From 517a443636daa1e8085cb4e5325524a54e8a8fd7 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Tue, 17 Oct 2023 11:30:23 +0200 Subject: Merging upstream version 1.43.0. Signed-off-by: Daniel Baumann --- health/health.d/vsphere.conf | 158 +++++++++---------------------------------- 1 file changed, 31 insertions(+), 127 deletions(-) (limited to 'health/health.d/vsphere.conf') diff --git a/health/health.d/vsphere.conf b/health/health.d/vsphere.conf index 1d8be6cb5..b8ad9aee4 100644 --- a/health/health.d/vsphere.conf +++ b/health/health.d/vsphere.conf @@ -1,28 +1,26 @@ # you can disable an alarm notification by setting the 'to' line to: silent -# -----------------------------------------------VM Specific------------------------------------------------------------ -# Memory +# -----------------------------------------------Virtual Machine-------------------------------------------------------- - template: vsphere_vm_mem_usage - on: vsphere.vm_mem_usage_percentage + template: vsphere_vm_cpu_utilization + on: vsphere.vm_cpu_utilization class: Utilization type: Virtual Machine -component: Memory +component: CPU hosts: * - calc: $used + lookup: average -10m unaligned match-names of used units: % every: 20s - warn: $this > (($status >= $WARNING) ? (80) : (90)) - crit: $this > (($status == $CRITICAL) ? (90) : (98)) + warn: $this > (($status >= $WARNING) ? (75) : (85)) + crit: $this > (($status == $CRITICAL) ? (85) : (95)) delay: down 15m multiplier 1.5 max 1h - info: virtual machine memory utilization - -# -----------------------------------------------HOST Specific---------------------------------------------------------- -# Memory + summary: vSphere CPU utilization for VM ${label:vm} + info: CPU utilization VM ${label:vm} host ${label:host} cluster ${label:cluster} datacenter ${label:datacenter} + to: silent - template: vsphere_host_mem_usage - on: vsphere.host_mem_usage_percentage + template: vsphere_vm_mem_utilization + on: vsphere.vm_mem_utilization class: Utilization type: Virtual Machine component: Memory @@ -33,69 +31,14 @@ component: Memory warn: $this > (($status >= $WARNING) ? (80) : (90)) crit: $this > (($status == $CRITICAL) ? (90) : (98)) delay: down 15m multiplier 1.5 max 1h - info: host memory utilization - -# Network errors - - template: vsphere_inbound_packets_errors - on: vsphere.net_errors_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of rx - units: packets - every: 1m - info: number of inbound errors for the network interface in the last 10 minutes - - template: vsphere_outbound_packets_errors - on: vsphere.net_errors_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of tx - units: packets - every: 1m - info: number of outbound errors for the network interface in the last 10 minutes - -# Network errors ratio + summary: vSphere memory utilization for VM ${label:vm} + info: Memory utilization VM ${label:vm} host ${label:host} cluster ${label:cluster} datacenter ${label:datacenter} + to: silent - template: vsphere_inbound_packets_errors_ratio - on: vsphere.net_packets_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_errors != nan AND $this > 1000) ? ($vsphere_inbound_packets_errors * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of inbound errors for the network interface over the last 10 minutes - to: sysadmin +# -----------------------------------------------ESXI host-------------------------------------------------------------- - template: vsphere_outbound_packets_errors_ratio - on: vsphere.net_packets_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_errors != nan AND $this > 1000) ? ($vsphere_outbound_packets_errors * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of outbound errors for the network interface over the last 10 minutes - to: sysadmin - -# -----------------------------------------------Common------------------------------------------------------------------- -# CPU - - template: vsphere_cpu_usage - on: vsphere.cpu_usage_total + template: vsphere_host_cpu_utilization + on: vsphere.host_cpu_utilization class: Utilization type: Virtual Machine component: CPU @@ -106,61 +49,22 @@ component: CPU warn: $this > (($status >= $WARNING) ? (75) : (85)) crit: $this > (($status == $CRITICAL) ? (85) : (95)) delay: down 15m multiplier 1.5 max 1h - info: average CPU utilization + summary: vSphere ESXi CPU utilization for host ${label:host} + info: CPU utilization ESXi host ${label:host} cluster ${label:cluster} datacenter ${label:datacenter} to: sysadmin -# Network drops - - template: vsphere_inbound_packets_dropped - on: vsphere.net_drops_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of rx - units: packets - every: 1m - info: number of inbound dropped packets for the network interface in the last 10 minutes - - template: vsphere_outbound_packets_dropped - on: vsphere.net_drops_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of tx - units: packets - every: 1m - info: number of outbound dropped packets for the network interface in the last 10 minutes - -# Network drops ratio - - template: vsphere_inbound_packets_dropped_ratio - on: vsphere.net_packets_total - class: Errors - type: Virtual Machine -component: Network - hosts: * - lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_inbound_packets_dropped * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of inbound dropped packets for the network interface over the last 10 minutes - to: sysadmin - - template: vsphere_outbound_packets_dropped_ratio - on: vsphere.net_packets_total - class: Errors + template: vsphere_host_mem_utilization + on: vsphere.host_mem_utilization + class: Utilization type: Virtual Machine -component: Network +component: Memory hosts: * - lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_outbound_packets_dropped * 100 / $this) : (0)) + calc: $used units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of outbound dropped packets for the network interface over the last 10 minutes + every: 20s + warn: $this > (($status >= $WARNING) ? (80) : (90)) + crit: $this > (($status == $CRITICAL) ? (90) : (98)) + delay: down 15m multiplier 1.5 max 1h + summary: vSphere ESXi Ram utilization for host ${label:host} + info: Memory utilization ESXi host ${label:host} cluster ${label:cluster} datacenter ${label:datacenter} to: sysadmin -- cgit v1.2.3