Adding upstream version 1.37.1.upstream/1.37.1upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 256 insertions, 0 deletions

diff --git a/health/health.d/net.conf b/health/health.d/net.conf
new file mode 100644
index 0000000..9d5b3b8
--- /dev/null
+++ b/health/health.d/net.conf
@@ -0,0 +1,256 @@
+
+# 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: *
+ families: *
+     calc: ( $nic_speed_max > 0 ) ? ( $nic_speed_max) : ( nan )
+    units: Mbit
+    every: 10s
+     info: network interface $family current speed
+
+ template: 1m_received_traffic_overflow
+       on: net.net
+    class: Workload
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: *
+   lookup: average -1m unaligned absolute of received
+     calc: ($interface_speed > 0) ? ($this * 100 / ($interface_speed)) : ( nan )
+    units: %
+    every: 10s
+     warn: $this > (($status >= $WARNING)  ? (85) : (90))
+    delay: up 1m down 1m multiplier 1.5 max 1h
+     info: average inbound utilization for the network interface $family over the last minute
+       to: sysadmin
+
+ template: 1m_sent_traffic_overflow
+       on: net.net
+    class: Workload
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: *
+   lookup: average -1m unaligned absolute of sent
+     calc: ($interface_speed > 0) ? ($this * 100 / ($interface_speed)) : ( nan )
+    units: %
+    every: 10s
+     warn: $this > (($status >= $WARNING)  ? (85) : (90))
+    delay: up 1m down 1m multiplier 1.5 max 1h
+     info: average outbound utilization for the network interface $family over the last minute
+       to: sysadmin
+
+# -----------------------------------------------------------------------------
+# 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: inbound_packets_dropped
+       on: net.drops
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: *
+   lookup: sum -10m unaligned absolute of inbound
+    units: packets
+    every: 1m
+     info: number of inbound dropped packets for the network interface $family in the last 10 minutes
+
+ template: outbound_packets_dropped
+       on: net.drops
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: *
+   lookup: sum -10m unaligned absolute of outbound
+    units: packets
+    every: 1m
+     info: number of outbound dropped packets for the network interface $family in the last 10 minutes
+
+ template: inbound_packets_dropped_ratio
+       on: net.packets
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: !wl* *
+   lookup: sum -10m unaligned absolute of received
+     calc: (($inbound_packets_dropped != nan AND $this > 10000) ? ($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 $family over the last 10 minutes
+       to: sysadmin
+
+ template: outbound_packets_dropped_ratio
+       on: net.packets
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: !wl* *
+   lookup: sum -10m unaligned absolute of sent
+     calc: (($outbound_packets_dropped != nan AND $this > 1000) ? ($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 $family over the last 10 minutes
+       to: sysadmin
+
+ template: wifi_inbound_packets_dropped_ratio
+       on: net.packets
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: wl*
+   lookup: sum -10m unaligned absolute of received
+     calc: (($inbound_packets_dropped != nan AND $this > 10000) ? ($inbound_packets_dropped * 100 / $this) : (0))
+    units: %
+    every: 1m
+     warn: $this >= 10
+    delay: up 1m down 1h multiplier 1.5 max 2h
+     info: ratio of inbound dropped packets for the network interface $family over the last 10 minutes
+       to: sysadmin
+
+ template: wifi_outbound_packets_dropped_ratio
+       on: net.packets
+    class: Errors
+     type: System
+component: Network
+       os: linux
+    hosts: *
+ families: wl*
+   lookup: sum -10m unaligned absolute of sent
+     calc: (($outbound_packets_dropped != nan AND $this > 1000) ? ($outbound_packets_dropped * 100 / $this) : (0))
+    units: %
+    every: 1m
+     warn: $this >= 10
+    delay: up 1m down 1h multiplier 1.5 max 2h
+     info: ratio of outbound dropped packets for the network interface $family over the last 10 minutes
+       to: sysadmin
+
+# -----------------------------------------------------------------------------
+# interface errors
+
+ template: interface_inbound_errors
+       on: net.errors
+    class: Errors
+     type: System
+component: Network
+       os: freebsd
+    hosts: *
+ families: *
+   lookup: sum -10m unaligned absolute of inbound
+    units: errors
+    every: 1m
+     warn: $this >= 5
+    delay: down 1h multiplier 1.5 max 2h
+     info: number of inbound errors for the network interface $family in the last 10 minutes
+       to: sysadmin
+
+ template: interface_outbound_errors
+       on: net.errors
+    class: Errors
+     type: System
+component: Network
+       os: freebsd
+    hosts: *
+ families: *
+   lookup: sum -10m unaligned absolute of outbound
+    units: errors
+    every: 1m
+     warn: $this >= 5
+    delay: down 1h multiplier 1.5 max 2h
+     info: number of outbound errors for the network interface $family in the last 10 minutes
+       to: sysadmin
+
+# -----------------------------------------------------------------------------
+# 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: *
+ families: *
+   lookup: sum -10m unaligned absolute
+    units: errors
+    every: 1m
+     warn: $this > 0
+    delay: down 1h multiplier 1.5 max 2h
+     info: number of FIFO errors for the network interface $family in the last 10 minutes
+       to: sysadmin
+
+# -----------------------------------------------------------------------------
+# 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: *
+ families: *
+   lookup: average -1m unaligned of received
+    units: packets
+    every: 10s
+     info: average number of packets received by the network interface $family over the last minute
+
+ template: 10s_received_packets_storm
+       on: net.packets
+    class: Workload
+     type: System
+component: Network
+       os: linux freebsd
+    hosts: *
+ families: *
+   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
+     info: ratio of average number of received packets for the network interface $family over the last 10 seconds, \
+           compared to the rate over the last minute
+       to: sysadmin