From: Faiyaz Ahmed Date: Mon, 12 Nov 2007 21:18:23 +0000 (+0000) Subject: tabs != spaces. I hate CVS. X-Git-Tag: 2.2-02~6 X-Git-Url: http://git.onelab.eu/?p=mom.git;a=commitdiff_plain;h=3a0a4118e3adf3aaa282a5adedc1671987ab3c1e tabs != spaces. I hate CVS. --- diff --git a/swapmon.py b/swapmon.py index 37743b1..f15879f 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 1.10 2006/08/16 16:18:45 faiyaza Exp $ +# $Id: swapmon.py,v 1.11 2006/12/02 19:11:47 mlhuang Exp $ # import syslog @@ -146,107 +146,104 @@ and what you are doing to address the problem. class Reset: - """ - Keeps track of state information for resets and kills - - resettimeleft - timeout before checking for next reset - resetcount - number of strikes - 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.killtimeleft = kill_timeout - self.killmail = 0 - - def __repr__(self): - return self.name - - def update(self): - # Count down for next check of reset slice. - if self.resettimeleft > 0: - self.resettimeleft -= 1 - if debug and verbose: - 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 strikes - self.resetcount = 0 - - - # Check to see if a slice needs to be killed. If it has been killed 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 debug: + """ + Keeps track of state information for resets and kills + + resettimeleft - timeout before checking for next reset + resetcount - number of strikes + 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.killtimeleft = kill_timeout + self.killmail = 0 + + def __repr__(self): + return self.name + + def update(self): + # Count down for next check of reset slice. + if self.resettimeleft > 0: + self.resettimeleft -= 1 + if debug and verbose: 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 strikes + self.resetcount = 0 + + + # Check to see if a slice needs to be killed. If it has been killed 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 debug: print kill_subject % params print kill_body % params - try: - pid = os.fork() - if pid == 0: - print "Slice %s is being killed." % self.name - 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 - 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 - - # Reset slice after checking to see if slice is out of timeout. - # Increment resetcount, check to see if larger than kill_thresh. - def reset(self, params): - # If its the first reset (came back after kill) - # or if its been reset before - # and we are out of the reset timeout. - if self.resetcount == 0 or self.resettimeleft == 0: - # Do we need to kill this slice? Check history first. - if self.checkkill(params): - return - # Update counters - 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) - except Exception, err: - print "Warning: Exception received while resetting slice %s:" \ - % self.name, err - 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() - + try: + pid = os.fork() + if pid == 0: + print "Slice %s is being killed." % self.name + 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) + 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 + + # Reset slice after checking to see if slice is out of timeout. + # Increment resetcount, check to see if larger than kill_thresh. + def reset(self, params): + # If its the first reset (came back after kill) + # or if its been reset before + # and we are out of the reset timeout. + if self.resetcount == 0 or self.resettimeleft == 0: + # Do we need to kill this slice? Check history first. + if self.checkkill(params): return + # Update counters + 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) + except Exception, err: + print "Warning: Exception received while resetting slice %s:" \ + % self.name, err + 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(): print """ @@ -277,7 +274,7 @@ def slicestat(names = None): 'vsize': virtual_kib, 'sz': potential_kib, 'rss': physical_kib, 'pcpu': cpu_percent, 'pmem': mem_percent}] 'vsize': total_virtual_kib, - 'sz': total_potential_kib, + 'sz': total_potential_kib, 'rss': total_physical_kib}} """ @@ -362,7 +359,7 @@ def slicestat(names = None): slice['procs'].append(proc) slice['vsize'] += proc['vsize'] - slice['sz'] += proc['sz'] + slice['sz'] += proc['sz'] slice['rss'] += proc['rss'] slices[proc['xid']] = slice @@ -379,14 +376,14 @@ def memtotal(): meminfo = open("/proc/meminfo", "r") for line in meminfo.readlines(): - try: - (name, value, kb) = line.split() - except: - continue - if name == "MemTotal:": - mem = int(value) - elif name == "SwapTotal:": - swap = int(value) + try: + (name, value, kb) = line.split() + except: + continue + if name == "MemTotal:": + mem = int(value) + elif name == "SwapTotal:": + swap = int(value) meminfo.close() return (mem, swap) @@ -422,11 +419,11 @@ def summary(slices = None, total_mem = None, total_swap = None): Return a summary of memory usage by slice. """ if not slices: - slices = slicestat() + slices = slicestat() slicelist = slices.values() slicelist.sort(lambda a, b: b['sz'] - a['sz']) if total_mem is None or total_swap is None: - (total_mem, total_swap) = memtotal() + (total_mem, total_swap) = memtotal() table = "%-20s%10s%24s%24s\n\n" % ("Slice", "Processes", "Memory Usage", "Potential Usage") for slice in slicelist: @@ -436,7 +433,7 @@ def summary(slices = None, total_mem = None, total_swap = None): 100. * slice['rss'] / total_mem, format_bytes(slice['sz'] * 1024, si = False), 100. * slice['sz'] / (total_mem + total_swap)) - + return table @@ -505,7 +502,7 @@ def main(): (version, slices) = pickle.load(f) f.close() # Check version of data file - if version != "$Id: swapmon.py,v 1.10 2006/08/16 16:18:45 faiyaza Exp $": + if version != "$Id: swapmon.py,v 1.11 2006/12/02 19:11:47 mlhuang Exp $": print "Not using old version '%s' data file %s" % (version, datafile) raise Exception @@ -522,7 +519,7 @@ def main(): # Delete data file os.unlink(datafile) except Exception: - version = "$Id: swapmon.py,v 1.10 2006/08/16 16:18:45 faiyaza Exp $" + version = "$Id: swapmon.py,v 1.11 2006/12/02 19:11:47 mlhuang Exp $" slices = {} # Query process table every 30 seconds, or when a large change in @@ -540,85 +537,82 @@ def main(): while True: used = swap_used() - for resetslice in resetlist.keys(): - resetlist[resetslice].update() - + for resetslice in resetlist.keys(): + resetlist[resetslice].update() + if last_used is None: last_used = used - if verbose: + if verbose: print "%d%% swap consumed" % used - if used >= reboot_thresh: - # Dump slice state before rebooting - if verbose: - print "Saving %s" % datafile - f = open(datafile, "w") - pickle.dump((version, slices), f) - f.close() - - # Goodbye, cruel world - print "%d%% swap consumed, rebooting" % used - if not debug: - bwlimit.run("/bin/sync; /sbin/reboot -f") - - elif used >= reset_thresh: - if debug: - print "Memory used = %s" %(used) - # Try and find a hog - slicelist = slices.values() - slicelist.sort(lambda a, b: b['rss'] - a['rss']) - for slice in slicelist: - percent = 100. * slice['rss'] / total_mem - - if slice['rss'] < rss_min: - continue - - print "%d%% swap consumed, slice %s is using %s (%d%%) of memory" % \ - (used, - slice['name'], - format_bytes(slice['rss'] * 1024, si = False), - percent) - - slice['procs'].sort(lambda a, b: b['rss'] - a['rss']) - - table = "%5s %10s %10s %10s %4s %4s %s\n\n" % ("PID", "VIRT", "SZ", "RES", '%CPU', '%MEM', 'COMMAND') - for proc in slice['procs']: - table += "%5s %10s %10s %10s %4.1f %4.1f %s\n" % \ - (proc['pid'], - format_bytes(proc['vsize'] * 1024, si = False), - format_bytes(proc['sz'] * 1024, si = False), - format_bytes(proc['rss'] * 1024, si = False), - proc['pcpu'], proc['pmem'], proc['cmd']) - - params = {'hostname': socket.gethostname(), - 'date': time.asctime(time.gmtime()) + " GMT", - 'table': table, - 'slice': slice['name'], - 'rss': format_bytes(slice['rss'] * 1024, si = False), - 'sz': format_bytes(slice['sz'] * 1024, si = False), - 'percent': percent} - - # 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 used >= reboot_thresh: + # Dump slice state before rebooting + if verbose: print "Saving %s" % datafile + f = open(datafile, "w") + pickle.dump((version, slices), f) + f.close() + + # Goodbye, cruel world + print "%d%% swap consumed, rebooting" % used + if not debug: bwlimit.run("/bin/sync; /sbin/reboot -f") + + elif used >= reset_thresh: + if debug: print "Memory used = %s" %(used) + # Try and find a hog + slicelist = slices.values() + slicelist.sort(lambda a, b: b['rss'] - a['rss']) + for slice in slicelist: + percent = 100. * slice['rss'] / total_mem + if slice['rss'] < rss_min: continue + print "%d%% swap consumed, slice %s is using %s (%d%%) of memory" % \ + (used, + slice['name'], + format_bytes(slice['rss'] * 1024, si = False), + percent) + + slice['procs'].sort(lambda a, b: b['rss'] - a['rss']) + table = "%5s %10s %10s %10s %4s %4s %s\n\n" % \ + ("PID", "VIRT", "SZ", "RES", '%CPU', '%MEM', 'COMMAND') + for proc in slice['procs']: + table += "%5s %10s %10s %10s %4.1f %4.1f %s\n" % \ + (proc['pid'], + format_bytes(proc['vsize'] * 1024, si = False), + format_bytes(proc['sz'] * 1024, si = False), + format_bytes(proc['rss'] * 1024, si = False), + proc['pcpu'], + proc['pmem'], + proc['cmd']) + + params = {'hostname': socket.gethostname(), + 'date': time.asctime(time.gmtime()) + " GMT", + 'table': table, + 'slice': slice['name'], + 'rss': format_bytes(slice['rss'] * 1024, si = False), + 'sz': format_bytes(slice['sz'] * 1024, si = False), + 'percent': percent} + + # 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: - # Reset slice - if not resetlist.has_key(slice['name']): - resetlist[slice['name']] = Reset(slice['name']) - resetlist[slice['name']].reset(params) - slices = slicestat(names) + print "Warning slice " + slice['name'] + slicemail(slice['name'], alarm_subject % params, + alarm_body % params) + else: + # Reset slice + if not resetlist.has_key(slice['name']): + resetlist[slice['name']] = Reset(slice['name']) + resetlist[slice['name']].reset(params) + slices = slicestat(names) if timer <= 0 or used >= (last_used + change_thresh): if used >= (last_used + change_thresh):