L3 agent: at some point an agent becomes unable to handle new routers

Bug #1566291 reported by Oleg Bondarev
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
neutron
Fix Released
High
Oleg Bondarev

Bug Description

Following seen in l3 agent logs:

2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent [-] Failed to process compatible router 'e341e0e2-5089-46e9-91f9-2099a156b27f'
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent Traceback (most recent call last):
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/agent.py", line 497, in _process_router_update
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent self._process_router_if_compatible(router)
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/agent.py", line 434, in _process_router_if_compatible
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent self._process_added_router(router)
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/agent.py", line 439, in _process_added_router
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent self._router_added(router['id'], router)
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/agent.py", line 340, in _router_added
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent ri = self._create_router(router_id, router)
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/agent.py", line 337, in _create_router
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent return legacy_router.LegacyRouter(*args, **kwargs)
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent File "/usr/lib/python2.7/dist-packages/neutron/agent/l3/router_info.py", line 61, in __init__
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent DEFAULT_ADDRESS_SCOPE: ADDRESS_SCOPE_MARK_IDS.pop()}
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent KeyError: 'pop from an empty set'
2016-04-05 09:30:09.033 24216 ERROR neutron.agent.l3.agent

So agent is constantly resyncing (causing load on neutron server) and unable to handle new routers.

I believe that set "ADDRESS_SCOPE_MARK_IDS = set(range(1024, 2048))" from router_info.py should not be agent global but it should be ADDRESS_SCOPE_MARK_IDS per router. Or at least need to return values back to the set when router is deleted.

Assaf Muller (amuller)
Changed in neutron:
importance: Undecided → High
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/301685

Changed in neutron:
status: New → In Progress
description: updated
tags: added: mitaka-backport-potential
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to neutron (stable/mitaka)

Fix proposed to branch: stable/mitaka
Review: https://review.openstack.org/303439

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to neutron (master)

Reviewed: https://review.openstack.org/301685
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=1cb43734808eded87210d2957d56b70c514d55c3
Submitter: Jenkins
Branch: master

commit 1cb43734808eded87210d2957d56b70c514d55c3
Author: Oleg Bondarev <email address hidden>
Date: Tue Apr 5 16:18:03 2016 +0300

    ADDRESS_SCOPE_MARK_IDS should not be global for L3 agent

    Otherwise agent becomes unable to handle more than 1024 routers
    (including deleted routers) and starts failing.

    It should be enough to distinguish address scopes inside router namespace,
    so this patch moves ADDRESS_SCOPE_MARK_IDS set to the RouterInfo class.

    Closes-Bug: #1566291
    Change-Id: I1e43bb3e68db4db93cc1dfc1383af0311bfb0f2d

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

Reviewed: https://review.openstack.org/303439
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=07fa3725c5a7fc68a41ed8af53ca2d3aad4c35b9
Submitter: Jenkins
Branch: stable/mitaka

commit 07fa3725c5a7fc68a41ed8af53ca2d3aad4c35b9
Author: Oleg Bondarev <email address hidden>
Date: Tue Apr 5 16:18:03 2016 +0300

    ADDRESS_SCOPE_MARK_IDS should not be global for L3 agent

    Otherwise agent becomes unable to handle more than 1024 routers
    (including deleted routers) and starts failing.

    It should be enough to distinguish address scopes inside router namespace,
    so this patch moves ADDRESS_SCOPE_MARK_IDS set to the RouterInfo class.

    Closes-Bug: #1566291
    Change-Id: I1e43bb3e68db4db93cc1dfc1383af0311bfb0f2d
    (cherry picked from commit 1cb43734808eded87210d2957d56b70c514d55c3)

tags: added: in-stable-mitaka
Revision history for this message
Doug Hellmann (doug-hellmann) wrote : Fix included in openstack/neutron 8.1.0

This issue was fixed in the openstack/neutron 8.1.0 release.

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/314250

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to neutron (master)
Download full text (36.9 KiB)

Reviewed: https://review.openstack.org/314250
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=3bf73801df169de40d365e6240e045266392ca63
Submitter: Jenkins
Branch: master

commit a323769143001d67fd1b3b4ba294e59accd09e0e
Author: Ryan Moats <email address hidden>
Date: Tue Oct 20 15:51:37 2015 +0000

    Revert "Improve performance of ensure_namespace"

    This reverts commit 81823e86328e62850a89aef9f0b609bfc0a6dacd.

    Unneeded optimization: this commit only improves execution
    time on the order of milliseconds, which is less than 1% of
    the total router update execution time at the network node.

    This also

    Closes-bug: #1574881

    Change-Id: Icbcdf4725ba7d2e743bb6761c9799ae436bd953b

commit 7fcf0253246832300f13b0aa4cea397215700572
Author: OpenStack Proposal Bot <email address hidden>
Date: Thu Apr 21 07:05:16 2016 +0000

    Imported Translations from Zanata

    For more information about this automatic import see:
    https://wiki.openstack.org/wiki/Translations/Infrastructure

    Change-Id: I9e930750dde85a9beb0b6f85eeea8a0962d3e020

commit 643b4431606421b09d05eb0ccde130adbf88df64
Author: OpenStack Proposal Bot <email address hidden>
Date: Tue Apr 19 06:52:48 2016 +0000

    Imported Translations from Zanata

    For more information about this automatic import see:
    https://wiki.openstack.org/wiki/Translations/Infrastructure

    Change-Id: I52d7460b3265b5460b9089e1cc58624640dc7230

commit 1ffea42ccdc14b7a6162c1895bd8f2aae48d5dae
Author: OpenStack Proposal Bot <email address hidden>
Date: Mon Apr 18 15:03:30 2016 +0000

    Updated from global requirements

    Change-Id: Icb27945b3f222af1d9ab2b62bf2169d82b6ae26c

commit b970ed5bdac60c0fa227f2fddaa9b842ba4f51a7
Author: Kevin Benton <email address hidden>
Date: Fri Apr 8 17:52:14 2016 -0700

    Clear DVR MAC on last agent deletion from host

    Once all agents are deleted from a host, the DVR MAC generated
    for that host should be deleted as well to prevent a buildup of
    pointless flows generated in the OVS agent for hosts that don't
    exist.

    Closes-Bug: #1568206
    Change-Id: I51e736aa0431980a595ecf810f148ca62d990d20
    (cherry picked from commit 92527c2de2afaf4862fddc101143e4d02858924d)

commit eee9e58ed258a48c69effef121f55fdaa5b68bd6
Author: Mike Bayer <email address hidden>
Date: Tue Feb 9 13:10:57 2016 -0500

    Add an option for WSGI pool size

    Neutron currently hardcodes the number of
    greenlets used to process requests in a process to 1000.
    As detailed in
    http://lists.openstack.org/pipermail/openstack-dev/2015-December/082717.html

    this can cause requests to wait within one process
    for available database connection while other processes
    remain available.

    By adding a wsgi_default_pool_size option functionally
    identical to that of Nova, we can lower the number of
    greenlets per process to be more in line with a typical
    max database connection pool size.

    DocImpact: a previously unused configuration value
               wsgi_default_pool_size is now used to a...

tags: added: neutron-proactive-backport-potential
Revision history for this message
Doug Hellmann (doug-hellmann) wrote : Fix included in openstack/neutron 9.0.0.0b1

This issue was fixed in the openstack/neutron 9.0.0.0b1 development milestone.

Revision history for this message
Brian Haley (brian-haley) wrote :

We should be all set with backports here as the code this was fixing was not in Liberty, just Mitaka, and that change was done.

tags: removed: mitaka-backport-potential neutron-proactive-backport-potential
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.