mDNS should Handle Bad File Descriptor

[Kiall Mac Innes]

While replying to an AXFR, a connection was interrupted(?) resulting in:

2015-03-28 19:30:01.201 INFO designate.policy [req-4e70463b-2824-4e3e-8529-e47604194a4f - - - - -] Policy check succeeded for rule 'all_tenants' on target {}
2015-03-28 19:30:01.202 ERROR designate.dnsutils [req-4e70463b-2824-4e3e-8529-e47604194a4f - - - - -] Failed to deserialize packet from 208.78.68.65:23357
2015-03-28 19:30:01.202 ERROR designate.service [req-4e70463b-2824-4e3e-8529-e47604194a4f - - - - -] Unhandled exception while processing request from 208.78.68.65:23357
2015-03-28 19:30:01.202 TRACE designate.service Traceback (most recent call last):
2015-03-28 19:30:01.202 TRACE designate.service File "/opt/stack/designate/designate/service.py", line 332, in _dns_handle
2015-03-28 19:30:01.202 TRACE designate.service client.send(tcp_response)
2015-03-28 19:30:01.202 TRACE designate.service File "/usr/local/lib/python2.7/dist-packages/eventlet/greenio/base.py", line 359, in send
2015-03-28 19:30:01.202 TRACE designate.service total_sent += fd.send(data[total_sent:], flags)
2015-03-28 19:30:01.202 TRACE designate.service File "/usr/lib/python2.7/socket.py", line 170, in _dummy
2015-03-28 19:30:01.202 TRACE designate.service raise error(EBADF, 'Bad file descriptor')
2015-03-28 19:30:01.202 TRACE designate.service error: [Errno 9] Bad file descriptor
2015-03-28 19:30:01.202 TRACE designate.service

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/176266

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/176266
Committed: https://git.openstack.org/cgit/openstack/designate/commit/?id=fb58ff9f7b6ffe8bd8d3b9129091f34dd47b1128
Submitter: Jenkins
Branch: master

commit fb58ff9f7b6ffe8bd8d3b9129091f34dd47b1128
Author: Kiall Mac Innes <email address hidden>
Date: Wed Apr 22 13:02:23 2015 +0100

    Handle socket.error's in mDNS

    This ensures we don't log stacktraces for conditions outside
    of our control, like a remote client closing the connection
    abruptly.

    Closes-Bug: 1437699
    Change-Id: Idd54ec4ef0dce1a731a0c6b0e77375bf5f42080c

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/kilo
Review: https://review.openstack.org/176810

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/176810
Committed: https://git.openstack.org/cgit/openstack/designate/commit/?id=4e937688a4051860ad92349e82cab557c8672958
Submitter: Jenkins
Branch: stable/kilo

commit 4e937688a4051860ad92349e82cab557c8672958
Author: Kiall Mac Innes <email address hidden>
Date: Wed Apr 22 13:02:23 2015 +0100

    Handle socket.error's in mDNS

    This ensures we don't log stacktraces for conditions outside
    of our control, like a remote client closing the connection
    abruptly.

    Closes-Bug: 1437699
    Change-Id: Idd54ec4ef0dce1a731a0c6b0e77375bf5f42080c
    (cherry picked from commit fb58ff9f7b6ffe8bd8d3b9129091f34dd47b1128)

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/179248

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/179248
Committed: https://git.openstack.org/cgit/openstack/designate/commit/?id=51a4d4dc1aac7b3eee62747b5000c22f6524c3f9
Submitter: Jenkins
Branch: master

commit f6ee98e4ed4529b0d699f3cf2d107a1ed8faa6f0
Author: stanzgy <email address hidden>
Date: Wed Apr 22 13:53:32 2015 +0800

    Remove duplicated index on table 'records' of pDNS backend

    Remove duplicated index `rec_name_index` on table 'records' of pDNS backend.

    Change-Id: I9e8723b464522a588f0e0ef6ff261b71609b0726
    Closes-Bug: 1446980

commit 2900cd391efae38be1121e58e72f5519f4a95e90
Author: Kiall Mac Innes <email address hidden>
Date: Wed Apr 22 15:42:29 2015 +0100

    central.update_status should be transactional

    Central's update status method issues a large number of SQL
    queries for large zones with many recent updates, as a result
    it's checking out many connections from our connection pool one
    after another, contributing to bug 1445123. By wrapping the
    call in a transaction, a single connection will be checked out
    of the pool.

    Partial-Bug: 1445123
    Change-Id: Idfde7578e1a17a6a0585835fe614f787102563fa
    (cherry picked from commit ac79df1567fb040067ed59d948e8a39fb90af177)

commit c7865e2c5e220b6fa9233c4985e657298d9935ae
Author: Graham Hayes <email address hidden>
Date: Thu Apr 23 16:54:04 2015 +0100

    Add the /v2/zones/<id>/nameservers endpoint back

    APIImpact
    Closes-Bug: #1447683

    Change-Id: I80b51e4e04fc7395d8d1f32a3b3d34f09520c75c

commit 56c8877b7738c434470e0867432fdd7c454a126e
Author: Tim Simmons <email address hidden>
Date: Wed Apr 15 21:04:46 2015 +0000

    Update JSON in howtos examples to match the API

    Some of the howto examples haven't been updated to match the removal
    of wrapping resources in the API, this fixes that

    Closes-Bug: 1447663
    Change-Id: I84a349e2848329a63f84b6073ba65ac5fd56877c
    (cherry picked from commit fc2f733cb53bcce7f03d5c932d99caebb2fe2ce9)

commit b791937b717789d6d0bc0011272ac345b18c46a0
Author: Kiall Mac Innes <email address hidden>
Date: Tue Apr 21 13:17:16 2015 +0100

    Service ThreadPool size should be configurable

    Change-Id: Icbf8873723f61990a76dae5f698b3cc370a48ff0
    Closes-Bug: 1444901
    (cherry picked from commit ad922201989ff0fde6e449e3c06a0208d809886d)

commit 4e937688a4051860ad92349e82cab557c8672958
Author: Kiall Mac Innes <email address hidden>
Date: Wed Apr 22 13:02:23 2015 +0100

    Handle socket.error's in mDNS

    This ensures we don't log stacktraces for conditions outside
    of our control, like a remote client closing the connection
    abruptly.

    Closes-Bug: 1437699
    Change-Id: Idd54ec4ef0dce1a731a0c6b0e77375bf5f42080c
    (cherry picked from commit fb58ff9f7b6ffe8bd8d3b9129091f34dd47b1128)

commit 4cefc818be22ebe097ab6f08569933f98ea33e3c
Author: Graham Hayes <email address hidden>
Date: Thu Apr 16 18:48:42 2015 +0200

    Move the Central rpcapi update_status to cast

    Moves the update_status call (used by pool manager) to a cast.

    Pool Manager doesn't actually do anything with the return value,
    and causes another choke point.

    Cl...