[or-cvs] r19657: {torflow} Add checks to filter out any nodes that clearly weren't arou (torflow/trunk/NetworkScanners/BwAuthority)

[mikeperry at seul.org]

Author: mikeperry
Date: 2009-06-08 00:48:04 -0400 (Mon, 08 Jun 2009)
New Revision: 19657

Modified:
   torflow/trunk/NetworkScanners/BwAuthority/aggregate.py
Log:

Add checks to filter out any nodes that clearly weren't
around during our last measurements from our messages
about missing them. There still are about 80 nodes
that are escaping measurement, possibly due to hibernation
changes.




Modified: torflow/trunk/NetworkScanners/BwAuthority/aggregate.py
===================================================================
--- torflow/trunk/NetworkScanners/BwAuthority/aggregate.py	2009-06-08 03:03:24 UTC (rev 19656)
+++ torflow/trunk/NetworkScanners/BwAuthority/aggregate.py	2009-06-08 04:48:04 UTC (rev 19657)
@@ -4,6 +4,7 @@
 import math
 import sys
 import socket
+import time
 
 sys.path.append("../../")
 from TorCtl.TorUtil import plog
@@ -202,7 +203,14 @@
 
   for n in prev_consensus.itervalues():
     if not n.measured:
-      plog("INFO", "Didn't measure "+n.idhex+"="+n.nickname)
+      if "Fast" in n.flags and "Running" in n.flags:
+        r = c.get_router(n)
+        if r and not r.down and r.bw > 0:
+          if time.mktime(r.published.utctimetuple()) - r.uptime \
+                 < oldest_timestamp:
+            # We still tend to miss about 80 nodes even with these
+            # checks.. Possibly going in and out of hibernation?
+            plog("INFO", "Didn't measure "+n.idhex+"="+n.nickname)
 
   n_print = nodes.values()
   n_print.sort(lambda x,y: int(x.new_bw) - int(y.new_bw))