OSO_DISK}"
    key: "disc.disk.tps[{#OSO_DISK}]"
    value_type: int
    description: "PCP disk.dev.totals metric measured over a period of time.  This shows how many disk transactions per second the disk is using"
    applications:
    - Disk

  - discoveryrule_key: disc.disk
    name: "Percent Utilized for disk {#OSO_DISK}"
    key: "disc.disk.putil[{#OSO_DISK}]"
    value_type: float
    description: "PCP disk.dev.avactive metric measured over a period of time.  This is the '%util' in the iostat command"
    applications:
    - Disk

  - discoveryrule_key: disc.network
    name: "Bytes per second IN on network interface {#OSO_NET_INTERFACE}"
    key: "disc.network.in.bytes[{#OSO_NET_INTERFACE}]"
    value_type: int
    units: B
    delta: 1
    description: "PCP network.interface.in.bytes metric.  This is setup as a delta in Zabbix to measure the speed per second"
    applications:
    - Network

  - discoveryrule_key: disc.network
    name: "Bytes per second OUT on network interface {#OSO_NET_INTERFACE}"
    key: "disc.network.out.bytes[{#OSO_NET_INTERFACE}]"
    value_type: int
    units: B
    delta: 1
    description: "PCP network.interface.out.bytes metric.  This is setup as a delta in Zabbix to measure the speed per second"
    applications:
    - Network

  ztriggerprototypes:
  - name: 'Filesystem: {#OSO_FILESYS} has less than 10% free disk space on {HOST.NAME}'
    expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>90'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc'
    priority: high

  # This has a dependency on the previous trigger
  - name: 'Filesystem: {#OSO_FILESYS} has less than 15% free disk space on {HOST.NAME}'
    expression: '{Template OS Linux:disc.filesys.full[{#OSO_FILESYS}].last()}>85'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc'
    priority: warn
    dependencies:
    - 'Filesystem: {#OSO_FILESYS} has less than 10% free disk space on {HOST.NAME}'

  - name: 'Filesystem: {#OSO_FILESYS} has less than 5% free inodes on {HOST.NAME}'
    expression: '{Template OS Linux:disc.filesys.inodes.pused[{#OSO_FILESYS}].last()}>95'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc'
    priority: high

  - name: 'Filesystem: {#OSO_FILESYS} has less than 10% free inodes on {HOST.NAME}'
    expression: '{Template OS Linux:disc.filesys.inodes.pused[{#OSO_FILESYS}].last()}>90'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_filesys_full.asciidoc'
    priority: warn
    dependencies:
    - 'Filesystem: {#OSO_FILESYS} has less than 5% free inodes on {HOST.NAME}'

  ztriggers:
  - name: 'Too many TOTAL processes on {HOST.NAME}'
    expression: '{Template OS Linux:proc.nprocs.last()}>5000'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_proc.asciidoc'
    priority: warn

  - name: 'Lack of available memory on {HOST.NAME}'
    expression: '{Template OS Linux:mem.freemem.last()}<30720000'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_memory.asciidoc'
    priority: warn
    description: 'Alert on less than 30MegaBytes.  This is 30 Million Bytes.  30000 KB x 1024'

    #  CPU Utilization #
  - name: 'CPU idle less than 5% on {HOST.NAME}'
    expression: '{Template OS Linux:kernel.all.cpu.idle.max(#5)}<5'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_cpu_idle.asciidoc'
    priority: average
    description: 'CPU is less than 5% idle'

  - name: 'CPU idle less than 10% on {HOST.NAME}'
    expression: '{Template OS Linux:kernel.all.cpu.idle.max(#5)}<10'
    url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_cpu_idle.asciidoc'
    priority: average
    description: 'CPU is less than 10% idle'
    dependencies:
    - 'CPU idle less than 5% on {HOST.NAME}'