Presence of stale DHCP directories on dhcp-agent node

Bug #1195770 reported by Armando Migliaccio
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
neutron
Fix Released
Low
Armando Migliaccio

Bug Description

This happens on master with the default settings for the dhcp agent.

Currently there is a case where dnsmasq directories under /opt/stack/quantum/dhcp or /var/lib/quantum/dhcp remain undeleted. This has two potentials issues: a) we might hit a limit on existing directories (even though I doubt we'll ever likely run into that), and b) this needless slows down the bootstrap of the dhcp agent, which has verify the activity of a dnsmasq agent even if their associated networks no longer exist.

For more context, see:

https://review.openstack.org/#/c/31950/

Tags: l3-ipam-dhcp
tags: added: l3-ipam-dhcp
Changed in neutron:
status: New → Confirmed
Changed in neutron:
assignee: nobody → Armando Migliaccio (armando-migliaccio)
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to neutron (master)

Fix proposed to branch: master
Review: https://review.openstack.org/46122

Changed in neutron:
status: Confirmed → In Progress
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to neutron (master)

Reviewed: https://review.openstack.org/46122
Committed: http://github.com/openstack/neutron/commit/eea6218e425766cefc62757da85b9de058d8e312
Submitter: Jenkins
Branch: master

commit eea6218e425766cefc62757da85b9de058d8e312
Author: armando-migliaccio <email address hidden>
Date: Wed Sep 11 11:49:48 2013 -0700

    Prevent stale DHCP directories for dhcp agent nodes

    This change fixes an issue that is most acute in test/dev
    environments but that may affect production environments
    as well: the presence of DHCP directories that no longer
    represent networks that exist in the Neutron DB. This
    issue can manifest itself over time if you delete networks
    from the Server while the Agent node is down.

    Without this fix, at the agent start-up the method
    existing_dhcp_networks will return an empty list, and
    as a consequence the directories that belong to networks
    that have been removed from the server will stay because
    the sync logic will not process them as it only looks at
    the mismatch between active networks on the agent and
    active networks on the server. With the fix instead, we
    return *all* existing dhcp networks; if they are meant to
    be active, the agent will bring them up, if they no longer
    exist on the server, then the agent will dispose of the
    resources. The dnsmasq driver will do the right thing when
    enabling or disabling the process.

    Fixes bug #1195770

    Change-Id: I194064a449801713051d01193adc706bcb687c82

Changed in neutron:
status: In Progress → Fix Committed
Changed in neutron:
importance: Undecided → Low
milestone: none → havana-rc1
Thierry Carrez (ttx)
Changed in neutron:
status: Fix Committed → Fix Released
Thierry Carrez (ttx)
Changed in neutron:
milestone: havana-rc1 → 2013.2
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.