sensu-plugins-edgelab 1.2.1 → 1.2.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: eed99dc209e0eb7a588c6b73d5f99e285c1e1825
-  data.tar.gz: b246d4c581b0e6a2c3d0fbb4145170617433940f
+  metadata.gz: ea0c07087191700425e9030541a612e62b7bce0c
+  data.tar.gz: 8f26d3185e70df103814a71d80bfeb0a0f345aed
 SHA512:
-  metadata.gz: 5494b1d94ae2f340c798fcd6bbadc132dbc1a11417504f614af89615d9f1fca866ac87e87930b75f30bbe0991348f17ecb5383bd198c5fb9d93d61cc545a594a
-  data.tar.gz: a67e3152c3f83f2c30091086b6ffef6d74a29a07b8b99e69bcd3da205f027db1497dc9c3b83a52ce0abc1aeae3ba68565534ca2bc73b4b99ba5bf7b122388d0d
+  metadata.gz: d785ffe04f0a3d3fe258c70f565e6c70dc2dec837bbd3fcf15eeaa22dd35cfd43cab1f5c3aafd6d923acc9e6bb7bac271f8267b706e95456423f93f5d10805d4
+  data.tar.gz: ebab2d70d72dae8eb52960e5f0d92218b21568458ab3f9727cbe55e46176daa8ec9ba659809e26022a6fc617f9abc4b9bfc148d482c62ab586e6f025ef92cdda

data/bin/check-nomad-jobs.rb

@@ -17,10 +17,14 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
          description: 'nomad server address',
          long: '--nomad SERVER',
          default: 'http://localhost:4646'
+  option :alloc_starting_time,
+         description: '',
+         long: '--alloc-starting-time',
+         default: 300
 
-  # Return, as array of hash, all registered jobs in Nomad
-  def nomad_jobs
-    url = config[:nomad] + '/v1/jobs'
+  # Call Nomad api and parse the json response
+  def api_call(endpoint)
+    url = config[:nomad] + endpoint
     begin
       response = RestClient.get(url)
     rescue => e
@@ -34,8 +38,40 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
     end
   end
 
+  # Check that allocations are in the desired status
+  def check_allocations(job, failed)
+    allocations = api_call "/v1/job/#{job['ID']}/allocations"
+
+    allocations.each do |alloc|
+      if alloc['DesiredStatus'] == 'run'
+        # Batch stay in run DesiredStatus even if task completed successfully.
+        next if job['Type'] == 'batch' and alloc['ClientStatus'] == 'complete'
+
+        alloc['TaskStates'].each do |_, state|
+          if state['State'] == 'dead'
+              failed << "Alloc #{alloc['Name']} is dead but desired status is 'run'"
+
+          # Check that pending alloc are not too old
+          elsif state['State'] == 'pending'
+            # Get the last event timestamp (which is in microseconds in Nomad)
+            event_time = state['Events'][-1]['Time'] / 1_000_000_000
+            starting_time = (Time.new - Time.at(event_time)).round
+
+            if starting_time > config[:alloc_starting_time]
+              failed << "Alloc #{alloc['Name']} is pending since #{starting_time} seconds"
+
+              # No need to check other task in the same task group.
+              break
+            end
+
+          end
+        end
+      end
+    end
+  end
+
   def run
-    jobs = nomad_jobs
+    jobs = api_call '/v1/jobs'
     if jobs.empty?
       critical 'No jobs found in Nomad.'
     end
@@ -43,11 +79,7 @@ class CheckNomadAllocations < Sensu::Plugin::Check::CLI
     failed = []
 
     jobs.each do |job|
-      job['JobSummary']['Summary'].each do |group, summary|
-        if summary['Failed'] != 0
-          failed << "#{job['Name']}.#{group}"
-        end
-      end
+      check_allocations 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: 1.2.1
+  version: 1.2.2
 platform: ruby
 authors:
 - Edgelab
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2017-01-12 00:00:00.000000000 Z
+date: 2017-01-23 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: sensu-plugin