yahoo-eng-team team mailing list archive

[Alan Pevec <1457900@xxxxxxxxxxxxxxxxxx>]

** Also affects: neutron/juno
   Importance: Undecided
       Status: New

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1457900

Title:
  dhcp_agents_per_network > 1 cause conflicts (NACKs) from dnsmasqs
  (break networks)

Status in neutron:
  Fix Released
Status in neutron icehouse series:
  Fix Released
Status in neutron juno series:
  New
Status in neutron kilo series:
  New

Bug description:
  If neutron was configured to have more than one DHCP agent per network
  (option dhcp_agents_per_network=2), it causes dnsmasq to reject leases
  of others dnsmasqs, creating mess and stopping instances to boot
  normally.

  Symptoms:

  Cirros (at the log):
  Sending discover...
  Sending select for 188.42.216.146...
  Received DHCP NAK
  Usage: /sbin/cirros-dhcpc <up|down>
  Sending discover...
  Sending select for 188.42.216.146...
  Received DHCP NAK
  Usage: /sbin/cirros-dhcpc <up|down>
  Sending discover...
  Sending select for 188.42.216.146...
  Received DHCP NAK

  Steps to reproduce:
  1. Set up neutron with VLANs and dhcp_agents_per_network=2 option in neutron.conf
  2. Set up two or more different nodes with enabled neutron-dhcp-agent
  3. Create VLAN neutron network with --enable-dhcp option
  4. Create instance with that network

  Expected behaviour:

  Instance recieve IP address via DHCP without problems or delays.

  Actual behaviour:

  Instance stuck in the network boot for long time.
  There are complains about NACKs in the logs of dhcp client.
  There are multiple NACKs on tcpdump on interfaces

  Additional analysis: It is very complex, so I attach example of two
  parallel tcpdumps from two dhcp namespaces in HTML format.

  
  Version: 2014.2.3

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1457900/+subscriptions