diff options
Diffstat (limited to '')
| -rw-r--r-- | health/health.d/net.conf | 258 |
1 files changed, 258 insertions, 0 deletions
diff --git a/health/health.d/net.conf b/health/health.d/net.conf new file mode 100644 index 00000000..2dfe6bba --- /dev/null +++ b/health/health.d/net.conf @@ -0,0 +1,258 @@ + +# you can disable an alarm notification by setting the 'to' line to: silent + +# ----------------------------------------------------------------------------- +# net traffic overflow + + template: interface_speed + on: net.net + class: Latency + type: System +component: Network + os: * + hosts: * + calc: ( $nic_speed_max > 0 ) ? ( $nic_speed_max / 1000) : ( nan ) + units: Mbit + every: 10s + info: Network interface ${label:device} current speed + + template: 1m_received_traffic_overflow + on: net.net + class: Workload + type: System +component: Network + os: linux + hosts: * + lookup: average -1m unaligned absolute of received + calc: ($interface_speed > 0) ? ($this * 100 / ($interface_speed * 1000)) : ( nan ) + units: % + every: 10s + warn: $this > (($status >= $WARNING) ? (85) : (90)) + delay: up 1m down 1m multiplier 1.5 max 1h + summary: System network interface ${label:device} inbound utilization + info: Average inbound utilization for the network interface ${label:device} over the last minute + to: silent + + template: 1m_sent_traffic_overflow + on: net.net + class: Workload + type: System +component: Network + os: linux + hosts: * + lookup: average -1m unaligned absolute of sent + calc: ($interface_speed > 0) ? ($this * 100 / ($interface_speed * 1000)) : ( nan ) + units: % + every: 10s + warn: $this > (($status >= $WARNING) ? (85) : (90)) + delay: up 1m down 1m multiplier 1.5 max 1h + summary: System network interface ${label:device} outbound utilization + info: Average outbound utilization for the network interface ${label:device} over the last minute + to: silent + +# ----------------------------------------------------------------------------- +# dropped packets + +# check if an interface is dropping packets +# the alarm is checked every 1 minute +# and examines the last 10 minutes of data +# +# it is possible to have expected packet drops on an interface for some network configurations +# look at the Monitoring Network Interfaces section in the proc.plugin documentation for more information + + template: net_interface_inbound_packets + on: net.packets + class: Workload + type: System +component: Network + os: * + hosts: * + lookup: sum -10m unaligned absolute of received + units: packets + every: 1m + summary: Network interface ${label:device} received packets + info: Received packets for the network interface ${label:device} in the last 10 minutes + + template: net_interface_outbound_packets + on: net.packets + class: Workload + type: System +component: Network + os: * + hosts: * + lookup: sum -10m unaligned absolute of sent + units: packets + every: 1m + summary: Network interface ${label:device} sent packets + info: Sent packets for the network interface ${label:device} in the last 10 minutes + + template: inbound_packets_dropped_ratio + on: net.drops + class: Errors + type: System +component: Network + os: * + hosts: * +chart labels: device=!wl* * + lookup: sum -10m unaligned absolute of inbound + calc: (($net_interface_inbound_packets > 10000) ? ($this * 100 / $net_interface_inbound_packets) : (0)) + units: % + every: 1m + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} inbound drops + info: Ratio of inbound dropped packets for the network interface ${label:device} over the last 10 minutes + to: silent + + template: outbound_packets_dropped_ratio + on: net.drops + class: Errors + type: System +component: Network + os: * + hosts: * +chart labels: device=!wl* * + lookup: sum -10m unaligned absolute of outbound + calc: (($net_interface_outbound_packets > 1000) ? ($this * 100 / $net_interface_outbound_packets) : (0)) + units: % + every: 1m + warn: $this >= 2 + delay: up 1m down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} outbound drops + info: Ratio of outbound dropped packets for the network interface ${label:device} over the last 10 minutes + to: silent + + template: wifi_inbound_packets_dropped_ratio + on: net.drops + class: Errors + type: System +component: Network + os: linux + hosts: * +chart labels: device=wl* + lookup: sum -10m unaligned absolute of received + calc: (($net_interface_inbound_packets > 10000) ? ($this * 100 / $net_interface_inbound_packets) : (0)) + units: % + every: 1m + warn: $this >= 10 + delay: up 1m down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} inbound drops ratio + info: Ratio of inbound dropped packets for the network interface ${label:device} over the last 10 minutes + to: silent + + template: wifi_outbound_packets_dropped_ratio + on: net.drops + class: Errors + type: System +component: Network + os: linux + hosts: * +chart labels: device=wl* + lookup: sum -10m unaligned absolute of sent + calc: (($net_interface_outbound_packets > 1000) ? ($this * 100 / $net_interface_outbound_packets) : (0)) + units: % + every: 1m + warn: $this >= 10 + delay: up 1m down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} outbound drops ratio + info: Ratio of outbound dropped packets for the network interface ${label:device} over the last 10 minutes + to: silent + +# ----------------------------------------------------------------------------- +# interface errors + + template: interface_inbound_errors + on: net.errors + class: Errors + type: System +component: Network + os: freebsd + hosts: * + lookup: sum -10m unaligned absolute of inbound + units: errors + every: 1m + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} inbound errors + info: Number of inbound errors for the network interface ${label:device} in the last 10 minutes + to: silent + + template: interface_outbound_errors + on: net.errors + class: Errors + type: System +component: Network + os: freebsd + hosts: * + lookup: sum -10m unaligned absolute of outbound + units: errors + every: 1m + warn: $this >= 5 + delay: down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} outbound errors + info: Number of outbound errors for the network interface ${label:device} in the last 10 minutes + to: silent + +# ----------------------------------------------------------------------------- +# FIFO errors + +# check if an interface is having FIFO +# buffer errors +# the alarm is checked every 1 minute +# and examines the last 10 minutes of data + + template: 10min_fifo_errors + on: net.fifo + class: Errors + type: System +component: Network + os: linux + hosts: * + lookup: sum -10m unaligned absolute + units: errors + every: 1m + warn: $this > 0 + delay: down 1h multiplier 1.5 max 2h + summary: System network interface ${label:device} FIFO errors + info: Number of FIFO errors for the network interface ${label:device} in the last 10 minutes + to: silent + +# ----------------------------------------------------------------------------- +# check for packet storms + +# 1. calculate the rate packets are received in 1m: 1m_received_packets_rate +# 2. do the same for the last 10s +# 3. raise an alarm if the later is 10x or 20x the first +# we assume the minimum packet storm should at least have +# 10000 packets/s, average of the last 10 seconds + + template: 1m_received_packets_rate + on: net.packets + class: Workload + type: System +component: Network + os: linux freebsd + hosts: * + lookup: average -1m unaligned of received + units: packets + every: 10s + info: Average number of packets received by the network interface ${label:device} over the last minute + + template: 10s_received_packets_storm + on: net.packets + class: Workload + type: System +component: Network + os: linux freebsd + hosts: * + lookup: average -10s unaligned of received + calc: $this * 100 / (($1m_received_packets_rate < 1000)?(1000):($1m_received_packets_rate)) + every: 10s + units: % + warn: $this > (($status >= $WARNING)?(200):(5000)) + crit: $this > (($status == $CRITICAL)?(5000):(6000)) + options: no-clear-notification + summary: System network interface ${label:device} inbound packet storm + info: Ratio of average number of received packets for the network interface ${label:device} over the last 10 seconds, \ + compared to the rate over the last minute + to: silent |