neutron_dhcp_agent accumulates zombie processes

[Vladislav Belogrudov]

while adding / deleting networks in Queens release of Kolla I can see many zombie processes, the number only increases in time

e.g.

(neutron-dhcp-agent)[root@network9 /]# ps aux | grep defunct
root 146 0.0 0.0 0 0 ? Zs 12:31 0:00 [dnsmasq] <defunct>
nobody 147 0.0 0.0 0 0 ? Z 12:31 0:00 [dnsmasq] <defunct>
neutron 153 0.0 0.0 0 0 ? Zs 12:31 0:00 [haproxy] <defunct>
root 281 0.0 0.0 0 0 ? Zs 12:31 0:00 [dnsmasq] <defunct>
nobody 282 0.0 0.0 0 0 ? Z 12:31 0:00 [dnsmasq] <defunct>
neutron 288 0.0 0.0 0 0 ? Zs 12:31 0:00 [haproxy] <defunct>
root 416 0.0 0.0 0 0 ? Zs 12:31 0:00 [dnsmasq] <defunct>
nobody 417 0.0 0.0 0 0 ? Z 12:31 0:00 [dnsmasq] <defunct>
neutron 423 0.0 0.0 0 0 ? Zs 12:31 0:00 [haproxy] <defunct>
root 2732 0.0 0.0 0 0 ? Zs 13:14 0:00 [dnsmasq] <defunct>
nobody 2733 0.0 0.0 0 0 ? Z 13:14 0:00 [dnsmasq] <defunct>
root 3017 0.0 0.0 0 0 ? Zs 13:14 0:00 [dnsmasq] <defunct>
nobody 3018 0.0 0.0 0 0 ? Z 13:14 0:00 [dnsmasq] <defunct>
root 3224 0.0 0.0 0 0 ? Zs 13:17 0:00 [dnsmasq] <defunct>
nobody 3225 0.0 0.0 0 0 ? Z 13:17 0:00 [dnsmasq] <defunct>
neutron 3231 0.0 0.0 0 0 ? Zs 13:17 0:00 [haproxy] <defunct>
root 3359 0.0 0.0 0 0 ? Zs 13:17 0:00 [dnsmasq] <defunct>
nobody 3360 0.0 0.0 0 0 ? Z 13:17 0:00 [dnsmasq] <defunct>
neutron 3366 0.0 0.0 0 0 ? Zs 13:17 0:00 [haproxy] <defunct>
root 3494 0.0 0.0 0 0 ? Zs 13:17 0:00 [dnsmasq] <defunct>
nobody 3495 0.0 0.0 0 0 ? Z 13:17 0:00 [dnsmasq] <defunct>
neutron 3501 0.0 0.0 0 0 ? Zs 13:17 0:00 [haproxy] <defunct>
root 4906 0.0 0.0 0 0 ? Zs 13:36 0:00 [dnsmasq] <defunct>
nobody 4907 0.0 0.0 0 0 ? Z 13:36 0:00 [dnsmasq] <defunct>
root 5189 0.0 0.0 0 0 ? Zs 13:36 0:00 [dnsmasq] <defunct>
nobody 5190 0.0 0.0 0 0 ? Z 13:36 0:00 [dnsmasq] <defunct>
root 5207 0.0 0.0 0 0 ? Zs 13:36 0:00 [dnsmasq] <defunct>
nobody 5208 0.0 0.0 0 0 ? Z 13:36 0:00 [dnsmasq] <defunct>

[Vladislav Belogrudov (vlad-belogrudov)]

i restarted dhcp container and it started to increase number of zombies without stopping:

(neutron-dhcp-agent)[root@network9 /]# ps aux | awk '/defunct/ { print $2 }'
213

[Vladislav Belogrudov (vlad-belogrudov)]

ignore last message, after container restart zombies are gone, awk was catching itself

[Vladislav Belogrudov (vlad-belogrudov)]

to increase number of zombies I used the following loop:

while true; do
    openstack network create --provider-network-type vxlan --share test_net
    openstack subnet create --network test_net --subnet-range 192.0.0.0/24 test_subnet
    openstack network delete test_net
done

after complete clean-up of networks zombies are still there. I also use l3 ha if this matters

[Vladislav Belogrudov (vlad-belogrudov)]

dumb-init was missing in my images