diff options
Diffstat (limited to 'health/health.d/vsphere.conf')
-rw-r--r-- | health/health.d/vsphere.conf | 263 |
1 files changed, 148 insertions, 115 deletions
diff --git a/health/health.d/vsphere.conf b/health/health.d/vsphere.conf index 3e1414c16..aee7c5cd4 100644 --- a/health/health.d/vsphere.conf +++ b/health/health.d/vsphere.conf @@ -4,138 +4,171 @@ # -----------------------------------------------VM Specific------------------------------------------------------------ # Memory -template: vsphere_vm_mem_usage - on: vsphere.vm_mem_usage_percentage - hosts: * - calc: $used - units: % - every: 20s - warn: $this > (($status >= $WARNING) ? (80) : (90)) - crit: $this > (($status == $CRITICAL) ? (90) : (98)) - delay: down 15m multiplier 1.5 max 1h - info: virtual machine memory utilization + template: vsphere_vm_mem_usage + on: vsphere.vm_mem_usage_percentage + class: Virtual Machine +component: Memory + type: Utilization + hosts: * + calc: $used + units: % + every: 20s + warn: $this > (($status >= $WARNING) ? (80) : (90)) + crit: $this > (($status == $CRITICAL) ? (90) : (98)) + delay: down 15m multiplier 1.5 max 1h + info: virtual machine memory utilization # -----------------------------------------------HOST Specific---------------------------------------------------------- # Memory -template: vsphere_host_mem_usage - on: vsphere.host_mem_usage_percentage - hosts: * - calc: $used - units: % - every: 20s - warn: $this > (($status >= $WARNING) ? (80) : (90)) - crit: $this > (($status == $CRITICAL) ? (90) : (98)) - delay: down 15m multiplier 1.5 max 1h - info: host memory utilization + template: vsphere_host_mem_usage + on: vsphere.host_mem_usage_percentage + class: Virtual Machine +component: Memory + type: Utilization + hosts: * + calc: $used + units: % + every: 20s + warn: $this > (($status >= $WARNING) ? (80) : (90)) + crit: $this > (($status == $CRITICAL) ? (90) : (98)) + delay: down 15m multiplier 1.5 max 1h + info: host memory utilization # Network errors -template: vsphere_inbound_packets_errors - on: vsphere.net_errors_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of rx - units: packets - every: 1m - info: number of inbound errors for the network interface in the last 10 minutes - -template: vsphere_outbound_packets_errors - on: vsphere.net_errors_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of tx - units: packets - every: 1m - info: number of outbound errors for the network interface in the last 10 minutes + template: vsphere_inbound_packets_errors + on: vsphere.net_errors_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of rx + units: packets + every: 1m + info: number of inbound errors for the network interface in the last 10 minutes + + template: vsphere_outbound_packets_errors + on: vsphere.net_errors_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of tx + units: packets + every: 1m + info: number of outbound errors for the network interface in the last 10 minutes # Network errors ratio -template: vsphere_inbound_packets_errors_ratio - on: vsphere.net_packets_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_errors != nan AND $this > 1000) ? ($vsphere_inbound_packets_errors * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of inbound errors for the network interface over the last 10 minutes - to: sysadmin - -template: vsphere_outbound_packets_errors_ratio - on: vsphere.net_packets_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_errors != nan AND $this > 1000) ? ($vsphere_outbound_packets_errors * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of outbound errors for the network interface over the last 10 minutes - to: sysadmin + template: vsphere_inbound_packets_errors_ratio + on: vsphere.net_packets_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of rx + calc: (($vsphere_inbound_packets_errors != nan AND $this > 1000) ? ($vsphere_inbound_packets_errors * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + info: ratio of inbound errors for the network interface over the last 10 minutes + to: sysadmin + + template: vsphere_outbound_packets_errors_ratio + on: vsphere.net_packets_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of tx + calc: (($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 - hosts: * - lookup: average -10m unaligned match-names of used - units: % - every: 20s - warn: $this > (($status >= $WARNING) ? (75) : (85)) - crit: $this > (($status == $CRITICAL) ? (85) : (95)) - delay: down 15m multiplier 1.5 max 1h - info: average CPU utilization - to: sysadmin + template: vsphere_cpu_usage + on: vsphere.cpu_usage_total + class: Virtual Machine +component: CPU + type: Utilization + hosts: * + lookup: average -10m unaligned match-names of used + units: % + every: 20s + warn: $this > (($status >= $WARNING) ? (75) : (85)) + crit: $this > (($status == $CRITICAL) ? (85) : (95)) + delay: down 15m multiplier 1.5 max 1h + info: average CPU utilization + to: sysadmin # Network drops -template: vsphere_inbound_packets_dropped - on: vsphere.net_drops_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of rx - units: packets - every: 1m - info: number of inbound dropped packets for the network interface in the last 10 minutes - -template: vsphere_outbound_packets_dropped - on: vsphere.net_drops_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of tx - units: packets - every: 1m - info: number of outbound dropped packets for the network interface in the last 10 minutes + template: vsphere_inbound_packets_dropped + on: vsphere.net_drops_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of rx + units: packets + every: 1m + info: number of inbound dropped packets for the network interface in the last 10 minutes + + template: vsphere_outbound_packets_dropped + on: vsphere.net_drops_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of tx + units: packets + every: 1m + info: number of outbound dropped packets for the network interface in the last 10 minutes # Network drops ratio -template: vsphere_inbound_packets_dropped_ratio - on: vsphere.net_packets_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_inbound_packets_dropped * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of inbound dropped packets for the network interface over the last 10 minutes - to: sysadmin - -template: vsphere_outbound_packets_dropped_ratio - on: vsphere.net_packets_total - hosts: * -families: * - lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_outbound_packets_dropped * 100 / $this) : (0)) - units: % - every: 1m - warn: $this >= 2 - delay: up 1m down 1h multiplier 1.5 max 2h - info: ratio of outbound dropped packets for the network interface over the last 10 minutes - to: sysadmin + template: vsphere_inbound_packets_dropped_ratio + on: vsphere.net_packets_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of rx + calc: (($vsphere_inbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_inbound_packets_dropped * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + info: ratio of inbound dropped packets for the network interface over the last 10 minutes + to: sysadmin + + template: vsphere_outbound_packets_dropped_ratio + on: vsphere.net_packets_total + class: Virtual Machine +component: Network + type: Errors + hosts: * + families: * + lookup: sum -10m unaligned absolute match-names of tx + calc: (($vsphere_outbound_packets_dropped != nan AND $this > 1000) ? ($vsphere_outbound_packets_dropped * 100 / $this) : (0)) + 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 + to: sysadmin |