yahoo-eng-team team mailing list archive

[Yair Fried <yfried@xxxxxxxxxx>]

Public bug reported:

When running Tempest from a node outside the Openstack cloud, tests tend to use the same Floating IP. The time between tests is so short that the external switch still has the old MAC (of port? VM?) in cache and therefore can't connect to the new setup.
This is only solved when the new VM tries to access external network and updates the switch DB with the new MAC.

I don't know how this can be fixed outside a change to neutron such as to:
1. change Floating IP scheduling (round-robin?
2. Force gratuitous ARP upon Floating IP association.

another option is to implement  support to ip-netns which will allow
more flexibility to network operations

How to reproduce:
1. setup Tempest on different node than your Openstack cloud (specifically L3 agent)
2. run 2 tests that use Floating IPs connection (such as network_basic_ops and cross_tenant)

Result:
Second test will fail on connectivity issus (floating ip unreachable)

Expected Result:
Both tests should pass

** Affects: neutron
     Importance: Undecided
         Status: New

** Affects: tempest
     Importance: Undecided
         Status: New

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

** Description changed:

  When running Tempest from a node outside the Openstack cloud, tests tend to use the same Floating IP. The time between tests is so short that the external switch still has the old MAC (of port? VM?) in cache and therefore can't connect to the new setup.
  This is only solved when the new VM tries to access external network and updates the switch DB with the new MAC.
  
- I don't know how this can be fixed outside a change to neutron such as
- to change Floating IP scheduling (round-robin?)
+ I don't know how this can be fixed outside a change to neutron such as to:
+ 1. change Floating IP scheduling (round-robin?
+ 2. Force gratuitous ARP upon Floating IP association.
  
  another option is to implement  support to ip-netns which will allow
  more flexibility to network operations
  
  How to reproduce:
  1. setup Tempest on different node than your Openstack cloud (specifically L3 agent)
  2. run 2 tests that use Floating IPs connection (such as network_basic_ops and cross_tenant)
  
  Result:
  Second test will fail on connectivity issus (floating ip unreachable)
  
  Expected Result:
  Both tests should pass

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1268604

Title:
  multiple tests using floating-ips fail on remote testing

Status in OpenStack Neutron (virtual network service):
  New
Status in Tempest:
  New

Bug description:
  When running Tempest from a node outside the Openstack cloud, tests tend to use the same Floating IP. The time between tests is so short that the external switch still has the old MAC (of port? VM?) in cache and therefore can't connect to the new setup.
  This is only solved when the new VM tries to access external network and updates the switch DB with the new MAC.

  I don't know how this can be fixed outside a change to neutron such as to:
  1. change Floating IP scheduling (round-robin?
  2. Force gratuitous ARP upon Floating IP association.

  another option is to implement  support to ip-netns which will allow
  more flexibility to network operations

  How to reproduce:
  1. setup Tempest on different node than your Openstack cloud (specifically L3 agent)
  2. run 2 tests that use Floating IPs connection (such as network_basic_ops and cross_tenant)

  Result:
  Second test will fail on connectivity issus (floating ip unreachable)

  Expected Result:
  Both tests should pass

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1268604/+subscriptions