self-cleaning obsolete cgroups and network interfaces from memory; fixes #1163; fixes...

[netdata.git] / conf.d / health.d / net.conf

diff --git a/conf.d/health.d/net.conf b/conf.d/health.d/net.conf
index f969f17b9c47cbe0e2f109392e9656d6311ebfad..0232395acfd2cd1e5c313a72a4ecd1e6997f43b4 100644 (file)
--- a/conf.d/health.d/net.conf
+++ b/conf.d/health.d/net.conf
@@ -1,67 +1,57 @@
-# -----------------------------------------------------------------------------
-# make sure we collect values for each interface
-
-template: interface_last_collected_secs
-      on: net.net
-    calc: $now - $last_collected_t
-   units: seconds ago
-   every: 10s
-    warn: $this > (($status >= $WARNING)  ? ($update_every) : ( 5 * $update_every))
-    crit: $this > (($status == $CRITICAL) ? ($update_every) : (60 * $update_every))
-   delay: down 5m multiplier 1.5 max 1h
-    info: number of seconds since the last successful data collection
-      to: sysadmin
-
 
 # -----------------------------------------------------------------------------
 # dropped packets
 
 # check if an interface is dropping packets
 # the alarm is checked every 1 minute
-# and examines the last hour of data
+# and examines the last 10 minutes of data
 
-template: 1hour_packet_drops_inbound
+template: inbound_packets_dropped
       on: net.drops
-  lookup: sum -1h unaligned absolute of inbound
+families: *
+  lookup: sum -10m unaligned absolute of inbound
    units: packets
    every: 1m
     warn: $this > 0
-   delay: down 30m multiplier 1.5 max 1h
-    info: interface inbound dropped packets in the last hour
+   delay: down 1h multiplier 1.5 max 2h
+    info: interface inbound dropped packets in the last 10 minutes
       to: sysadmin
 
-template: 1hour_packet_drops_outbound
+template: outbound_packets_dropped
       on: net.drops
-  lookup: sum -1h unaligned absolute of outbound
+families: *
+  lookup: sum -10m unaligned absolute of outbound
    units: packets
    every: 1m
     warn: $this > 0
-   delay: down 30m multiplier 1.5 max 1h
-    info: interface outbound dropped packets in the last hour
+   delay: down 1h multiplier 1.5 max 2h
+    info: interface outbound dropped packets in the last 10 minutes
       to: sysadmin
 
-template: 1hour_packet_drops_ratio_inbound
+template: inbound_packets_dropped_ratio
       on: net.packets
-  lookup: sum -1h unaligned absolute of received
-    calc: (($1hour_packet_drops_inbound != nan AND $this > 0) ? ($1hour_packet_drops_inbound * 100 / $this) : (0))
+families: *
+  lookup: sum -10m unaligned absolute of received
+    calc: (($inbound_packets_dropped != nan AND $this > 0) ? ($inbound_packets_dropped * 100 / $this) : (0))
    units: %
    every: 1m
     warn: $this > 0.5
     crit: $this > 3
-   delay: down 30m multiplier 1.5 max 1h
-    info: the ratio of inbound dropped packets vs the total number of received packets of the network interface, during the last hour
+   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: 1hour_packet_drops_ratio_outbound
+template: outbound_packets_dropped_ratio
       on: net.packets
-  lookup: sum -1h unaligned absolute of sent
-    calc: (($1hour_packet_drops_outbound != nan AND $this > 0) ? ($1hour_packet_drops_outbound * 100 / $this) : (0))
+families: *
+  lookup: sum -10m unaligned absolute of sent
+    calc: (($outbound_packets_dropped != nan AND $this > 0) ? ($outbound_packets_dropped * 100 / $this) : (0))
    units: %
    every: 1m
     warn: $this > 0.5
     crit: $this > 3
-   delay: down 30m multiplier 1.5 max 1h
-    info: the ratio of outbound dropped packets vs the total number of sent packets of the network interface, during the last hour
+   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
 
 
@@ -71,16 +61,17 @@ template: 1hour_packet_drops_ratio_outbound
 # check if an interface is having FIFO
 # buffer errors
 # the alarm is checked every 1 minute
-# and examines the last hour of data
+# and examines the last 10 minutes of data
 
-template: 1hour_fifo_errors
+template: 10min_fifo_errors
       on: net.fifo
-  lookup: sum -1h unaligned absolute
+families: *
+  lookup: sum -10m unaligned absolute
    units: errors
    every: 1m
     warn: $this > 0
-   delay: down 30m multiplier 1.5 max 1h
-    info: interface fifo errors in the last hour
+   delay: down 1h multiplier 1.5 max 2h
+    info: interface fifo errors in the last 10 minutes
       to: sysadmin
 
 
@@ -95,6 +86,7 @@ template: 1hour_fifo_errors
 
 template: 1m_received_packets_rate
       on: net.packets
+families: *
   lookup: average -1m of received
    units: packets
    every: 10s
@@ -102,12 +94,14 @@ template: 1m_received_packets_rate
 
 template: 10s_received_packets_storm
       on: net.packets
+families: *
   lookup: average -10s of received
     calc: $this * 100 / (($1m_received_packets_rate < 1000)?(1000):($1m_received_packets_rate))
    every: 10s
    units: %
    warn: $this > (($status >= $WARNING)?(200):(1000))
    crit: $this > (($status >= $WARNING)?(1000):(2000))
+options: no-clear-notification
    info: the % of the rate of received packets in the last 10 seconds, compared to the rate of the last minute
-     to: silent
+     to: sysadmin