diff options
Diffstat (limited to 'health/health.d')
| -rw-r--r-- | health/health.d/vsphere.conf | 157 |
1 files changed, 157 insertions, 0 deletions
diff --git a/health/health.d/vsphere.conf b/health/health.d/vsphere.conf new file mode 100644 index 000000000..d8b2be190 --- /dev/null +++ b/health/health.d/vsphere.conf @@ -0,0 +1,157 @@ + +# you can disable an alarm notification by setting the 'to' line to: silent + +# -----------------------------------------------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: used RAM + +# -----------------------------------------------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: used RAM + +# 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 + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + info: interface inbound dropped packets in the last 10 minutes + to: sysadmin + +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 + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + info: interface outbound dropped packets in the last 10 minutes + to: sysadmin + +# Network errors ratio + +template: vsphere_inbound_packets_errors_ratio + on: vsphere.net_packets_total + hosts: * +families: * + lookup: sum -10m unaligned absolute match-names of rx + calc: (($vsphere_inbound_packets_errors != nan AND $this > 0) ? ($vsphere_inbound_packets_errors * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 0.1 + crit: $this >= 2 + delay: down 1h multiplier 1.5 max 2h + info: the ratio of inbound errors vs the total number of received packets of the network interface, during the last 10 minutes + to: sysadmin + +template: vsphere_outbound_packets_errors_ratio + on: vsphere.net_packets_total + hosts: * +families: * + lookup: sum -10m unaligned absolute match-names of tx + calc: (($vsphere_outbound_packets_errors != nan AND $this > 0) ? ($vsphere_outbound_packets_errors * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 0.1 + crit: $this >= 2 + delay: down 1h multiplier 1.5 max 2h + info: the ratio of outbound errors vs the total number of sent packets of the network interface, during 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: cpu utilization for the last 10 minutes + 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 + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + info: interface inbound dropped packets in the last 10 minutes + to: sysadmin + +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 + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + info: interface outbound dropped packets in the last 10 minutes + to: sysadmin + +# Network drops ratio + +template: vsphere_inbound_packets_dropped_ratio + on: vsphere.net_packets_total + hosts: * +families: * + lookup: sum -10m unaligned absolute match-names of rx + calc: (($vsphere_inbound_packets_dropped != nan AND $this > 0) ? ($vsphere_inbound_packets_dropped * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 0.1 + crit: $this >= 2 + delay: down 1h multiplier 1.5 max 2h + info: the ratio of inbound dropped packets vs the total number of received packets of the network interface, during the last 10 minutes + to: sysadmin + +template: vsphere_outbound_packets_dropped_ratio + on: vsphere.net_packets_total + hosts: * +families: * + lookup: sum -10m unaligned absolute match-names of tx + calc: (($vsphere_outbound_packets_dropped != nan AND $this > 0) ? ($vsphere_outbound_packets_dropped * 100 / $this) : (0)) + units: % + every: 1m + warn: $this >= 0.1 + crit: $this >= 2 + delay: down 1h multiplier 1.5 max 2h + info: the ratio of outbound dropped packets vs the total number of sent packets of the network interface, during the last 10 minutes + to: sysadmin |