sensu-plugins-edgelab 1.3.1 → 1.4.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 4a8a2860bfd4180b4d7a0199449c484799418376
-  data.tar.gz: 110365a9d4347b06e9bf626fb8d65993bf1bd27a
+  metadata.gz: aa20291918973ec9b435b65626dec763c7b36be2
+  data.tar.gz: 27fc4397fb9b6454d504239f469abe596a465806
 SHA512:
-  metadata.gz: 3385985143bdc67ebc208d152727852ffbb71ad96a90146b69a26fea321b695aa6035691c5f9e9113a9fc6331f4bc4c283ebd5fcbe98851244cb2c4fad1b8e4d
-  data.tar.gz: d11e760ad23c4ff5e013f1d3bc5083cce74640c509aa57a9e88ccf871f270314a9f515af235287a11b4100403143896bb926d48ea93c1d3c847314a9666a409e
+  metadata.gz: 2c3b8052c10cd420be8c30891cd453ef1bac9997c2b5d2c319f6573919f8ad469351d58065b695a6e61253db99fbc0714604f2f1c39fced570d21d11edb15147
+  data.tar.gz: 3b18f436aa32a778743c465ae63fa3fe55019a1a7fbc5fa08808e078b53e02c0efbb82a72e8e0443a1e3281ca3d0fd681342ba67b22707996dc3ecf2918b8e98

data/bin/check-nomad-jobs.rb

@@ -17,11 +17,22 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
          description: 'Nomad server URL',
          long: '--nomad SERVER',
          default: 'http://localhost:4646'
+
   option :alloc_starting_time,
          description: '',
-         long: '--alloc-starting-time',
+         long: '--alloc-starting-time SECONDS',
          default: 300
 
+  option :alloc_restarts_count,
+         description: 'Limit number of restarts in restarts interval',
+         long: '--alloc-restarts-count COUNT',
+         default: 3
+
+  option :alloc_restarts_interval,
+         description: 'Interval in seconds for the limit number of restarts',
+         long: '--alloc-restarts-interval SECONDS',
+         default: 3600
+
   # Call Nomad api and parse the JSON response
   def api_call(endpoint)
     url = config[:nomad] + endpoint
@@ -125,6 +136,32 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
     end
   end
 
+  # Check that running allocations are not restarting endlessly.
+  def check_restarts(job, failed)
+    allocations = api_call "/v1/job/#{job['ID']}/allocations"
+    now = Time.new.to_i
+
+    allocations.each do |alloc|
+      if %w(running pending).include? alloc['ClientStatus']
+        alloc['TaskStates'].each do |_, state|
+          restarts = 0
+          state['Events'].each do |event|
+            if event['Type'] == 'Restarting'
+              event_time = event['Time'] / 1_000_000_000
+              if (now - event_time) < config[:alloc_restarts_interval].to_i
+                restarts += 1
+              end
+            end
+          end
+
+          if restarts >= config[:alloc_restarts_count].to_i
+            failed << "Alloc #{alloc['Name']} restart #{restarts} times"
+          end
+        end
+      end
+    end
+  end
+
   def run
     jobs = api_call '/v1/jobs'
     if jobs.empty?
@@ -136,6 +173,7 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
     jobs.each do |job|
       check_evaluations job, failed
       check_allocations job, failed
+      check_restarts job, failed
     end
 
     if failed.any?

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: sensu-plugins-edgelab
 version: !ruby/object:Gem::Version
-  version: 1.3.1
+  version: 1.4.0
 platform: ruby
 authors:
 - Edgelab
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2017-02-09 00:00:00.000000000 Z
+date: 2017-02-10 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: sensu-plugin
@@ -111,12 +111,12 @@ dependencies:
 description: Sensu plugins developed by Edgelab
 email: 
 executables:
-- check-nomad-leader.rb
+- check-swarm-cluster.rb
+- metrics-mysql-processes.rb
 - metrics-swarm-cluster.rb
 - check-apt-expired-keys.rb
 - check-nomad-jobs.rb
-- check-swarm-cluster.rb
-- metrics-mysql-processes.rb
+- check-nomad-leader.rb
 extensions: []
 extra_rdoc_files: []
 files: