|
1 # -*- ruby -*- |
|
2 # vim: set noet nosta sw=4 ts=4 : |
|
3 |
|
4 require 'arborist/monitor/snmp' unless defined?( Arborist::Monitor::SNMP ) |
|
5 |
|
6 # Machine load/cpu checks. |
|
7 # |
|
8 # Sets current 1, 5, and 15 minute loads under the 'load' attribute, |
|
9 # and calculates/warns on cpu overutilization. |
|
10 # |
|
11 class Arborist::Monitor::SNMP::CPU |
|
12 include Arborist::Monitor::SNMP |
|
13 |
|
14 extend Configurability, Loggability |
|
15 log_to :arborist_snmp |
|
16 |
|
17 # OIDS for discovering system load. |
|
18 # |
|
19 OIDS = { |
|
20 load: '1.3.6.1.4.1.2021.10.1.3', |
|
21 cpu: '1.3.6.1.2.1.25.3.3.1.2' |
|
22 } |
|
23 |
|
24 # When walking load OIDS, the iterator count matches |
|
25 # these labels. |
|
26 # |
|
27 LOADKEYS = { |
|
28 1 => :load1, |
|
29 2 => :load5, |
|
30 3 => :load15 |
|
31 } |
|
32 |
|
33 |
|
34 # Global defaults for instances of this monitor |
|
35 # |
|
36 configurability( 'arborist.snmp.cpu' ) do |
|
37 # What overutilization percentage qualifies as a warning |
|
38 setting :warn_at, default: 80 |
|
39 end |
|
40 |
|
41 |
|
42 ### Return the properties used by this monitor. |
|
43 ### |
|
44 def self::node_properties |
|
45 return USED_PROPERTIES |
|
46 end |
|
47 |
|
48 |
|
49 ### Class #run creates a new instance and immediately runs it. |
|
50 ### |
|
51 def self::run( nodes ) |
|
52 return new.run( nodes ) |
|
53 end |
|
54 |
|
55 |
|
56 ### Perform the monitoring checks. |
|
57 ### |
|
58 def run( nodes ) |
|
59 super do |host, snmp| |
|
60 self.find_load( host, snmp ) |
|
61 end |
|
62 end |
|
63 |
|
64 |
|
65 ######### |
|
66 protected |
|
67 ######### |
|
68 |
|
69 ### Return system CPU data. |
|
70 ### |
|
71 def cpu( snmp ) |
|
72 return snmp.walk( OIDS[:cpu] ) |
|
73 end |
|
74 |
|
75 |
|
76 ### Find load data, add additional niceties for reporting. |
|
77 ### |
|
78 def format_load( snmp ) |
|
79 info = { cpu: {}, load: {} } |
|
80 cpus = self.cpu( snmp ) |
|
81 |
|
82 info[ :cpu ][ :count ] = cpus.size |
|
83 |
|
84 # Windows SNMP doesn't have a concept of "load" over time, |
|
85 # so we have to just use the current averaged CPU usage. |
|
86 # |
|
87 # This means that windows machines will very likely want to |
|
88 # adjust the default "overutilization" number, considering |
|
89 # it's really just how much of the CPU is used at the time of |
|
90 # the monitor run, along with liberal use of the Observer "only |
|
91 # alert after X events" pragmas. |
|
92 # |
|
93 if self.system =~ /windows\s+/i |
|
94 info[ :cpu ][ :usage ] = cpus.values.inject( :+ ).to_f / cpus.size |
|
95 info[ :message ] = "System is %0.1f%% in use." % [ info[ :cpu ][ :usage ] ] |
|
96 |
|
97 # UCDavis stuff is better for alerting only after there has been |
|
98 # an extended load event. Use the 5 minute average to avoid |
|
99 # state changes on transient spikes. |
|
100 # |
|
101 else |
|
102 snmp.walk( OIDS[:load] ).each_with_index do |(_, value), idx| |
|
103 next unless LOADKEYS[ idx + 1 ] |
|
104 info[ :load ][ LOADKEYS[idx + 1] ] = value.to_f |
|
105 end |
|
106 |
|
107 percentage = (( ( info[:load][ :load5 ] / cpus.size ) - 1 ) * 100 ).round( 1 ) |
|
108 |
|
109 if percentage < 0 |
|
110 info[ :message ] = "System is %0.1f%% idle." % [ percentage.abs ] |
|
111 info[ :cpu ][ :usage ] = percentage + 100 |
|
112 else |
|
113 info[ :message ] = "System is %0.1f%% overloaded." % [ percentage ] |
|
114 info[ :cpu ][ :usage ] = percentage |
|
115 end |
|
116 end |
|
117 |
|
118 return info |
|
119 end |
|
120 |
|
121 |
|
122 ### Collect the load information for +host+ from an existing |
|
123 ### (and open) +snmp+ connection. |
|
124 ### |
|
125 def find_load( host, snmp ) |
|
126 info = self.format_load( snmp ) |
|
127 |
|
128 config = identifiers[ host ].last || {} |
|
129 warn_at = config[ 'warn_at' ] || self.class.warn_at |
|
130 usage = info.dig( :cpu, :usage ) || 0 |
|
131 |
|
132 if usage >= warn_at |
|
133 info[ :warning ] = "%0.1f utilization exceeds %0.1f percent" % [ usage, warn_at ] |
|
134 end |
|
135 |
|
136 self.results[ host ] = info |
|
137 end |
|
138 |
|
139 end # class Arborist::Monitor::SNMP::CPU |
|
140 |