yahoo-eng-team team mailing list archive

[Rodolfo Alonso <2092983@xxxxxxxxxxxxxxxxxx>]

Hello zhuyudong:

The code is not the same at all. During the years between Train release
and master (Epoxy), we have improved the queries (subnets in
particular), improved the subnets operations (e.g.: [1]), we have
implemented a new DB engine facade [2], bumped the SQLAlchemy mayor
version (2.0) and many other minor fixes and improvements that cannot be
related in one single message.

I've tested (with master branch) your script (adapted to OSC commands)
and I cannot reproduce the issue, even increasing the sleep time in the
Neutron code.

I'm marking this bug as invalid. Please update your environment to a
newer (and supported) version.

Regards.

[1]https://review.opendev.org/c/openstack/neutron/+/713045
[2]https://review.opendev.org/q/topic:%22bp/enginefacade-switch%22


** Changed in: neutron
       Status: Incomplete => Invalid

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/2092983

Title:
  deadlock： delete the network

Status in neutron:
  Invalid

Bug description:
  When a network is deleted, the port is deleted first and then the
  subnet is deleted. When a port is deleted, neutron-dhcp-agent is
  notified. The agent code goes to self.schedule_resync("Agent port was
  deleted", port.network_id), which causes sync_state to be performed on
  the network. If the subnet is not deleted, sync_state may deactivate
  DHCP. In this case, DHCP port creation and subnet deletion are
  performed at the same time, triggering deadlock.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/2092983/+subscriptions