Merge pull request #79 from Rajendra-R/flush_metrics

#76 implemented flush grouped metrics when it program terminates

basescript/basescript.py

@@ -45,15 +45,18 @@ def __init__(self, args=None):
         if self.args.metric_grouping_interval is None:
             self.args.metric_grouping_interval = self.METRIC_GROUPING_INTERVAL
 
-        self.log = init_logger(
+        log = init_logger(
             fmt=self.args.log_format,
             quiet=self.args.quiet,
             level=self.args.log_level,
             fpath=self.args.log_file,
             pre_hooks=self.define_log_pre_format_hooks(),
             post_hooks=self.define_log_post_format_hooks(),
             metric_grouping_interval=self.args.metric_grouping_interval
-        ).bind(name=self.args.name)
+        )
+
+        self._flush_metrics_q = log._force_flush_q
+        self.log = log.bind(name=self.args.name)
 
         self.stats = Dummy()
 
@@ -73,12 +76,13 @@ def start(self):
                 raise
         except KeyboardInterrupt:
             self.log.warning("exited via keyboard interrupt")
-            sys.exit(1)
         except:
             self.log.exception("exited start function")
             # set exit code so we know it did not end successfully
             # TODO different exit codes based on signals ?
-            sys.exit(1)
+        finally:
+            self._flush_metrics_q.put(None, block=True)
+            self._flush_metrics_q.put(None, block=True, timeout=1)
 
         self.log.info("exited successfully")
 

basescript/log.py

@@ -5,16 +5,18 @@
 import socket
 import logging
 import numbers
+import queue
 from threading import Thread, Lock
 from datetime import datetime
 from functools import wraps
 
-from deeputil import Dummy
+from deeputil import Dummy, keeprunning
 import structlog
 
 # stdlib to structlog handlers should be configured only once.
 _GLOBAL_LOG_CONFIGURED = False
 
+FORCE_FLUSH_Q_SIZE = 1
 HOSTNAME = socket.gethostname()
 METRICS_STATE = {}
 METRICS_STATE_LOCK = Lock()
@@ -270,11 +272,18 @@ def _structlog_default_keys_processor(logger_class, log_method, event):
 
     return event
 
+@keeprunning()
 def dump_metrics(log, interval):
     global METRICS_STATE
 
+    terminate = False
+
     while True:
-        time.sleep(interval)
+        try:
+            log._force_flush_q.get(block=True, timeout=interval)
+            terminate = True
+        except queue.Empty:
+            pass
 
         METRICS_STATE_LOCK.acquire()
         m = METRICS_STATE
@@ -292,6 +301,9 @@ def dump_metrics(log, interval):
             fn = getattr(log, level)
             fn(event, type='metric', __grouped__=True, num=n, **d)
 
+        if terminate:
+            break
+
 def metrics_grouping_processor(logger_class, log_method, event):
     if event.get('type') == 'logged_metric':
         event['type'] = 'metric'
@@ -459,6 +471,8 @@ def init_logger(
     level = getattr(logging, level.upper())
     log.setLevel(level)
 
+    log._force_flush_q = queue.Queue(maxsize=FORCE_FLUSH_Q_SIZE)
+
     if metric_grouping_interval:
         keep_running = Thread(target=dump_metrics, args=(log, metric_grouping_interval))
         keep_running.daemon = True

setup.py

@@ -22,7 +22,7 @@ def get_long_description():
 
 long_description = get_long_description()
 
-version = '0.2.6'
+version = '0.2.7'
 setup(
     name="basescript",
     version=version,