Bug #1454738 “Nexus VLAN config hangs when replay enabled on RHE...” : Series kilo : Bugs : networking-cisco

Carol Bouchard (caboucha) on 2015-05-13

Changed in networking-cisco:
assignee:	nobody → Carol Bouchard (caboucha)
tags:	added: cisco ml2

Rich Curran (rcurran) on 2015-06-05

Changed in networking-cisco:
assignee:	Carol Bouchard (caboucha) → Rich Curran (rcurran)

Leon Zachery (lzachery) on 2015-06-05

tags:

added: e-rel

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-06-05: Fix proposed to networking-cisco (master)

#1

Fix proposed to branch: master
Review: https://review.openstack.org/188891

Changed in networking-cisco:
status:	New → In Progress

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-06-10: Fix merged to networking-cisco (master)

#2

Reviewed: https://review.openstack.org/188891
Committed: https://git.openstack.org/cgit/stackforge/networking-cisco/commit/?id=347073a248ecd844f259c6ef821c5cc27d1b07d1
Submitter: Jenkins
Branch: master

commit 347073a248ecd844f259c6ef821c5cc27d1b07d1
Author: Rich Curran <email address hidden>
Date: Fri Jun 5 15:00:14 2015 -0400

ML2 cisco_nexus MD: Config hangs when replay enabled

When multiple neutron-server processes are created (rpc_workers > 0)
the cisco_nexus monitor_thread was hanging.

    Similar issues seen with other MDs. More info:
    http://lists.openstack.org/pipermail/openstack-dev/2015-May/063515.html
    http://lists.openstack.org/pipermail/openstack-dev/2015-June/065558.html
    https://bugs.launchpad.net/vmware-nsx/+bug/1420278

    09-Jun - Upstream community should be fixing this problem properly
    in near future - https://review.openstack.org/#/c/189391/1
    Once this fix is pushed up then a new bug can be used to update this
    patch to use the new method for avoiding these multi-process/thread
    issues.

Change-Id: I1cbf6a4f9bc2e795720c75aca10abd4d8d458434
Closes-Bug: #1454738

Changed in networking-cisco:
status:	In Progress → Fix Committed

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-06-15: Fix proposed to networking-cisco (master)

#3

Fix proposed to branch: master
Review: https://review.openstack.org/191823

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-06-23: Change abandoned on networking-cisco (master)

#4

Change abandoned by rcurran (<email address hidden>) on branch: master
Review: https://review.openstack.org/191823
Reason: submitted change under different patch
(this patch was not complete)

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-08-04: Fix proposed to networking-cisco (stable/kilo)

#5

Fix proposed to branch: stable/kilo
Review: https://review.openstack.org/209241

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-08-04:

#6

Fix proposed to branch: stable/kilo
Review: https://review.openstack.org/209248

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-08-05: Fix merged to networking-cisco (stable/kilo)

#7

Reviewed: https://review.openstack.org/209248
Committed: https://git.openstack.org/cgit/openstack/networking-cisco/commit/?id=ffabc773febb9a8df7853588ae27a4fe3bc4069b
Submitter: Jenkins
Branch: stable/kilo

commit ffabc773febb9a8df7853588ae27a4fe3bc4069b
Author: Rich Curran <email address hidden>
Date: Sat Jun 20 13:38:27 2015 -0400

ML2 cisco_nexus MD: Multiprocess replay issue

    When api or rpc_workers are set > 0 replay is not working in multi-node setups.
    Issue is that access to nexus switches is to set to False (can't access) by
    default in the monitor code. In multiprocess configurations the monitor
    thread (thread code determines if switch is accessible) only runs in the
    main process (mech_cisco_nexus.py initialize() method is only called once,
    regardless of number of processes). In all other processes the "access
    switch" setting never gets set to True (can access).

    This patch allows all non-parent processes to continue event handler
    processing. Only the parent process (which starts the monitor thread) will
    verify if the switch is accessible during event processing.

    Change-Id: I637aba7615df2711fdf058e57800c28e3741dadb
    Partial-Bug: #1454738
    (cherry picked from commit 7d37c15e8fc0e20ad9e3b0606d7a2fc534bfdd4a)

tags:

added: in-stable-kilo

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-08-05:

#8

Reviewed: https://review.openstack.org/209241
Committed: https://git.openstack.org/cgit/openstack/networking-cisco/commit/?id=b60296644660303fb2341ca6495611621fc486e7
Submitter: Jenkins
Branch: stable/kilo

commit b60296644660303fb2341ca6495611621fc486e7
Author: Rich Curran <email address hidden>
Date: Fri Jun 5 15:00:14 2015 -0400

ML2 cisco_nexus MD: Config hangs when replay enabled

When multiple neutron-server processes are created (rpc_workers > 0)
the cisco_nexus monitor_thread was hanging.

    Similar issues seen with other MDs. More info:
    http://lists.openstack.org/pipermail/openstack-dev/2015-May/063515.html
    http://lists.openstack.org/pipermail/openstack-dev/2015-June/065558.html
    https://bugs.launchpad.net/vmware-nsx/+bug/1420278

    09-Jun - Upstream community should be fixing this problem properly
    in near future - https://review.openstack.org/#/c/189391/1
    Once this fix is pushed up then a new bug can be used to update this
    patch to use the new method for avoiding these multi-process/thread
    issues.

    Change-Id: I1cbf6a4f9bc2e795720c75aca10abd4d8d458434
    Closes-Bug: #1454738
    (cherry picked from commit 347073a248ecd844f259c6ef821c5cc27d1b07d1)

Sam Betts (sambetts) on 2015-11-04

Changed in networking-cisco:
milestone:	none → 1.1.0

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-11-17: Fix proposed to networking-cisco (master)

#9

Fix proposed to branch: master
Review: https://review.openstack.org/246547

Revision history for this message

OpenStack Infra (hudson-openstack) wrote on 2015-12-08: Fix merged to networking-cisco (master)

#10

Download full text (23.9 KiB)

Reviewed: https://review.openstack.org/246547
Committed: https://git.openstack.org/cgit/openstack/networking-cisco/commit/?id=7b1eb2b6d5e55563c084f60e44adc1d32706eb17
Submitter: Jenkins
Branch: master

commit d9b9a6421d7ff92e920ed21b01ebc7bf49e38bd6
Author: Sam Betts <email address hidden>
Date: Tue Sep 29 09:18:10 2015 +0100

Set default branch for stable/kilo

Change-Id: I31f51ff60f95639f459839f4c7d929d5ec7c458d

commit f08fb31f20c2d8cc1e6b71784cdfd9604895e16d
Author: Rich Curran <email address hidden>
Date: Thu Sep 3 13:23:52 2015 -0400

ML2 cisco_nexus MD: VLAN not created on switch

    As described in DE588,
    "With neutron multiworkers configured, there is a potential race condition
    issue where some of the VLANs will not be configured on one or more N9k
    switches.

    /etc/neutron/neutron.conf
    -------------------------
    api_workers=3
    rpc_workers=3"

    Fix is to allow the vlan create command to be sent down to a switch
    under most event conditions. Long term fix will be to introduce a new
    column in the port binding DB table that indicates the true state of the
    entry/row.

    Closes-Bug: #1491940
    Change-Id: If1da1fcf16a450c1a4107da9970b18fc64936896
    (cherry picked from commit 0e48a16e77fc5ec5fd485a85f97f3650126fb6fe)

commit d400749e43e9d5a1fc92683b40159afce81edc95
Author: Carol Bouchard <email address hidden>
Date: Thu Sep 3 15:19:48 2015 -0400

Create knob to prevent caching ssh connection

    Create a new initialization knob named never_cache_ssh_connection.
    This boolean is False by default allowing multiple ssh connections
    to the Nexus switch to be cached as it behaves today. When there
    are multiple neutron processes/controllers and/or non-neutron ssh(xml)
    connections, this is an issue since processes hold onto a connection
    while the Nexus devices supports a maximum of 8 sessions. As a result,
    further ssh connections will fail. In this case, the boolean should be
    set to True causing each connection to be closed when a neutron event
    is complete.

    Change-Id: I61ec303856b757dd8d9d43110fec8e7844ab7c6d
    Closes-bug: #1491108
    (cherry picked from commit 23551a4198c61e2e25a6382f27d47b0665f054b8)

commit 0050ea7f1fb3c22214d7ca49cfe641da86123e2c
Author: Carol Bouchard <email address hidden>
Date: Wed Sep 2 11:10:42 2015 -0400

Bubble up exceptions when Nexus replay enabled

There are several changes made surrounding this bug.

    1) When replay is enabled, we should bubble exceptions
       for received port create/update/delete post_commit
       transactions. This was suppressed earlier by
       1422738.

    2) When an exception is encountered during a
       post_commit transaction, the driver will no longer
       mark the switch state to inactive to force a replay.
       This is no longer needed since 1481856 was introduced.
       So from this point on, only the replay thread will
       determine the state of the connection to the switch.

3) In addition to accommodating 1 & 2 above, more detail
data verification was added to the test code.

Change-Id: I97...

Reviewed:  https://review.openstack.org/246547
Committed: https://git.openstack.org/cgit/openstack/networking-cisco/commit/?id=7b1eb2b6d5e55563c084f60e44adc1d32706eb17
Submitter: Jenkins
Branch:    master

commit d9b9a6421d7ff92e920ed21b01ebc7bf49e38bd6
Author: Sam Betts <sam@code-smash.net>
Date:   Tue Sep 29 09:18:10 2015 +0100

Set default branch for stable/kilo
    
    Change-Id: I31f51ff60f95639f459839f4c7d929d5ec7c458d

commit f08fb31f20c2d8cc1e6b71784cdfd9604895e16d
Author: Rich Curran <rcurran@cisco.com>
Date:   Thu Sep 3 13:23:52 2015 -0400

ML2 cisco_nexus MD: VLAN not created on switch
    
    As described in DE588,
    "With neutron multiworkers configured, there is a potential race condition
    issue where some of the VLANs will not be configured on one or more N9k
    switches.
    
    /etc/neutron/neutron.conf
    -------------------------
    api_workers=3
    rpc_workers=3"
    
    Fix is to allow the vlan create command to be sent down to a switch
    under most event conditions. Long term fix will be to introduce a new
    column in the port binding DB table that indicates the true state of the
    entry/row.
    
    Closes-Bug: #1491940
    Change-Id: If1da1fcf16a450c1a4107da9970b18fc64936896
    (cherry picked from commit 0e48a16e77fc5ec5fd485a85f97f3650126fb6fe)

commit d400749e43e9d5a1fc92683b40159afce81edc95
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Thu Sep 3 15:19:48 2015 -0400

Create knob to prevent caching ssh connection
    
    Create a new initialization knob named never_cache_ssh_connection.
    This boolean is False by default allowing multiple ssh connections
    to the Nexus switch to be cached as it behaves today.  When there
    are multiple neutron processes/controllers and/or non-neutron ssh(xml)
    connections, this is an issue since processes hold onto a connection
    while the Nexus devices supports a maximum of 8 sessions.  As a result,
    further ssh connections will fail.  In this case, the boolean should be
    set to True causing each connection to be closed when a neutron event
    is complete.
    
    Change-Id: I61ec303856b757dd8d9d43110fec8e7844ab7c6d
    Closes-bug:  #1491108
    (cherry picked from commit 23551a4198c61e2e25a6382f27d47b0665f054b8)

commit 0050ea7f1fb3c22214d7ca49cfe641da86123e2c
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Wed Sep 2 11:10:42 2015 -0400

Bubble up exceptions when Nexus replay enabled
    
    There are several changes made surrounding this bug.
    
    1) When replay is enabled, we should bubble exceptions
       for received port create/update/delete post_commit
       transactions.  This was suppressed earlier by
       1422738.
    
    2) When an exception is encountered during a
       post_commit transaction, the driver will no longer
       mark the switch state to inactive to force a replay.
       This is no longer needed since 1481856 was introduced.
       So from this point on, only the replay thread will
       determine the state of the connection to the switch.
    
    3) In addition to accommodating 1 & 2 above, more detail
       data verification was added to the test code.
    
    Change-Id: I97e4de2d5f56522eb0212cef4f804aa4a83957ec
    Closes-bug:  #1481859
    (cherry picked from commit 6b86a859db5fcf931df263b01e3bf1bd68f0797f)

commit 54fca8a047810304c69990dce03052e45f21cc23
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Tue Aug 11 13:14:17 2015 -0400

Quick retry connect to resolve stale ncclient handle
    
    When the Nexus switch reboots, OpenStack Nexus Driver will be left
    with a stale ncclient handle resulting in an exception.  When an
    exception is encountered, another retry is performed within the driver.
    The first exception information is saved so if the second exception
    fails again, we report on the first exception.
    As part of this fix, I've removed code which supports 2 versions
    of ncclient since the older version no longer meets the version
    configured in our requirements.txt.
    
    Change-Id: I182440c3a19e7c57e4dfe69e4013ea7bf87aa7ab
    Closes-bug:  #1481856
    (cherry picked from commit 46d364dc9f3b24e654939f822ea048e3cbf227c6)

commit 0c496e1d7425984bf9686b11b5c0c9c8ece23bf3
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Wed Aug 19 16:56:21 2015 -0400

Update requirements.txt for ML2 Nexus
    
    The version number to the lxml library is not correct.
    The ML2 Nexus driver requires version 3.3.3 NOT 0.3.3
    
    Change-Id: I25d21b9b470517caaffce44561b85bdb0c4e8649
    Closes-bug:  #1486164
    (cherry picked from commit 64aaa417f40dc10e321906764e70c924b6178e64)

commit 393254fcfbe3165e4253801bc3be03e15201c36d
Author: Andrew Boik <dboik@cisco.com>
Date:   Mon Jun 29 09:58:51 2015 -0400

Update requirements.txt
    
    Add ncclient and lxml versioned dependencies for ML2 Nexus,
    UCS SDK for ML2 UCSM Mech Driver.
    
    Change-Id: Ie7d4c8e173d91c1c3a9e8c77fb333b923094f258
    (cherry picked from commit d62b1c8663ea40852056924c21fe7cd79a26b0fd)

commit 75fd522b36f7b67dc4152e461f4e5dfa26b4ff31
Author: bdemers <bdemers@apache.org>
Date:   Tue Aug 11 20:57:41 2015 -0400

Remove duplicate entrypoints in setup.cfg
    
    Delete the entrypoints that are duplicates of neutron's entrypoints.
    Duplicate entrypoints cause packaging to fail.
    
    This is a cherry pick from master, but in master the driver entrypoints
    were cleaned up by driver decomposition patches. In this patch for stable
    we simply remove the driver entrypoints since they are still in neutron
    stable/kilo core.
    
    Closes Bug: 1483938
    
    Change-Id: I0301c37434dc7bca389ead83acd86686f14a8094
    (cherry picked from commit 6ba3a63e0194a36b2034defe3a630db856024c4c)

commit 178f40f2a43192687188661d5fcedf394321e191
Author: Sandhya Dasu <sadasu@cisco.com>
Date:   Thu Jul 9 19:01:52 2015 -0400

Cisco UCSM driver updates to handle duplicate creations
    
    Each create, delete or update operation on the UCS Manager is part of its own
    transaction to handle duplicate creations on UCS Manager better. So, when a
    create operation, tries to create something that already exists, the plugin
    correctly ignores the duplicate exception and proceeds to the next step
    (instead of skipping it).
    
    Change-Id: I8f650dc8d656ac8ee4a839420b68423ad65123eb
    Closes-Bug: #1464500
    Closes-Bug: #1470568
    (cherry picked from commit 4c8d9adb552b3c88deccac8f3289f622a2d7445e)

commit 11f5f29af3e5c4a2ed4b42471e32db49180693df
Author: Sandhya Dasu <sadasu@cisco.com>
Date:   Tue Mar 24 09:16:19 2015 -0400

Clean up of UCS Manager connection handle management.
    
    With the current changes, the UCS Manager Network driver raises a (new)
    exception when unable to Logout of UCS Manager using the handle passed in.
    
    Change-Id: I5b77273f6b48766a18b676b3e2fdc502ed9e7cd6
    Closes-Bug: #1433662
    (cherry picked from commit a3da889d7c693713ec6c614d282dd9c5db822895)

commit ad010718f978763e399f0bf9a0976ba51d3334eb
Author: Yanping Qu <yanping@cisco.com>
Date:   Thu Jul 16 13:06:26 2015 -0700

Fix Cisco CSR1KV script issues
    
    (1) Use service flags q-ciscorouter and ciscocfgagent for enabling
    CSR1KV.
    (2) Don't declare Q_CISCO_PLUGIN_SUBPLUGINS in cisco_neutron.
    
    Change-Id: I8407d50b975644d99bc623a7dd29e89a7ea11535
    Close-Bug: #1475406
    (cherry picked from commit 6e073199e4139366c85d287bed1ab4bdd19cc9ce)

commit a8c4bd753ba254b062612c1bcd85000656ebfa44
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Wed Aug 5 12:36:16 2015 -0400

Replace retry count with replay failure stats
    
    This code-set eliminates checking switch config replay attempts against
    max configurable attempts.  Instead it will continue to contact the
    switch forever.
    
    I've repurposed some of the logic to provide statistics to
    identify if there are issues in replay process.  There are two
    statistics: 1) contact 2) config.  The contact stat identifies
    a problem contacting the switch.  The operator would need to
    see if there is something wrong with the nexus switch or the path
    to the switch. The config stat identifies an issues with the config
    set we're trying to send to the switch.  This latter case may require
    part of the config to be removed.
    
    Change-Id: Icb15d0b2d97f1c1c9f583989823a4dae256f3e65
    Closes-bug:  #1469839
    (cherry picked from commit e6673d534cdff2d5c8434ef9928a766b4304b003)

commit db1bd250b95abfc267c8a75891ba56105cbeed8c
Author: Yanping Qu <yanping@cisco.com>
Date:   Tue Jul 7 12:42:38 2015 -0700

Add scripts to enable CSR FWaaS service
    
    Add setup scripts to install neutron-fwaas, configure CSR FWaaS service.
    A new service flag cisco-fwaas is introduced in localrc / local.conf to
    enable the service.
    
    Change-Id: Ibb64f8081de8eed0d905527e6ebc77f445f3d4ed
    Close-Bug: #1472345
    (cherry picked from commit 1c32261bdb0f32843d7d02fa7620cb7f210b0a28)

commit f39c6a55613a274d6d0e67409533edefbca6f9a7
Author: Yanping Qu <yanping@cisco.com>
Date:   Thu May 7 11:37:11 2015 -0700

Fix N1kv trunk driver: same mac assigned to ports created
    
    When update p_spec for creating the ports on trunk network, mac_address in
    p_spec['port'] should be updated and set to attributes.ATTR_NOT_SPECIFIED.
    As such a new mac address will be generated for each port created.
    
    Change-Id: I9515915263ed5ceebd4c1f007d4fc53d824ff64c
    Close-Bug: #1452845
    (cherry picked from commit 25cb20f4aa86ae59063fe1ad1b43a0fa5b5878d6)

commit a118483327f7a217dfedfe69da3ef91f9ec6a169
Author: Tom Holtzen <tholtzen@cisco.com>
Date:   Fri May 1 15:07:36 2015 -0700

Update netorking-cisco files for due to neutrons port dictionary subnet being replaced with
    a subnets list.
    
    Change-Id: I1e873c0d7e26d0d4742008562d7a04391bb45f60
    Closes-Bug: #1450130
    (cherry picked from commit 69d7d4870740666d96539fbcfe19d2849c001a8e)

commit b60296644660303fb2341ca6495611621fc486e7
Author: Rich Curran <rcurran@cisco.com>
Date:   Fri Jun 5 15:00:14 2015 -0400

ML2 cisco_nexus MD: Config hangs when replay enabled
    
    When multiple neutron-server processes are created (rpc_workers > 0)
    the cisco_nexus monitor_thread was hanging.
    
    Similar issues seen with other MDs. More info:
    http://lists.openstack.org/pipermail/openstack-dev/2015-May/063515.html
    http://lists.openstack.org/pipermail/openstack-dev/2015-June/065558.html
    https://bugs.launchpad.net/vmware-nsx/+bug/1420278
    
    09-Jun - Upstream community should be fixing this problem properly
    in near future - https://review.openstack.org/#/c/189391/1
    Once this fix is pushed up then a new bug can be used to update this
    patch to use the new method for avoiding these multi-process/thread
    issues.
    
    Change-Id: I1cbf6a4f9bc2e795720c75aca10abd4d8d458434
    Closes-Bug: #1454738
    (cherry picked from commit 347073a248ecd844f259c6ef821c5cc27d1b07d1)

commit 76f7be8758145c61e960ed37e5c93262252f56ff
Author: Sam Betts <sam@code-smash.net>
Date:   Wed Jul 29 14:24:11 2015 +0100

Move UCSM from extension drivers to mech drivers
    
    The UCSM mech driver endpoint was in the wrong place in setup.cfg, this
    patch moves it in the right place.
    
    Change-Id: I4b99c2990cdbb8009b75a7f1862666e2a69b26e8
    (cherry picked from commit 923dfd79ef6eb0418d5d56e5d1ce225334424963)

commit ffabc773febb9a8df7853588ae27a4fe3bc4069b
Author: Rich Curran <rcurran@cisco.com>
Date:   Sat Jun 20 13:38:27 2015 -0400

ML2 cisco_nexus MD: Multiprocess replay issue
    
    When api or rpc_workers are set > 0 replay is not working in multi-node setups.
    Issue is that access to nexus switches is to set to False (can't access) by
    default in the monitor code. In multiprocess configurations the monitor
    thread (thread code determines if switch is accessible) only runs in the
    main process (mech_cisco_nexus.py initialize() method is only called once,
    regardless of number of processes). In all other processes the "access
    switch" setting never gets set to True (can access).
    
    This patch allows all non-parent processes to continue event handler
    processing. Only the parent process (which starts the monitor thread) will
    verify if the switch is accessible during event processing.
    
    Change-Id: I637aba7615df2711fdf058e57800c28e3741dadb
    Partial-Bug: #1454738
    (cherry picked from commit 7d37c15e8fc0e20ad9e3b0606d7a2fc534bfdd4a)

commit 77d4a60fbce7f81275c3cdd9fec3b28a1ca0c57c
Author: Rich Curran <rcurran@cisco.com>
Date:   Wed Jul 1 13:06:07 2015 -0400

ML2 cisco_nexus MD: If configured, close ssh sessions
    
    Most (if not all) Nexus switches have a limit of eight ssh sessions open
    at one time. With the openstack/neutron introduction of rpc and api workers
    (setting these variables results in additional neutron-server processes
    being forked) when more than eight processes are in use (including one for
    the parent neutron-server process) the ssh limit on the nexus switches is
    exceeded and exceptions are taken. More importantly is that communication
    to the nexus switches can not be made for the nineth (or above) processes.
    
    This nexus switch limitation will be fixed by closing each session after
    each event if the overall number of processes is >= eight.
    
    Change-Id: Ie953c955c6e045a1b252e6973f53b38284833f37
    Closes-Bug: #1454734
    (cherry picked from commit 1eec7fb821e4d75fb6f875f9edeb2977bfcd88ef)

commit 825cf6d1239600917f8fa545cc3745517d363838
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Fri Jul 24 15:06:10 2015 -0400

Part II-Detect switch failure earlier-Port Create
    
    This changeset addresses comments received post merge.
    ref: https://review.openstack.org/#/c/205080/
    
    Change-Id: I070725601e2135328897fdbc84e1cde0d930aecf
    Closes-bug:  #1476723
    (cherry picked from commit d447a1f964b71a3d770ba2060733c21585035f48)

commit 9b7b57097b2bd34f42ca5adce1e3342a91b4d3f8
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Thu Jun 25 11:03:03 2015 -0400

Retry count not reset on successful replay
    
    Retry count currently is only reset when a valid config request is made.
    It was an oversight in original design to not reset the retry
    counter when replay completed successfully.  Also added log entries
    for success and failure to help test team quickly detect results.
    
    Change-Id: I0c06c0e7860020f26fd610a5391a6087ac5917fc
    Closes-bug:  #1466511
    (cherry picked from commit b6e438a4bc45f5f4e2824be478cb83c9afc5ff35)

commit 6afe5d8a6d11db4bc2db29e6a84dc709672b1d69
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Fri Jul 24 10:39:22 2015 -0400

ML2 Nexus decomposition not complete for Nexus
    
    I found file ml2_conf_cisco.ini exists in two repos:  The
    openstack/neutron and openstack/networking-cisco.  The file
    in openstack/networking-cisco was missing changes for the
    ML2 Nexus Driver.  This changeset adds those changes.
    
    Change-Id: I15c93fad3842bf6dc2a7f6ab4614dc849062029e
    Partially-implements: blueprint core-vendor-decomposition
    (cherry picked from commit 830545442e6652978d8324318fd068e019ce5251)

commit ac84fcb861bd594a5a3773c32e06b3e58a729308
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Thu Jul 23 12:34:26 2015 -0400

Delete fails after switch reset (replay off)
    
    Many changes have occurred  since this bug was written.  We now have
    a newer N9K image and dhcp port entries are now created when the CLI
    'neutron subnet-create' is executed.  I went thru the steps as
    defined by the bug. The first failure after switch restart is due
    to the ncclient session being stale (fixed by 1399998).  This is
    expected when replay is not enabled.  When attempting delete on
    2nd retry when ncclient session new, the N9K switch returns the error
    message 'None of the VLANs exist'.  This changeset accepts this
    message as an acceptable condition and prevents exceptions from
    bubbling up to higher layer.
    
    Change-Id: I072dc05590d8c7d376c491c931d18befa507af30
    Closes-bug:  #1430308
    (cherry picked from commit 6d2be75e00fca5b844562dba9c1f10a1c038a0a9)

commit 97720feb4ef4d75fa190a23ac10038d29582b001
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Tue Jul 21 11:02:31 2015 -0400

Call to get nexus type for Nexus 9372PX fails
    
    Started seeing the error "Get call failed to return Nexus type" when
    setting up network with 2 switches.  There was only one Nexus switch
    in which error appeared on and it was the Nexus9372PX.  It seems
    the XML output from EXEC_GET_INVENTORY_SNIPPET from both these
    N9K were not consistent.  The word Chassis was capitalized in
    9396 while not in 9372.
    
    Change-Id: Ib0f44fe59229e0b2f945addcbcd624ed40e68340
    Closes-bug:  #1466164
    (cherry picked from commit 2a43c81b28bd99f13bb4eaeffc36a2b91b249e34)

commit 87fb3d6f75f9b0ae574df17b494421126a636199
Author: Carol Bouchard <caboucha@cisco.com>
Date:   Thu Jul 23 10:27:38 2015 -0400

Detect switch failure earlier during port create
    
    This code set performs a get call to the switch earlier during the call
    to create_port_postcommit(). This will detect if the switch is up.  If
    it fails, we will no longer have persistent retries and other
    components will clean up their data bases which does not happen with
    failed calls to update_port_postcommit().
    
    Change-Id: Ibc8d7163f27e5c040e502b1743d805996aff6643
    Closes-bug:  #1476723
    (cherry picked from commit 81a7c328276659d917c053964e8eb33408f568cc)

commit b38e47a37977634df14846ba38aa38d7239a1adc
Author: Yanping Qu <yanping@cisco.com>
Date:   Tue Jun 30 14:59:23 2015 -0700

Enable the CSR1kv devstack plugin for Kilo
    
    Add CSR1kv scripts, localrc and local.conf samples, README.TXT.
    
    Introduce a new service flag cisco-csr that can be set in localrc
    or local.conf to enable CSR1kv routing.
    
    Modify plugin.sh to invoke CSR1kv scripts accordingly at the proper
    stage of running stack.sh
    
    Change-Id: Ibdcd111c6a3c30624fd30f6e4686fb2763549333
    Close-Bug: 1460160
    (cherry picked up from commit Ibdcd111c6a3c30624fd30f6e4686fb2763549333)

commit 365cd0f94e579a4c885e6ea9c94f5df241fb2288
Author: shivrao <shivrao@cisco.com>
Date:   Wed Jul 15 13:44:03 2015 -0700

Sanitize policy profile table on neutron restart
    
    On neutron restart, VSM IPs can be reconfigured. We need to sanitize
    the policy profile table and delete entries for VSMs that are not
    configured. Delete will only take place if the profile is not in use
    or if there is another VSM with the profile configured.
    
    Change-Id: I631417da254d12184b71314391737173ab2e03c1
    Closes-Bug: #1446892
    (cherry picked from commit 2c1876be1690063b13c4a3aa5ad0e42144ca8009)

commit 4a6a4040a71096b31ca5c283fd0df15fb87aeb38
Author: Sourabh Patwardhan <sopatwar@cisco.com>
Date:   Mon Jun 15 15:39:59 2015 -0700

Cisco Nexus1000V: Retry mechanism for VSM REST calls
    
    Added a config option and retry mechanism to handle connection
    timeouts encountered during VSM REST API calls from the client.
    
    Change-Id: Ic9db4c61d950f675d77257aff0d11a4d5f76c3f6
    Closes-Bug: #1465426
    (cherry picked from commit b20dac9c062333fc8952e98c8fc2c0bd43686851)

commit 7bcec734cbc658f4cd0792c625aff1a3edc73208
Author: Saksham Varma <sakvarma@cisco.com>
Date:   Mon Jun 1 12:01:04 2015 -0700

Moved N1kv section from neutron tree to stackforge
    
    Since most of N1kv plugin code resides in stackforge/networking-cisco
    repo, it's best to move the N1kv section here. Also added sync_interval
    parameter to make time duration b/w consecutive VSM-neutron syncs
    configurable.
    
    Change-Id: Ia26b504a4135b12f003a89e3f19ff650e0d271a4
    Closes-Bug: #1441400
    (cherry picked from commit 9f53a8cdb2188f4ded9b3215ebf5ea68e264c281)

commit 4970a3e279995faf9aff402c96d4b16796a00ef5
Author: Saksham Varma <sakvarma@cisco.com>
Date:   Wed Jun 10 15:50:26 2015 -0700

N1Kv: Force syncing BDs on neutron restart
    
    Sync had no way to detect a BD mismatch if VSM and Neutron networks
    are already in sync. This adds functionality to sync bridge
    domains on neutron restart, in such cases -- when a BD might be
    missing for a n/w segment on VSM, or if a stale BD is sticking
    around on the VSM
    
    Change-Id: I3a89914a58cc99c0f0484cb068ea230439802c0f
    Closes-Bug: #1463950
    (cherry picked from commit ef46d7c0d2b735bd12550db662fd06d2974217d5)

commit f078a701931986a2755d340d5f4a7cc2ab095bb3
Author: Henry Gessau <gessau@cisco.com>
Date:   Fri Jun 12 21:11:09 2015 -0400

s/stackforge/openstack/g
    
    Change-Id: I9100d66c4af912ecd6e8f6d0c2c3cbebe536171e

commit 151f6f6836491b77e0e788089e0cf9edbe9b7e00
Author: Jeremy Stanley <fungi@yuggoth.org>
Date:   Sat Jun 13 00:51:00 2015 +0000

Update .gitreview file for project rename
    
    Change-Id: I0e2ae0487a1c10dfaa4faf7da0c8ad769ed08553

commit 876c25fbf7e3aa7f8a44dd88560a030e609648d5
Author: Sam Betts <sam@code-smash.net>
Date:   Fri Jun 12 09:09:20 2015 +0100

Bump minor version number to enable development
    
    and Account for random vni allocation in vxlan tests
    
    Change [1] introduced random tunnel id allocation. Update the nexus
    vxlan tests to account for this.
    
    [1] https://review.openstack.org/182901
    
    Cherry-picked from: I689b9dbba436b9823d6a75bee42293a7dcedbc7d
    Change-Id: I66c8a9c70f67ea718bfe814fc10079df241401d9

commit a5e7f6a3f0f824ec313449273cf9b283cf1fd3b9
Author: Saksham Varma <sakvarma@cisco.com>
Date:   Thu Apr 16 15:53:13 2015 -0700

Sync notification to VSM & major sync refactoring
    
    Send sync notification (start/end/no-change) to NSM, which will then
    send an MTS msg with this notification to the FWM.
    
    Change-Id: Ie0b935b75b6b6cc77b9621d6b16b526e910ec3f4
    Closes-Bug: #1440212
    (cherry picked from commit a4991886097a084ed2913bbe84732816b2940b12)

commit 24edb9fd14584020a8b242a8b351befc5ddafb7e
Author: Sam Betts <sam@code-smash.net>
Date:   Wed May 6 15:57:28 2015 +0100

Ensure that we're neutron stable/kilo compatible
    
    This patch both reverts a patch that made us incompatible with neutron
    stable/kilo and updates the requirements.txt and tox.ini so that it will
    install neutron stable/kilo for UTs and also ensures "python setup.py
    install" verifies that a compatible version of neutron is installed.
    
    This reverts commit 3d9b38c758e49db0798e6bad3607c6bcd203d065.
    
    Change-Id: I5c84854069d556835001151462cb512b7f12ebe1

commit 013aedf90122c99ac91d6d2be83844c096d13da0
Author: Steven Hillman <sthillma@cisco.com>
Date:   Fri Feb 20 15:46:17 2015 -0800

Send tenant_id 0 for shared networks to VSM
    
    We need to indicate if a network is shared for VSM. In order
    to do so, we will now send tenant_id = '0' for shared networks. This
    change additionally ensures the tenant_id is properly updated when the
    network is updated to be a shared network after it is created.
    
    Also removed broken test case in monolithic plugin UTs and fixed
    copyright lines in UT files.
    
    Closes-Bug: #1440214
    Change-Id: I50c0d19a47f82994a9b17fc70ea9e2fba6deec39
    (cherry picked from commit cb774a512f4d7ba44ba5055e4be7c2eb9d639fd6)

commit 888ae8ff553b5ea4e715d0161f2c59b15e3ec1cc
Author: Steven Hillman <sthillma@cisco.com>
Date:   Thu Apr 2 16:17:27 2015 -0700

Implement update port for N1kV ML2 mechanism driver
    
    Added support for update port operation in order to have this operation
    work successfully. On update port we check if this is the the first time
    binding of a router or DHCP port as currently that is the only operation
    we want to support updates for.
    
    Closes-Bug: #1447369
    Change-Id: I69624d7b0cb0eb82bc5cf6ed19202e076bee7cc4
    (cherry picked from commit e7230174792199f932de92492e537fe263d18824)

commit 1ea0799efae39623a0933e60c56ee26e40bd03ad
Author: Sam Betts <sam@code-smash.net>
Date:   Fri May 1 16:31:04 2015 +0100

Update to version 2015.1.0
    
    Change-Id: I50806ce0cadd1828e963b9b80c5569037706f9ac
    (cherry picked from commit 03e26dc7527488ea2508c992ecb6076bb460eb08)

Changed in networking-cisco:
status:	Fix Committed → Fix Released

	Status	Importance	Assigned to	Milestone
networking-cisco	Fix Released	Undecided	Rich Curran	networking-cisco 2.0.0
Kilo	Fix Committed	Undecided	Rich Curran	networking-cisco 1.1.0
Liberty	Fix Released	Undecided	Rich Curran	networking-cisco 2.0.0

networking-cisco

Nexus VLAN config hangs when replay enabled on RHEL set-up

Bug Description

Other bug subscribers

Remote bug watches