diff options
Diffstat (limited to 'health/health.d/vsphere.conf')
| -rw-r--r-- | health/health.d/vsphere.conf | 62 |
1 files changed, 23 insertions, 39 deletions
diff --git a/health/health.d/vsphere.conf b/health/health.d/vsphere.conf index d8b2be19..3e1414c1 100644 --- a/health/health.d/vsphere.conf +++ b/health/health.d/vsphere.conf @@ -13,7 +13,7 @@ template: vsphere_vm_mem_usage warn: $this > (($status >= $WARNING) ? (80) : (90)) crit: $this > (($status == $CRITICAL) ? (90) : (98)) delay: down 15m multiplier 1.5 max 1h - info: used RAM + info: virtual machine memory utilization # -----------------------------------------------HOST Specific---------------------------------------------------------- # Memory @@ -27,7 +27,7 @@ template: vsphere_host_mem_usage warn: $this > (($status >= $WARNING) ? (80) : (90)) crit: $this > (($status == $CRITICAL) ? (90) : (98)) delay: down 15m multiplier 1.5 max 1h - info: used RAM + info: host memory utilization # Network errors @@ -38,10 +38,7 @@ 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 + info: number of inbound errors for the network interface in the last 10 minutes template: vsphere_outbound_packets_errors on: vsphere.net_errors_total @@ -50,10 +47,7 @@ families: * lookup: sum -10m unaligned absolute match-names of tx units: packets every: 1m - warn: $this >= 5 - delay: down 1h multiplier 1.5 max 2h - info: interface outbound dropped packets in the last 10 minutes - to: sysadmin + info: number of outbound errors for the network interface in the last 10 minutes # Network errors ratio @@ -62,13 +56,12 @@ template: vsphere_inbound_packets_errors_ratio hosts: * families: * lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_errors != nan AND $this > 0) ? ($vsphere_inbound_packets_errors * 100 / $this) : (0)) + calc: (($vsphere_inbound_packets_errors != nan AND $this > 1000) ? ($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 + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + info: ratio of inbound errors for the network interface over the last 10 minutes to: sysadmin template: vsphere_outbound_packets_errors_ratio @@ -76,13 +69,12 @@ template: vsphere_outbound_packets_errors_ratio hosts: * families: * lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_errors != nan AND $this > 0) ? ($vsphere_outbound_packets_errors * 100 / $this) : (0)) + calc: (($vsphere_outbound_packets_errors != nan AND $this > 1000) ? ($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 + 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------------------------------------------------------------------- @@ -97,7 +89,7 @@ template: vsphere_cpu_usage 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 + info: average CPU utilization to: sysadmin # Network drops @@ -109,10 +101,7 @@ 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 + 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 @@ -121,10 +110,7 @@ families: * lookup: sum -10m unaligned absolute match-names of tx units: packets every: 1m - warn: $this >= 5 - delay: down 1h multiplier 1.5 max 2h - info: interface outbound dropped packets in the last 10 minutes - to: sysadmin + info: number of outbound dropped packets for the network interface in the last 10 minutes # Network drops ratio @@ -133,13 +119,12 @@ template: vsphere_inbound_packets_dropped_ratio hosts: * families: * lookup: sum -10m unaligned absolute match-names of rx - calc: (($vsphere_inbound_packets_dropped != nan AND $this > 0) ? ($vsphere_inbound_packets_dropped * 100 / $this) : (0)) + calc: (($vsphere_inbound_packets_dropped != nan AND $this > 1000) ? ($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 + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + info: ratio of inbound dropped packets for the network interface over the last 10 minutes to: sysadmin template: vsphere_outbound_packets_dropped_ratio @@ -147,11 +132,10 @@ template: vsphere_outbound_packets_dropped_ratio hosts: * families: * lookup: sum -10m unaligned absolute match-names of tx - calc: (($vsphere_outbound_packets_dropped != nan AND $this > 0) ? ($vsphere_outbound_packets_dropped * 100 / $this) : (0)) + calc: (($vsphere_outbound_packets_dropped != nan AND $this > 1000) ? ($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 + 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 |