diff options
Diffstat (limited to 'roles/os_zabbix/vars/template_os_linux.yml')
| -rw-r--r-- | roles/os_zabbix/vars/template_os_linux.yml | 199 |
1 files changed, 143 insertions, 56 deletions
diff --git a/roles/os_zabbix/vars/template_os_linux.yml b/roles/os_zabbix/vars/template_os_linux.yml index 1c9d10bb0..69432273f 100644 --- a/roles/os_zabbix/vars/template_os_linux.yml +++ b/roles/os_zabbix/vars/template_os_linux.yml @@ -10,17 +10,20 @@ g_template_os_linux: - key: kernel.all.cpu.wait.total applications: - Kernel - value_type: int + value_type: float + units: '%' - key: kernel.all.cpu.irq.hard applications: - Kernel - value_type: int + value_type: float + units: '%' - key: kernel.all.cpu.idle applications: - Kernel - value_type: int + value_type: float + units: '%' - key: kernel.uname.distro applications: @@ -35,7 +38,8 @@ g_template_os_linux: - key: kernel.all.cpu.irq.soft applications: - Kernel - value_type: int + value_type: float + units: '%' - key: kernel.all.load.15_minute applications: @@ -45,32 +49,19 @@ g_template_os_linux: - key: kernel.all.cpu.sys applications: - Kernel - value_type: int + value_type: float + units: '%' - key: kernel.all.load.5_minute applications: - Kernel value_type: float - - key: mem.freemem - applications: - - Memory - value_type: int - - key: kernel.all.cpu.nice applications: - Kernel - value_type: int - - - key: mem.util.bufmem - applications: - - Memory - value_type: int - - - key: swap.used - applications: - - Memory - value_type: int + value_type: float + units: '%' - key: kernel.all.load.1_minute applications: @@ -82,92 +73,188 @@ g_template_os_linux: - Kernel value_type: string - - key: swap.length + - key: kernel.all.uptime applications: - - Memory + - Kernel value_type: int - - key: mem.physmem + - key: kernel.all.cpu.user applications: - - Memory - value_type: int + - Kernel + value_type: float + units: '%' - - key: kernel.all.uptime + - key: kernel.uname.machine applications: - Kernel - value_type: int + value_type: string - - key: swap.free + - key: hinv.ncpu applications: - - Memory + - Kernel value_type: int - - key: mem.util.used + - key: kernel.all.cpu.steal applications: - - Memory - value_type: int + - Kernel + value_type: float + units: '%' - - key: kernel.all.cpu.user + - key: kernel.all.pswitch applications: - Kernel value_type: int - - key: kernel.uname.machine + - key: kernel.uname.release applications: - Kernel value_type: string - - key: hinv.ncpu + - key: proc.nprocs applications: - Kernel value_type: int - - key: mem.util.cached + # Memory Items + - key: mem.freemem applications: - Memory value_type: int + description: "PCP: free system memory metric from /proc/meminfo" + multiplier: 1024 + units: B - - key: kernel.all.cpu.steal + - key: mem.util.bufmem applications: - - Kernel + - Memory value_type: int + description: "PCP: Memory allocated for buffer_heads.; I/O buffers metric from /proc/meminfo" + multiplier: 1024 + units: B - - key: kernel.all.pswitch + - key: swap.used applications: - - Kernel + - Memory value_type: int + description: "PCP: swap used metric from /proc/meminfo" + multiplier: 1024 + units: B - - key: kernel.uname.release + - key: swap.length applications: - - Kernel - value_type: string + - Memory + value_type: int + description: "PCP: total swap available metric from /proc/meminfo" + multiplier: 1024 + units: B - - key: proc.nprocs + - key: mem.physmem applications: - - Kernel + - 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: filesys.avail + - key: swap.free applications: - - Disk + - Memory value_type: int + description: "PCP: swap free metric from /proc/meminfo" + multiplier: 1024 + units: B - - key: filesys.capacity + - key: mem.util.available applications: - - Disk + - 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: filesys.free + - key: mem.util.used applications: - - Disk + - 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: filesys.full + - key: mem.util.cached applications: - - Disk + - 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" + + zitemprototypes: + - discoveryrule_key: disc.filesys + name: "disc.filesys.full.{#OSO_FILESYS}" + key: "disc.filesys.full[{#OSO_FILESYS}]" value_type: float - - - key: filesys.used + 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 + + ztriggerprototypes: + - 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 + + - 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 + + - 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 + + - 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 + + 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.last()}<5 and {Template OS Linux:kernel.all.cpu.idle.last(#2)}<5' + url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_cpu_idle.asciidoc' + priority: high + description: 'CPU is less than 5% idle' + + - name: 'CPU idle less than 10% on {HOST.NAME}' + expression: '{Template OS Linux:kernel.all.cpu.idle.last()}<10 and {Template OS Linux:kernel.all.cpu.idle.last(#2)}<10' + url: 'https://github.com/openshift/ops-sop/blob/master/V3/Alerts/check_cpu_idle.asciidoc' + priority: warn + description: 'CPU is less than 10% idle' + dependencies: + - 'CPU idle less than 5% on {HOST.NAME}' |