Merge pull request #1441 from DataDog/leo/php_check

[php-fpm] new check, get perf insight of your php-fpm install

.travis.yml

@@ -64,6 +64,7 @@ env:
     - TRAVIS_FLAVOR=etcd
     - TRAVIS_FLAVOR=pgbouncer
     - TRAVIS_FLAVOR=supervisord
+    - TRAVIS_FLAVOR=phpfpm
 
 # Override travis defaults with empty jobs
 before_install: echo "OVERRIDING TRAVIS STEPS"

Rakefile

@@ -19,6 +19,7 @@ require './ci/mongo'
 require './ci/mysql'
 require './ci/nginx'
 require './ci/pgbouncer'
+require './ci/phpfpm'
 require './ci/postgres'
 require './ci/rabbitmq'
 require './ci/redis'

checks.d/php_fpm.py

@@ -0,0 +1,125 @@
+# 3p
+import requests
+
+# project
+from checks import AgentCheck
+from util import headers
+
+
+class PHPFPMCheck(AgentCheck):
+    """
+    Tracks basic php-fpm metrics via the status module
+    Requires php-fpm pools to have the status option.
+    See http://www.php.net/manual/de/install.fpm.configuration.php#pm.status-path for more details
+    """
+
+    SERVICE_CHECK_NAME = 'php_fpm.can_ping'
+
+    GAUGES = {
+        'listen queue': 'php_fpm.listen_queue.size',
+        'idle processes': 'php_fpm.processes.idle',
+        'active processes': 'php_fpm.processes.active',
+        'total processes': 'php_fpm.processes.total',
+    }
+
+    RATES = {
+        'max children reached': 'php_fpm.processes.max_reached'
+    }
+
+    COUNTERS = {
+        'accepted conn': 'php_fpm.requests.accepted',
+        'slow requests': 'php_fpm.requests.slow'
+    }
+
+    def check(self, instance):
+        status_url = instance.get('status_url')
+        ping_url = instance.get('ping_url')
+
+        auth = None
+        user = instance.get('user')
+        password = instance.get('password')
+
+        tags = instance.get('tags', [])
+
+        if user and password:
+            auth = (user, password)
+
+        if status_url is None and ping_url is None:
+            raise Exception("No status_url or ping_url specified for this instance")
+
+        pool = None
+        status_exception = None
+        if status_url is not None:
+            try:
+                pool = self._process_status(status_url, auth, tags)
+            except Exception as e:
+                status_exception = e
+                pass
+
+        if ping_url is not None:
+            self._process_ping(ping_url, auth, tags, pool)
+
+        if status_exception is not None:
+            raise status_exception
+
+    def _process_status(self, status_url, auth, tags):
+        data = {}
+        try:
+            # TODO: adding the 'full' parameter gets you per-process detailed
+            # informations, which could be nice to parse and output as metrics
+            resp = requests.get(status_url, auth=auth,
+                                headers=headers(self.agentConfig),
+                                params={'json': True})
+            resp.raise_for_status()
+
+            data = resp.json()
+        except Exception as e:
+            self.log.error("Failed to get metrics from {0}.\nError {1}".format(status_url, e))
+            raise
+
+        pool_name = data.get('pool', 'default')
+        metric_tags = tags + ["pool:{0}".format(pool_name)]
+
+        for key, mname in self.GAUGES.iteritems():
+            if key not in data:
+                self.log.warn("Gauge metric {0} is missing from FPM status".format(key))
+                continue
+            self.gauge(mname, int(data[key]), tags=metric_tags)
+
+        for key, mname in self.RATES.iteritems():
+            if key not in data:
+                self.log.warn("Rate metric {0} is missing from FPM status".format(key))
+                continue
+            self.rate(mname, int(data[key]), tags=metric_tags)
+
+        for key, mname in self.COUNTERS.iteritems():
+            if key not in data:
+                self.log.warn("Counter metric {0} is missing from FPM status".format(key))
+                continue
+            self.increment(mname, int(data[key]), tags=metric_tags)
+
+        # return pool, to tag the service check with it if we have one
+        return pool_name
+
+    def _process_ping(self, ping_url, auth, tags, pool_name):
+        sc_tags = tags[:]
+        if pool_name is not None:
+            sc_tags.append("pool:{0}".format(pool_name))
+
+        try:
+            # TODO: adding the 'full' parameter gets you per-process detailed
+            # informations, which could be nice to parse and output as metrics
+            resp = requests.get(ping_url, auth=auth,
+                                headers=headers(self.agentConfig))
+            resp.raise_for_status()
+
+            if 'pong' not in resp.text:
+                raise Exception("Received unexpected reply to ping {0}".format(resp.text))
+
+        except Exception as e:
+            self.log.error("Failed to ping FPM pool {0} on URL {1}."
+                           "\nError {2}".format(pool_name, ping_url, e))
+            self.service_check(self.SERVICE_CHECK_NAME,
+                               AgentCheck.CRITICAL, tags=sc_tags, message=str(e))
+        else:
+            self.service_check(self.SERVICE_CHECK_NAME, AgentCheck.OK, tags=sc_tags)

ci/phpfpm.rb

@@ -0,0 +1,85 @@
+require './ci/common'
+
+def nginx_version
+  ENV['NGINX_VERSION'] || '1.7.9'
+end
+
+def php_version
+  ENV['PHP_VERSION'] || '5.6.6'
+end
+
+def phpfpm_rootdir
+  "#{ENV['INTEGRATIONS_DIR']}/phpfpm_#{php_version}"
+end
+
+namespace :ci do
+  namespace :phpfpm do |flavor|
+    task :before_install => ['ci:common:before_install']
+
+    task :install => ['ci:common:install'] do
+      unless Dir.exist? File.expand_path(phpfpm_rootdir)
+        sh %(curl -s -L\
+             -o $VOLATILE_DIR/nginx-#{nginx_version}.tar.gz\
+             http://nginx.org/download/nginx-#{nginx_version}.tar.gz)
+        sh %(mkdir -p #{phpfpm_rootdir})
+        sh %(mkdir -p $VOLATILE_DIR/nginx)
+        sh %(tar zxf $VOLATILE_DIR/nginx-#{nginx_version}.tar.gz\
+             -C $VOLATILE_DIR/nginx --strip-components=1)
+        sh %(cd $VOLATILE_DIR/nginx\
+             && ./configure --prefix=#{phpfpm_rootdir} --with-http_stub_status_module\
+             && make -j $CONCURRENCY\
+             && make install)
+        sh %(curl -s -L\
+             -o $VOLATILE_DIR/php-#{php_version}.tar.bz2\
+             http://us1.php.net/get/php-#{php_version}.tar.bz2/from/this/mirror)
+        sh %(mkdir -p $VOLATILE_DIR/php)
+        sh %(tar jxf $VOLATILE_DIR/php-#{php_version}.tar.bz2\
+             -C $VOLATILE_DIR/php --strip-components=1)
+        sh %(cd $VOLATILE_DIR/php\
+             && ./configure --prefix=#{phpfpm_rootdir} --enable-fpm\
+             && make -j $CONCURRENCY\
+             && make install)
+      end
+    end
+
+    task :before_script => ['ci:common:before_script'] do
+      sh %(cp $TRAVIS_BUILD_DIR/ci/resources/phpfpm/nginx.conf\
+           #{phpfpm_rootdir}/conf/nginx.conf)
+      sh %(cp $TRAVIS_BUILD_DIR/ci/resources/phpfpm/php-fpm.conf\
+           #{phpfpm_rootdir}/etc/php-fpm.conf)
+      sh %(#{phpfpm_rootdir}/sbin/nginx -g "pid #{ENV['VOLATILE_DIR']}/nginx.pid;")
+      sh %(#{phpfpm_rootdir}/sbin/php-fpm -g #{ENV['VOLATILE_DIR']}/php-fpm.pid)
+    end
+
+    task :script => ['ci:common:script'] do
+      this_provides = [
+        'phpfpm'
+      ]
+      Rake::Task['ci:common:run_tests'].invoke(this_provides)
+    end
+
+    task :cleanup => ['ci:common:cleanup'] do
+      sh %(kill `cat $VOLATILE_DIR/nginx.pid`)
+      sh %(kill `cat $VOLATILE_DIR/php-fpm.pid`)
+    end
+
+    task :execute do
+      exception = nil
+      begin
+        %w(before_install install before_script script).each do |t|
+          Rake::Task["#{flavor.scope.path}:#{t}"].invoke
+        end
+      rescue => e
+        exception = e
+        puts "Failed task: #{e.class} #{e.message}".red
+      end
+      if ENV['SKIP_CLEANUP']
+        puts 'Skipping cleanup, disposable environments are great'.yellow
+      else
+        puts 'Cleaning up'
+        Rake::Task["#{flavor.scope.path}:cleanup"].invoke
+      end
+      fail exception if exception
+    end
+  end
+end

ci/resources/phpfpm/nginx.conf

@@ -0,0 +1,21 @@
+worker_processes  1;
+events {
+    worker_connections  1024;
+}
+http {
+    include       mime.types;
+    default_type  application/octet-stream;
+    sendfile        on;
+    keepalive_timeout  65;
+    server {
+        listen       42424;
+        server_name  localhost;
+        location ~ /(status|ping|\.*\.php)$ {
+            root           html;
+            fastcgi_pass   127.0.0.1:9000;
+            fastcgi_index  index.php;
+            fastcgi_param  SCRIPT_FILENAME  /scripts$fastcgi_script_name;
+            include        fastcgi_params;
+        }
+    }
+}

ci/resources/phpfpm/php-fpm.conf

@@ -0,0 +1,11 @@
+[www]
+user = nobody
+group = nobody
+listen = 127.0.0.1:9000
+pm = dynamic
+pm.max_children = 5
+pm.start_servers = 2
+pm.min_spare_servers = 1
+pm.max_spare_servers = 3
+pm.status_path = /status
+ping.path = /ping

conf.d/php_fpm.yaml.example

@@ -0,0 +1,24 @@
+init_config:
+
+instances:
+  - # Get metrics from your FPM pool with this URL
+    status_url: http://localhost/status
+    # Get a reliable service check of you FPM pool with that one
+    ping_url: http://localhost/ping
+    # These 2 URLs should follow the options from your FPM pool
+    # See http://php.net/manual/en/install.fpm.configuration.php
+    #   * pm.status_path
+    #   * ping.path
+    # You should configure your fastcgi passthru (nginx/apache) to
+    # catch these URLs and redirect them through the FPM pool target
+    # you want to monitor (FPM `listen` directive in the config, usually
+    # a UNIX socket or TCP socket.
+    #
+    # Use this if you have basic authentication on these pages
+    # user: bits
+    # password: D4T4D0G
+    #
+    # Array of custom tags
+    # By default metrics and service check will be tagged by pool and host
+    # tags:
+    #   - instance:foo

tests/common.py

@@ -180,7 +180,10 @@ def coverage_report(self):
                 tested_metrics += 1
             else:
                 untested_metrics.append(m)
-        coverage_metrics=100.0 * tested_metrics / total_metrics
+        if total_metrics == 0:
+            coverage_metrics = 100.0
+        else:
+            coverage_metrics = 100.0 * tested_metrics / total_metrics
 
         total_sc = len(self.service_checks)
         tested_sc = 0
@@ -190,7 +193,11 @@ def coverage_report(self):
                 tested_sc += 1
             else:
                 untested_sc.append(sc)
-        coverage_sc=100.0 * tested_sc / total_sc
+
+        if total_sc == 0:
+            coverage_sc = 100.0
+        else:
+            coverage_sc = 100.0 * tested_sc / total_sc
 
         coverage = """Coverage
 ========================================
@@ -240,8 +247,6 @@ def assertMetric(self, metric_name, value=None, tags=None, count=None, at_least=
         if count is not None:
             log.debug(" * should have exactly {0} data points".format(count))
         if at_least is not None:
-            if count is not None:
-                log.warn("Tolerance param will be ignored b/c count is present")
             log.debug(" * should have at least {0} data points".format(at_least))
 
         candidates = []
@@ -266,8 +271,6 @@ def assertMetricTagPrefix(self, metric_name, tag_prefix, count=None, at_least=1)
         if count is not None:
             log.debug(" * should have exactly {0} data points".format(count))
         if at_least is not None:
-            if count is not None:
-                log.warn("Tolerance param will be ignored b/c count is present")
             log.debug(" * should have at least {0} data points".format(at_least))
 
         candidates = []
@@ -290,8 +293,6 @@ def assertMetricTag(self, metric_name, tag, count=None, at_least=1):
         if count is not None:
             log.debug(" * should have exactly {0} data points".format(count))
         if at_least is not None:
-            if count is not None:
-                log.warn("Tolerance param will be ignored b/c count is present")
             log.debug(" * should have at least {0} data points".format(at_least))
 
         candidates = []