From 43992e6511c97e457662820c2eea58dfc40f071b Mon Sep 17 00:00:00 2001 From: Mark Huang Date: Wed, 19 Jul 2006 19:47:03 +0000 Subject: [PATCH] merge 0.4-7.planetlab changes --- pl_mom.spec | 17 ++++++++-- swapmon.py | 92 ++++++++++++++++++++++++++++++++--------------------- 2 files changed, 70 insertions(+), 39 deletions(-) diff --git a/pl_mom.spec b/pl_mom.spec index f98a708..5ea1e72 100644 --- a/pl_mom.spec +++ b/pl_mom.spec @@ -1,6 +1,6 @@ %define name pl_mom %define version 0.4 -%define release 6%{?pldistro:.%{pldistro}}%{?date:.%{date}} +%define release 7%{?pldistro:.%{pldistro}}%{?date:.%{date}} Summary: PlanetLab node monitoring tools Name: %{name} @@ -104,8 +104,19 @@ exit 0 %{_sysconfdir}/cron.d/pl_mom %changelog +* Wed Jul 19 2006 Faiyaz Ahmed 0.4-7.planetlab +- swapmon: Added state info so same reset email does not keep getting + sent. If slice is already being killed, will wait for a timeout + before trying to kill again. + +* Mon Jul 17 2006 Faiyaz Ahmed 0.4-6.planetlab +- swapmon: Added state information per slice. Slices are now killed + after consuming more memory than the threshold more than twice. + Also fixed the slice restart issue where slices would be restarted + and repeated restart emails would be sent. + * Mon Jul 10 2006 Faiyaz Ahmed 0.4-5.planetlab -- bwlimit: Initialize default limits +- bwmon: Initialize default limits * Mon Jul 10 2006 Faiyaz Ahmed 0.4-4.planetlab -- bwlimit: Use cached values if NM stops responding +- bwmon: Use cached values if NM stops responding diff --git a/swapmon.py b/swapmon.py index 7122481..671663d 100755 --- a/swapmon.py +++ b/swapmon.py @@ -10,7 +10,7 @@ # Faiyaz Ahmed # Copyright (C) 2004-2006 The Trustees of Princeton University # -# $Id: swapmon.py,v 1.6 2006/07/17 19:31:27 faiyaza Exp $ +# $Id: swapmon.py,v 1.9 2006/07/19 19:40:55 faiyaza Exp $ # import syslog @@ -45,19 +45,22 @@ period = 30 change_thresh = 5 # Swap utilization at which the largest consumer of physical memory is reset -reset_thresh = 75 +reset_thresh = 80 # Swap utilization at which the machine is rebooted reboot_thresh = 95 # Time to wait before checking slice again after reset -reset_timeout = 10 +reset_timeout = 15 # Number of strikes before killing (strike, strike, kill) kill_thresh = 2 -# Time to wait before removing slice from kill queue -kill_timeout = 30 +# Time to wait before removing slice from kill queue (probation) +kill_timeout = 120 + +# Don't email the same message more than once in the same emailtimeout interval +email_timeout = 1800 # Minimum physical memory utilization to be considered the largest consumer min_thresh = 10 @@ -146,14 +149,19 @@ class Reset: resettimeleft - timeout before checking for next reset resetcount - number of strikes - killtimeleft - time out before removing from kill queue + killtimeleft - time out before removing from kill queue + {kill,reset}mail - Time of last email + kill - State of kill. If slice is already being killed, wait before retry. """ def __init__(self,name): self.name = name self.resettimeleft = reset_timeout self.resetcount = 0 + self.resetmail = 0 + self.kill = False self.killtimeleft = kill_timeout + self.killmail = 0 def __repr__(self): return self.name @@ -163,22 +171,25 @@ class Reset: if self.resettimeleft > 0: self.resettimeleft -= 1 if debug and verbose: - print "%s has %s seconds to die and has been reset %s times" \ - %(self.name, self.resettimeleft, self.resetcount) print "%s has %s seconds in probation" \ %(self.name, self.killtimeleft) if self.killtimeleft > 0: # Count down kill probation timer (killtimeleft) self.killtimeleft -= 1 + if self.killtimeleft == 1: + print "%s is out of probation" % self.name else: - # Once out of probation period (killtimeleft), remove strike + # Once out of probation period (killtimeleft), remove strikes self.resetcount = 0 + self.kill = False # Check to see if a slice needs to be killed. If it has rules more than kill_thresh in # the probation period (kill_timeout) send an email, kill the slice. def checkkill(self,params): - if self.killtimeleft > 0 and self.resetcount >= kill_thresh: + if self.killtimeleft > 0 and self.resetcount >= kill_thresh and \ + self.kill == False: + self.kill = True if debug: print kill_subject % params print kill_body % params @@ -186,13 +197,18 @@ class Reset: pid = os.fork() if pid == 0: print "Slice %s is being killed." % self.name - vserver = VServer(self.name) - vserver.stop() + vserver = VServer(self.name) + vserver.stop() + os._exit(0) else: os.waitpid(pid,0) except Exception, err: - print "Warning: Exception received while killing slice %s: %s" % self.name, err - slicemail(self.name, kill_subject % params, kill_body % params) + print "Warning: Exception received while killing slice %s: %s" \ + % self.name, err + if (time.time() - self.killmail) > email_timeout: + slicemail(self.name, kill_subject % params, kill_body % params) + print "Sending KILL email for slice %s" % self.name + self.killmail = time.time() return True return False @@ -209,24 +225,28 @@ class Reset: self.resetcount += 1 self.killtimeleft = kill_timeout self.resettimeleft = reset_timeout + print "%s has %s seconds to die and has been reset %s times" \ + %(self.name, self.resettimeleft, self.resetcount) if debug: print reset_subject % params print reset_body % params try: pid = os.fork() if pid == 0: - print "Resetting slice " + self.name - vserver = VServer(self.name) - vserver.stop() - vserver.start(wait = False) - os._exit(0) - else: - os.waitpid(pid, 0) + print "Resetting slice " + self.name + vserver = VServer(self.name) + vserver.stop() + vserver.start(wait = False) + os._exit(0) + else: + os.waitpid(pid,0) except Exception, err: - print "Warning: Exception received while resetting slice %s:" \ + print "Warning: Exception received while resetting slice %s:" \ % self.name, err - slicemail(self.name, reset_subject % params, reset_body % params) - + if (time.time() - self.resetmail) > email_timeout: + slicemail(self.name, reset_subject % params, reset_body % params) + print "Sending Reset email for slice %s" % self.name + self.resetmail = time.time() def usage(): @@ -470,7 +490,7 @@ def main(): (version, slices) = pickle.load(f) f.close() # Check version of data file - if version != "$Id: swapmon.py,v 1.6 2006/07/17 19:31:27 faiyaza Exp $": + if version != "$Id: swapmon.py,v 1.9 2006/07/19 19:40:55 faiyaza Exp $": print "Not using old version '%s' data file %s" % (version, datafile) raise Exception @@ -487,7 +507,7 @@ def main(): # Delete data file os.unlink(datafile) except Exception: - version = "$Id: swapmon.py,v 1.6 2006/07/17 19:31:27 faiyaza Exp $" + version = "$Id: swapmon.py,v 1.9 2006/07/19 19:40:55 faiyaza Exp $" slices = {} # Query process table every 30 seconds, or when a large change in @@ -562,16 +582,16 @@ def main(): # Match slice name against system slice patterns is_system_slice = filter(None, [re.match(pattern, slice['name']) for pattern in system_slices]) - if is_system_slice: - # Do not reset system slices, just warn once - if slice['name'] not in warned: - warned.append(slice['name']) - if debug: - print alarm_subject % params - print alarm_body % params - else: - print "Warning slice " + slice['name'] - slicemail(slice['name'], alarm_subject % params, alarm_body % params) + if is_system_slice: + if slice['name'] not in warned: + warned.append(slice['name']) + if debug: + print alarm_subject % params + print alarm_body % params + else: + print "Warning slice " + slice['name'] + slicemail(slice['name'], alarm_subject % params, + alarm_body % params) else: # Reset slice if not resetlist.has_key(slice['name']): -- 2.47.0