Changeset 31 for trunk/daemon

Timestamp:

04/01/05 13:31:53 (19 years ago)

Author:

bastiaans

Message:

daemon/togad.py:

Bug was caused by incorrect timestamping

• Now host reporting time is used

New bug: when a ganglia monitor has a big interval
it will report the same value for a second xmlthread run.

• One cannot update an rrd twice with the same timestamp
• Will have to check timestamp of metrics each run

This implicates that for every metric for every host the
timestamp of the last xml run must be kept in memory

• Might as well rewrite rrd code with in-memory metric storage
• And do scheduled/queued rrd updates, in stead of every xml run: big performance increase

File:

• trunk/daemon/togad.py (modified) (4 diffs)

trunk/daemon/togad.py

@@ -18 +18 @@
 # <=9  = RRD activity,gmetad config parsing
 # <=8  = host processing
-# <=7  = daemon threading
+# <=7  = daemon threading - NOTE: Daemon will 'halt on all errors' from this level
 #
 DEBUG_LEVEL = 8
@@ -96 +96 @@
                         myMetric['name'] = attrs.get('NAME',"")
                         myMetric['val'] = attrs.get('VAL',"")
-                        myMetric['time'] = self.time
+                        myMetric['time'] = self.hostReported
                         myMetric['type'] = attrs.get('TYPE',"")
 
@@ -117 +117 @@
                         correct_serial = self.rrd.checkNewRrdPeriod( self.hostName, mytime )
 
-                        debug_msg( 8, 'time %s: Storing metrics for %s' %(mytime, self.hostName) )
+                        #debug_msg( 8, 'time %s: Storing metrics for %s' %(mytime, self.hostName) )
+                        print 'Storing metrics for %s:' %(self.hostName),
                         self.storeMetrics( self.hostName, correct_serial )
 
@@ -128 +129 @@
 
                                 self.rrd.createCheck( hostname, metric, timeserial )    
+                                print ' [%s.%s]' %(metric['name'], metric['time']),
                                 self.rrd.update( hostname, metric, timeserial )
                                 debug_msg( 9, 'stored metric %s for %s: %s' %( hostname, metric['name'], metric['val'] ) )
                                 #sys.exit(1)
+                print