Hello,

I'm working on a failure model for a process production system and ran into the exception below. My model is generating two kinds of Process objects where one interrupts the other. The model generates 4 Coil objects (the interrupted) and as many Leak objects (the interrupter) as simulation time allow, usually a few. I think an instance of Leak is causing the KeyError, but I don't know how it is causing it. This happens only when I put the simulation into a loop (initialize(), simulate(), etc.), and always on the second simulation (using fixed seed in the random number generator). I searched the mailing list archives, but there is not much about KeyError's. Does anyone have an idea of what's causing this?

-José


1555.0 Coil 8 leaves. Total cycle time =47.0. Cycle Num=28
1556.5 Coil 6 leaves. Total cycle time =46.5. Cycle Num=29
1566.0 Coil 5 leaves. Total cycle time =47.0. Cycle Num=29
1584.0 Leak 5 created.
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 145, in <module>
    simulate(until=10*365)    # days
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 2009, in simulate
    a=_e._nextev()         
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 554, in _nextev
    tt=tempwho._nextpoint.next()
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 91, in PEM
    self.interrupt(c)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 404, in interrupt
    reactivate(victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 597, in reactivate
    a.cancel(obj)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 380, in cancel
    _e._unpost(whom=victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 524, in _unpost
    thislist=self.events[thistime]
KeyError: 3661.0


-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/
_______________________________________________
Simpy-users mailing list
Simpy-users <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/simpy-users

I suggest you let us have a listing of the program so we can try and repeat the error.

On 9/6/07, Jose Trejo-Meza < jl.trejo <at> gmail.com> wrote:

Hello,

I'm working on a failure model for a process production system and ran into the exception below. My model is generating two kinds of Process objects where one interrupts the other. The model generates 4 Coil objects (the interrupted) and as many Leak objects (the interrupter) as simulation time allow, usually a few. I think an instance of Leak is causing the KeyError, but I don't know how it is causing it. This happens only when I put the simulation into a loop (initialize(), simulate(), etc.), and always on the second simulation (using fixed seed in the random number generator). I searched the mailing list archives, but there is not much about KeyError's. Does anyone have an idea of what's causing this?

-José


1555.0 Coil 8 leaves. Total cycle time =47.0. Cycle Num=28
1556.5 Coil 6 leaves. Total cycle time =46.5. Cycle Num=29
1566.0 Coil 5 leaves. Total cycle time =47.0. Cycle Num=29
1584.0 Leak 5 created.
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 145, in <module>
    simulate(until=10*365)    # days
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 2009, in simulate
    a=_e._nextev()         
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 554, in _nextev
    tt=tempwho._nextpoint.next()
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 91, in PEM
    self.interrupt(c)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 404, in interrupt
    reactivate(victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 597, in reactivate
    a.cancel(obj)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 380, in cancel
    _e._unpost(whom=victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 524, in _unpost
    thislist=self.events[thistime]
KeyError: 3661.0

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/
_______________________________________________
Simpy-users mailing list
Simpy-users <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/simpy-users

José,

your problem is caused by a simple programming error: you don't reset the class attribute "TrafficCoils.coilsList" to [ ] at the beginning of each run. Thus, processes (i.e., instances of class "Coils") stay in that list after a run. (I suspect, that you have to reset "Leak.leakTotal" to 0, too. But I am  not sure, maybe you want to count "leakTotal" acrosss runs. Does not do any harm, anyhow. And what about "Coils.coilsTotal"? Does this refer to a total for one run, or all? )

 

Another proposal: you don't need to call "stopSimulation()" after coming out of "simulate(. . .)". Does not do any harm, but is unnecessary.

 

The script is easily repaired by adding the resetting code to the "run" loop like so:

 

. . . . .

for run in range(RUNS):
    print "=== SIMULATION: %s"\
        %(run)
    initialize()
## Begin KGM repair
    TrafficCoils.coilsList = []
## End KGM repair
    decoking=Resource(capacity=numberCoilsPerHeater)
    trafficC=TrafficCoils()
    activate(trafficC, trafficC.generateCycles ())
    trafficL=TrafficLeaks()
    activate(trafficL,trafficL.generateLeaks())
    simulate(until=10*365)    # days

## Begin KGM repair
    #stopSimulation() 

## End KGM repair
    print (Coils.coilsTotal)

 

An interesting application and nice code. Thanks for sharing this with us!

 

Keep posting on the list, please!

 

Klaus Muller

From: simpy-users-bounces <at> lists.sourceforge.net [mailto:simpy-users-bounces <at> lists.sourceforge.net ] On Behalf Of Jose Trejo-Meza
Sent: Thursday, September 06, 2007 12:51 AM
To: Tony Vignaux
Cc: simpy-users <at> lists.sourceforge.net
Subject: Re: [Simpy-users] KeyError on interruption

Tony,

Thanks for replying. My program can be found below. I performed a debugging step through. The keyError is being generated by a key (3661, event time) of the victim that is being "_unpost" (??) and that happens to be past the simulation end time (3650), when interrupting the victim (a Coils object) from the Leak process.

-José

from SimPy.Simulation import Process, Resource, now, initialize, activate, simulate, hold, request, release, stopSimulation
from random import Random, expovariate, weibullvariate

numberCoilsPerHeater=4
numberHeaters=9
productionRate=48
initDecokeCycle=60.0
decokeThreshold=15
cycleReduction=0.50
#decokingTime=1
decokingTime=1 #Testing
RUNS=10
   
class Coils(Process):
    """Coils subject to failure"""
    coilsTotal=0
   
    def __init__(self):
        Coils.coilsTotal+=1
        self.coilNum=Coils.coilsTotal
        Process.__init__(self)
        self.id="Coil " + str(Coils.coilsTotal)

    def decokeTime(self):
        return self.dt    # hr
   
    def decokeCycle(self):
        if now()==0:
            self.dc=initDecokeCycle
        elif self.dc<=decokeThreshold:
            self.dc=initDecokeCycle
        else:
            self.dc-=cycleReduction
        return self.dc
   
    def leakWhileWorking(self):
        self.dc = initDecokeCycle
        yield hold,self,self.dc
   
    def leakWhileDecoking(self):
        self.dc = initDecokeCycle
   
    def PEM(self):
        while True:
            arrive=now()
            yield hold,self,self.decokeCycle ()
            if self.interrupted():
                print "*** %s %s leak interruption Time left in cycle: %s"\
                    %(now(), self.id,self.interruptLeft)
                #self.leakWhileWorking()
                self.dc=initDecokeCycle
                self.dt=self.dt-self.dt # No decoking
                yield hold,self,self.dc+Leak.leakReplaceTime
                self.interruptReset()
            #print "%s %s enters."\
            #    %(arrive,self.id)
            yield request,self,decoking    # get a helpdesk line
            atlast=now()
            yield hold,self,self.decokeTime()    # get service
            self.dt = decokingTime
            if self.interrupted():
                print "*** %s %s leak interruption Time left in decoking: %s"\
                    %(now(), self.id, self.interruptLeft)
                self.dc = initDecokeCycle
                yield hold,self,Leak.leakReplaceTime
                self.interruptReset()
                #self.leakWhileDecoking()
            yield release,self,decoking    # hang up line
            cycle=now()-self.cycleStart
            self.cycles+=1
            print "%s %s leaves. Total cycle time =%s. Cycle Num=%s"\
                %(now(),self.id,cycle,self.cycles)
            self.cycleStart=now()

class Leak(Process):
    leakTotal=0
    leakReplaceTime=3
   
    def __init__(self):
        Process.__init__(self)
        Leak.leakTotal+=1
        self.coilNum=Leak.leakTotal
        self.id="Leak " + str( Leak.leakTotal)
        print "%s %s created."\
            %(now(), self.id)
   
    def PEM(self):
        n=0
        failureTime=now()
        for c in TrafficCoils.coilsList:
            self.interrupt(c)
        Leak.leakTotal += 1
        yield hold,self,Leak.leakReplaceTime

class TrafficCoils(Process):
    """Generates Coil decoke cycles"""
    arrivalRate=1/initDecokeCycle    # (per hour) arrival rate
    rv=Random(4567)    # a random number generator
    coilsList = []
   
    def __init__(self):
        Process.__init__(self)
       
    def generateCycles(self):
        n=0
        timeToNext=0
        while n<numberCoilsPerHeater:
            coil=Coils()    # new customer call
            TrafficCoils.coilsList.append(coil)
            coil.cycles=0
            coil.cycleStart=now()
            coil.dc=initDecokeCycle
            coil.dt=decokingTime
            yield hold,self,timeToNext
            timeToNext=round( TrafficCoils.rv.expovariate(TrafficCoils.arrivalRate),0)
            activate(coil,coil.PEM())
            n+=1
           
class TrafficLeaks(Process):
    """Generates Leaks (failures) in the coils"""
    rv=Random(4567)
    Walpha=63.38*30
    Wbeta=5.79
   
    def __init__(self):
        Process.__init__(self)

    def generateLeaks(self):
        while True:
            leak=Leak()
            leak.cycles=0
            leak.cycleStart=now()
            timeToNext=round(TrafficLeaks.rv.weibullvariate(TrafficLeaks.Walpha, TrafficLeaks.Wbeta),0)
            yield hold,self,timeToNext
            activate(leak,leak.PEM())

for run in range(RUNS):
    print "=== SIMULATION: %s"\
        %(run)
    initialize()
    decoking=Resource(capacity=numberCoilsPerHeater)
    trafficC=TrafficCoils()
    activate(trafficC, trafficC.generateCycles ())
    trafficL=TrafficLeaks()
    activate(trafficL,trafficL.generateLeaks())
    simulate(until=10*365)    # days
    stopSimulation()
    print (Coils.coilsTotal)

      
      

On 9/5/07, Tony Vignaux <vignaux <at> gmail.com> wrote:

I suggest you let us have a listing of the program so we can try and repeat the error.

On 9/6/07, Jose Trejo-Meza < jl.trejo <at> gmail.com> wrote:

Hello,

I'm working on a failure model for a process production system and ran into the exception below. My model is generating two kinds of Process objects where one interrupts the other. The model generates 4 Coil objects (the interrupted) and as many Leak objects (the interrupter) as simulation time allow, usually a few. I think an instance of Leak is causing the KeyError, but I don't know how it is causing it. This happens only when I put the simulation into a loop (initialize(), simulate(), etc.), and always on the second simulation (using fixed seed in the random number generator). I searched the mailing list archives, but there is not much about KeyError's. Does anyone have an idea of what's causing this?

-José


1555.0 Coil 8 leaves. Total cycle time =47.0. Cycle Num=28
1556.5 Coil 6 leaves. Total cycle time =46.5. Cycle Num=29
1566.0 Coil 5 leaves. Total cycle time =47.0. Cycle Num=29
1584.0 Leak 5 created.
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 145, in <module>
    simulate(until=10*365)    # days
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 2009, in simulate
    a=_e._nextev()         
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 554, in _nextev
    tt=tempwho._nextpoint.next()
  File "C:\lang\Workspace\Heaters\src\mainSim.py", line 91, in PEM
    self.interrupt(c)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 404, in interrupt
    reactivate(victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 597, in reactivate
    a.cancel(obj)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 380, in cancel
    _e._unpost(whom=victim)
  File "C:\lang\Python25\lib\site-packages\SimPy\Simulation.py", line 524, in _unpost
    thislist=self.events[thistime]
KeyError: 3661.0

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >>  http://get.splunk.com/
_______________________________________________
Simpy-users mailing list
Simpy-users <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/simpy-users

This posting seeks your feedback on a proposal to add to the SimPy API.

Tony Vignaux and I propose to add a user-callable method "iactivate" to
class Process which will allow a simplified/shortened form of process
activation. This would be in addition to the "activate" function.

Here is the proposal:

"""Another way of activating a process, p is the iactivate Process method
attribute.

For a subclass which has a generator function (containing a yield) called
"PEM", this can be called in the
following ways:
• p.iactivate(p.PEM([args]) [,{at=t|delay=period}] [,prior=False])
if p is a Process. Here the generator function, PEM, can have any identifier
(such as run, lifecycle, etc). It
can have parameters.
• p.iactivate([p.PEM()] [,{at=t|delay=period}] [,prior=False])
if p is a Process and the generator function is given the standard
identifier, PEM. The identifier, PEM, is
recognised as a generator method but may not have parameters. Note that the
call p.PEM() is optional.
Thus a valid call would be p.iactivate() to activate the generator PEM at
time now without attributes.
An anonymous Process object can be created and activated in one command
using iactivate using the standard
PEM identifier, PEM.
• P([args]).iactivate([{at=t|delay=period}] [,prior=False])
Here, P is the identifier for the Process and not for a Process object as
was p, in the previous statements,
above. The generator method with the standard identifier, PEM, may not have
parameters.

iactivate can also be used if the name of the generator function is
different. It can then be called like this:

• p.iactivate(p.genname([args][,{at=t|delay=period}] [,prior=False])
"""
(end of proposal)

What we would like to hear from you is:
-- what do you think of the proposal?
-- do you think that another name instead of "PEM" is more appropriate for
the generator function, e.g. "lifecycle"?
-- do you think that the programming user who wants to use the short form of
iactivate should be able to select a generator function name from a list
with several default names?

Please, let us have your feedback!

Klaus Müller


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Simpy-developer mailing list
Simpy-developer <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/simpy-developer