Merge pull request #19 from brian-brazil/standard-exports

Standard exports

Author: brian-brazil

AUTHORS.md

@@ -5,4 +5,5 @@ Maintainers of this repository:
 The following individuals have contributed code to this repository
 (listed in alphabetical order):
 
+* Andrea Fagan <[email protected]>
 * Brian Brazil <[email protected]>

README.md

@@ -133,6 +133,12 @@ c.labels('get', '/').inc()
 c.labels('post', '/submit').inc()
 ```
 
+### Process Collector
+
+The Python Client automatically exports metrics about process CPU usage, RAM,
+file descriptors and start time. These all have the prefix `process\_`, and
+are only currently available on Linux.
+
 ## Exporting
 
 There are several options for exporting metrics.
@@ -180,4 +186,5 @@ g.set(1)
 write_to_textfile('/configured/textfile/path/raid.prom', registry)
 ```
 
-A separate registry is used, as the default registry may contain other metrics.
+A separate registry is used, as the default registry may contain other metrics
+such as those from the Process Collector.

prometheus_client/__init__.py

@@ -4,6 +4,7 @@
 
 import copy
 import re
+import resource
 import os
 import time
 import threading
@@ -419,7 +420,7 @@ def generate_latest(registry=REGISTRY):
                 labelstr = '{{{0}}}'.format(','.join(
                     ['{0}="{1}"'.format(
                      k, v.replace('\\', r'\\').replace('\n', r'\n').replace('"', r'\"'))
-                     for k, v in labels.items()]))
+                     for k, v in sorted(labels.items())]))
             else:
                 labelstr = ''
             output.append('{0}{1} {2}\n'.format(name, labelstr, _floatToGoString(value)))
@@ -446,6 +447,83 @@ def write_to_textfile(path, registry):
     os.rename(tmppath, path)
 
 
+class ProcessCollector(object):
+    """Collector for Standard Exports such as cpu and memory."""
+    def __init__(self, namespace='', pid='self', proc='/proc', registry=REGISTRY):
+        self._namespace = namespace
+        self._pid = os.path.join(proc, str(pid))
+        self._proc = proc
+        self._pagesize = resource.getpagesize()
+        if namespace:
+            self._prefix = namespace + '_process_'
+        else:
+            self._prefix = 'process_'
+        self._ticks = 100.0
+        try:
+            self._ticks = os.sysconf('SC_CLK_TCK')
+        except (ValueError, TypeError):
+            pass
+
+        # This is used to test if we can access /proc.
+        self._btime = 0
+        try:
+            self._btime = self._boot_time()
+        except IOError:
+            pass
+        if registry:
+          registry.register(self)
+
+    def _boot_time(self):
+        with open(os.path.join(self._proc, 'stat')) as stat:
+            for line in stat:
+                if line.startswith('btime '):
+                    return float(line.split()[1])
+
+    def collect(self):
+        if not self._btime:
+            return []
+
+        result = []
+        try:
+            with open(os.path.join(self._pid, 'stat')) as stat:
+                parts = (stat.read().split(')')[-1].split())
+            vmem = Metric(self._prefix + 'virtual_memory_bytes', 'Virtual memory size in bytes', 'gauge')
+            vmem.add_sample(self._prefix + 'virtual_memory_bytes', {}, float(parts[20]))
+            rss = Metric(self._prefix + 'resident_memory_bytes', 'Resident memory size in bytes', 'gauge')
+            rss.add_sample(self._prefix + 'resident_memory_bytes', {}, float(parts[21]) * self._pagesize)
+            start_time = Metric(self._prefix + 'start_time_seconds',
+                                'Start time of the process since unix epoch in seconds.', 'gauge')
+            start_time_secs = float(parts[19]) / self._ticks
+            start_time.add_sample(self._prefix + 'start_time_seconds',{} , start_time_secs + self._btime)
+            utime = float(parts[11]) / self._ticks
+            stime = float(parts[12]) / self._ticks
+            cpu = Metric(self._prefix + 'cpu_seconds_total',
+                         'Total user and system CPU time spent in seconds.', 'counter')
+            cpu.add_sample(self._prefix + 'cpu_seconds_total', {}, utime + stime)
+            result.extend([vmem, rss, start_time, cpu])
+        except IOError:
+            pass
+
+        try:
+            max_fds = Metric(self._prefix + 'max_fds', 'Maximum number of open file descriptors.', 'gauge')
+            with open(os.path.join(self._pid, 'limits')) as limits:
+                for line in limits:
+                    if line.startswith('Max open file'):
+                        max_fds.add_sample(self._prefix + 'max_fds', {}, float(line.split()[3]))
+                        break
+            open_fds = Metric(self._prefix + 'open_fds', 'Number of open file descriptors.', 'gauge')
+            open_fds.add_sample(self._prefix + 'open_fds', {}, len(os.listdir(os.path.join(self._pid, 'fd'))))
+            result.extend([open_fds, max_fds])
+        except IOError:
+            pass
+
+        return result
+
+
+PROCESS_COLLECTOR = ProcessCollector()
+"""Default ProcessCollector in default Registry REGISTRY."""
+
+
 if __name__ == '__main__':
     c = Counter('cc', 'A counter')
     c.inc()

tests/proc/26231/fd/0

@@ -0,0 +1,15 @@
+Limit                     Soft Limit           Hard Limit           Units
+Max cpu time              unlimited            unlimited            seconds
+Max file size             unlimited            unlimited            bytes
+Max data size             unlimited            unlimited            bytes
+Max stack size            8388608              unlimited            bytes
+Max core file size        0                    unlimited            bytes
+Max resident set          unlimited            unlimited            bytes
+Max processes             62898                62898                processes
+Max open files            2048                 4096                 files
+Max locked memory         65536                65536                bytes
+Max address space         unlimited            unlimited            bytes
+Max file locks            unlimited            unlimited            locks
+Max pending signals       62898                62898                signals
+Max msgqueue size         819200               819200               bytes
+Max nice priority         0                    0

tests/proc/26231/fd/1

@@ -0,0 +1 @@
+26231 (vim) R 5392 7446 5392 34835 7446 4218880 32533 309516 26 82 1677 44 158 99 20 0 1 0 82375 56274944 1981 18446744073709551615 4194304 6294284 140736914091744 140736914087944 139965136429984 0 0 12288 1870679807 0 0 0 17 0 0 0 31 0 0 8391624 8481048 16420864 140736914093252 140736914093279 140736914093279 140736914096107 0