Custom python process killed when chaning switch's configuration

DanielDelgado — Fri, 21 Feb 2020 20:54:00 GMT

Hi community.

#!/usr/bin/python

'''This script create a socket in EXOS that listen arp paquets at data plane and send that ARP request to the vlan “foo”.

We have “silent hosts” that we need to answer ARP request in the vlan foo. Once the silent host send the ARP reply, we can assignate a role and vlan using Identify Manager.
'''

import sys, logging
from exos.api import expkt as socket

def main():
    if not hasattr(sys, 'expy') or not sys.expy:
        print "Must be run within expy"
        return

    sock = socket.socket(socket.PF_EXOS_PACKET, socket.SOCK_RAW, socket.htons(0x0806))
    sock.setup_filter("filter", action=socket.UP_AND_CONTINUE, ethertype=0x0806)

    while True :
        pkt, add = sock.recvfrom(8192)
        sock.sendto(pkt, vlan_name="foo")

try:
    main()
except BaseException, e:
    print ("Exception on startup, {}".format(e))

I am testing with an X460G2-24t-G4 , 21.1.5.2 patch1-7 and 30.5.1.15 software version, both of them has this issue I will describe.

I create the “arpClone” process:

* TEST_LAB.69 # create process arpClone python-module arpClone start auto vr VR-Default

* TEST_LAB.69 # show process "arpClone"

Process Name Version Restart State Start Time

-------------------------------------------------------------------------

arpClone User 0 LoadCfg Fri Feb 21 13:47:47 2020

* TEST_LAB.70 # debug ems show trace "arpClone" all

02/21/2020 13:47:48.589265 [1] <arpClone:ipml> Created service group 1

02/21/2020 13:47:48.589734 [2] <arpClone:ipml> Added service 0x41aff0 to group 1

02/21/2020 13:47:48.589773 [3] <arpClone:ipml> Added service 0x41aee0 to group 1

…...

02/21/2020 13:47:48.747425 [99] <arpClone:dm> Added Object: upm, ID 50000022, Index 34, State READY, Flags 0000, (nil), LEN 104

...

The process start ok, is traced , and send the ARP request via socket to the vlan “foo”.

So far so good!

But there is a problem…

This process is killed, after any change of configuration’s switch. If create a vlan, for example:

* TEST_LAB.71 # create vlan 45

(the prompt is lost about 5 seconds..)

* TEST_LAB.72 # debug ems show trace "arpClone" all

ERROR: Process "arpClone" did not respond in time with trace information.

At this moment, process is still working, the socket is sending ARP request to the vlan “foo”.

But after about 6 minutes, the process is killed and I can show this log:

* TEST_LAB.72 # debug ems show trace "arpClone" all

ERROR: Process "arpClone" did not respond in time with trace information.

* TEST_LAB.74 # show process "arpClone"

Process Name Version Restart State Start Time

-------------------------------------------------------------------------

arpClone User 0 LoadCfg Fri Feb 21 13:47:47 2020

02/21/2020 10:52:47.26 <Warn:cm.database> Configuration database unlocked

02/21/2020 10:52:47.25 <Erro:DM.Error> Process arpClone Failed

02/21/2020 10:52:47.24 <Warn:cm.database> Configuration database locked

02/21/2020 10:52:47.24 <Erro:EPM.proc_conn_lost> Connection lost with process arpClone

02/21/2020 10:52:45.76 <Crit:Kern.Alert> 776cc15c 00000000

...

02/21/2020 10:52:45.50 <Crit:Kern.Alert> 776cc13c 00000000

02/21/2020 10:52:45.49 <Crit:Kern.Alert> Code:

02/21/2020 10:52:45.49 <Crit:Kern.Alert>

02/21/2020 10:52:45.49 <Crit:Kern.Alert> Process expy pid 4203 died with signal 6

02/21/2020 10:52:45.40 <Warn:EPM.proc_kill> Process arpClone ID 4203 killed

02/21/2020 10:52:45.40 <Erro:EPM.Msg.timer_thread> Because the main (2012753728) thread of process 4203, has not responded within 41 periods of 10 seconds, the process will be terminated.

How may I solve this? Did I forget to configure anything about process to refresh process’s information when changing the config?

Any one can help me?

Thanks and best regards.

Daniel Delgado,

topic Custom python process killed when chaning switch's configuration in ExtremeSwitching (EXOS/Switch Engine)

Custom python process killed when chaning switch's configuration