#opendaylight-ovsdb: ovsdb_weekly_call
Meeting started by tbachman at 20:00:11 UTC
(full logs).
Meeting summary
- Agenda (tbachman, 20:00:24)
- https://meetings.opendaylight.org/opendaylight-ovsdb/2015/ovsdb_weekly_call/opendaylight-ovsdb-ovsdb_weekly_call.2015-01-20-20.00.html
Minutes from last week’s meeting (tbachman,
20:00:34)
- It's my birthday today (mestery,
20:02:08)
- shague suggests the "50 Cent" version of "It's
Your Birthday" for mestery's song. (mestery,
20:04:04)
- trello board (tbachman, 20:06:11)
- shague finished getting all the IT tests into
the integration directory (tbachman,
20:06:52)
- shague and edwarnicke have been coordinating on
getting neutron into it’s own MD-SAL piece (tbachman,
20:07:15)
- plaurin_ has finished porting of L3 services,
and is now working on creating the UT for the services (tbachman,
20:07:51)
- mestery is hoping that Swami can set up a call
to go over the testing (tbachman,
20:08:12)
- ACTION: Swami to set
up a call to go over UT, possibly on Friday (tbachman,
20:08:30)
- DPDK integration with Eric Multanen, Sharad,
and edwarnicke is ongoing — working on the models (tbachman,
20:09:03)
- Eric asks if the capability to support DPDK
ports is going to be in the stackforge effort (tbachman,
20:10:10)
- shague says that the OVS L2 agent is what’s
talking to OVSDB — the agent is being replaced by ODL
integration (tbachman,
20:10:42)
- mestery says that ODL is effectively doing the
work that the agent was doing (tbachman,
20:11:02)
- mestery says that before the kilo release, the
entire plugin was upstream in neutron, and did contain back-end
logic to talk to ODL. As part of the decomposition in neutron, there
will be a super-thin plugin that pulls in the stackforge
project (tbachman,
20:11:41)
- Eric asks where the openstack folks are going
to get the DPDK port information (tbachman,
20:12:08)
- shague says there are some neutron calls that
talk to the OVS agent, which goes back into nova so it can use DPDK
ports; since the L2 agent isn’t there any more, where would that
belong/go with ODL? (tbachman,
20:12:47)
- Eric says that they want to know if the given
host and bridge can support a DPDK port type, so that nova can go
and spawn that VM (tbachman,
20:13:30)
- mestery says those VIFs are created by
Nova (tbachman,
20:14:08)
- mestery asks if this can be as simple as a
check of OVS versions? (tbachman,
20:14:24)
- shague says he doesn’t think the version
matters, b/c when the VIF gets spun up, they need to query to see if
it is a DPDK supported port; shague thinks blp added some support to
bridge tables to indicate this (tbachman,
20:15:36)
- mestery says this would have to be exposed in
ODL (tbachman,
20:15:53)
- mestery asks who’s doing this query
(tbachman,
20:15:58)
- shague says it’s in openstack - they come down
thru neutron L2 agent (tbachman,
20:16:14)
- mestery says that even if we put the data in
the database, that’s not going to work (tbachman,
20:16:28)
- shague asks if the ODL stackforge stuff can
address this (tbachman,
20:16:36)
- mestery says we might have to expose those
details as part of the port binding (tbachman,
20:16:51)
- mestery says that requires some coordination in
the back-end ODL side, as the ML2 driver doesn’t have anything
running on the host (tbachman,
20:17:08)
- stackforge efforts (tbachman, 20:17:43)
- recording started 20-minutes into the
meeting (tbachman,
20:20:29)
- as part of the kilo development work, blueprint
was approved where all of the plugins and drivers are going through
the decomposition process (tbachman,
20:21:04)
- there was a lot of code in neutron dealing with
proprietary or opensource backends, where reviewers didn’t
understand the backends (tbachman,
20:21:24)
- https://launchpad.net/networking-odl
"OpenDaylight backend controller integration with Neutron" (flaviof,
20:21:25)
- there are over 25 plugins and drivers in
neutrons, and only 14 reviewers (tbachman,
20:21:39)
- The result is that everyone can have a thin
shim in neutron which is a pass thru (tbachman,
20:21:55)
- the back-ends can live in stackforge, github,
etc., and can iterate much faster (tbachman,
20:22:17)
- there are still CI requirements on the
backends (tbachman,
20:22:25)
- The ODL mechanism driver is the first to be
fully decomposed upstream (tbachman,
20:22:36)
- https://github.com/stackforge/networking-odl
Stackforge ODL project (tbachman,
20:23:59)
- Users will be able to use pip to install, or
setup.py to install locally as well (tbachman,
20:24:28)
- https://review.openstack.org/gitweb?p=stackforge/networking-odl.git
networking-odl gerrit (flaviof,
20:24:43)
- mestery says there’s been a good confluence of
people who are new to either OpenStack or OpenDaylight that have
been jumping in (tbachman,
20:25:37)
- Swami, plaurin_, et. al. have joined in the
development (tbachman,
20:26:09)
- mestery says that once this infra is in place,
the tempest testing can be addressed (tbachman,
20:26:59)
- mestery filed a bunch of bugs on them, some of
which are hard to replicate (tbachman,
20:27:10)
- flaviof says ODL also has an effort to take
neutron out of the controller, which will isolate it and help with
other efforts (tbachman,
20:28:21)
- mestery says that CI is pretty hosed at the
moment, but it didn’t make sense to work on CI until the stackforge
stuff was set up (tbachman,
20:28:50)
- mestery and colindixon says that a small,
focused, sprint can happen via hangout/bluejeans/!webex (tbachman,
20:29:24)
- edwarnicke asks if this is connected with the
ODL integration team (tbachman,
20:29:40)
- edwarnicke says that’s the right place for the
CI to live, as that’s the thing that always gets checked for every
bug fix release (tbachman,
20:30:01)
- vishnoianil_ says there’s a plan to integration
neutron to the integration team (tbachman,
20:30:24)
- mestery says he’d like to move the jenkins
config into Jenkins Job Builder as well (tbachman,
20:30:40)
- mestery says he doesn’t think the neutron
tempest ever ran against the jobs on the OVSDB side (tbachman,
20:31:01)
- edwarnicke asks where integration stands on
migrating their jobs into Jenkins Job Builder (tbachman,
20:31:15)
- shague says moving their stuff into JJB is kind
of hard; wonders if integration doing this will help out here
(tbachman,
20:31:47)
- edwarnicke says he knows this is a high
priority for the integration group (tbachman,
20:32:01)
- flaviof asks if ovsdb needs a CI in openstack
that doesn’t use OVS, would that be better when the neutron has
moved out of the controller (tbachman,
20:33:56)
- edwarnicke says all the providers would live
downstream; everything meets in the integration project, which makes
it the optimal place to do this stuff (tbachman,
20:34:14)
- flaviof says there are a lot of hooks into OVS,
and isn’t sure how much of that could be “faked out" (tbachman,
20:34:35)
- colindixon says the distribution selected could
run the respective tests against that distribution; that could be
part of the hack-fest work (tbachman,
20:35:10)
- mestery says he feels these jobs need to live
in OVSDB; configuring things like VTN and OpenDove is outside the
scope of what we’d want to do (tbachman,
20:36:04)
- colindixon says we shouldn’t take on
resposibility for that as part of the OVSDB project (tbachman,
20:36:17)
- edwarnicke says the multiple implementations of
neutron stuff makes it more efficient to have this in
integration (tbachman,
20:36:50)
- flaviof asks how agnostic tempest is of
OVS (tbachman,
20:37:33)
- mestery says it’s pretty agnostic; however,
things have to be configured for ODL; if OpenDove and VTN folks need
to drive this testing, they’d have to provide this
configuration (tbachman,
20:38:05)
- mestery says right now they’re pulling released
versions of ODL; the matrix gets complex quickly, which is one
reason for just using released versions (tbachman,
20:38:38)
- mestery wants to make sure the upstream
openstack CI stuff is in OVSDB (tbachman,
20:39:11)
Meeting ended at 20:41:28 UTC
(full logs).
Action items
- Swami to set up a call to go over UT, possibly on Friday
People present (lines said)
- tbachman (86)
- mestery (11)
- odl_meetbot (5)
- flaviof (3)
- shague (2)
- plaurin_ (1)
- colindixon (1)
- abhijitkumbhare (1)
Generated by MeetBot 0.1.4.