#opendaylight-group-policy: gbp_status_arch
Meeting started by tbachman at 18:00:03 UTC
(full logs).
Meeting summary
- agenda (tbachman, 18:00:11)
- https://wiki.opendaylight.org/view/Group_Policy:Sub-Groups:STATUS#Team_Meeting
Agend for today’s meeting (tbachman,
18:00:17)
- https://meetings.opendaylight.org/opendaylight-group-policy/2015/gbp_status_arch/opendaylight-group-policy-gbp_status_arch.2015-01-02-18.00.html
The last meeting minutes that we have (tbachman,
18:00:42)
- GBP as Neutron Provider (tbachman, 18:01:03)
- GBP can bring benefits to NFV, but to do that,
we need to implement a provider to the neutron API (tbachman,
18:05:06)
- There are some model enhancements made in order
to support this (tbachman,
18:05:59)
- the ODL neutron service will be re-used, and
will interface to a neutron-to-GBP translator (tbachman,
18:06:14)
- https://docs.google.com/presentation/d/1AO47EYrDMuTAypcpYbL3XOKFggZXu9x00XpGVwJLr9o/edit#slide=id.p
presentation that alagalah_ is doing (tbachman,
18:06:29)
- Endpoints belong to groups, and the groups
interact diretionally using contracts (tbachman,
18:06:42)
- Endpoint Groups can specify a default netowkr
context for members (tbachman,
18:07:49)
- Neutron maps well into GBP — network is a
bridge domain, port is an Endpoint, Subnet is a subnet, Security
group is an Endpoint Group, and Routers are an L3 context
(tbachman,
18:09:02)
- alagalah_ says that policy dictates what
traffic can pass; with security groups in neutron, a new EPG is
created to map to the neutron security group (tbachman,
18:10:11)
- an endpoint (port) can then participate in the
security group EPG and it’s other EPG (tbachman,
18:10:53)
- dbainbri asks if there’s any conflicts when an
EP belongs to more than one security group (tbachman,
18:11:06)
- alagalah_ says he hasn’t seen anything concrete
in terms of conflicts; because it’s a white list model, there’s no
ordering, but if there’s ordering needed as a use case, it wouldn’t
be hard to add it to the model (tbachman,
18:11:55)
- mickey_spiegel says that you don’t have
conflicts is b/c it’s all allow rules (tbachman,
18:12:13)
- Prem asks if there will be an hierarchical
ordering with security groups (tbachman,
18:12:31)
- alagalah_ says that the existing model supports
inheritance, so it could support that (tbachman,
18:12:43)
- A NAT Endpoing Group can be created and an
Endpoint can join that group in order to implement NAT (tbachman,
18:13:57)
- A similar concept can be used with LBaaS
(tbachman,
18:14:18)
- for SFC, it’s a bit different, where the
service chain is defined by the contract; the EPGs define the
relationship to select the contract that implements the chain
(tbachman,
18:14:52)
- alagalah_ shows that OpenStack passes calls to
the ODL Neutron service (via ODL pass-thru neutron Plugin)
(tbachman,
18:15:15)
- There will be a neutron to GBP translation
layer, which then passes things on to GBP (tbachman,
18:15:30)
- The neutron to GBP translator doesn’t exist
today; we need some changes to the model (EP in multiple
EPGs) (tbachman,
18:16:00)
- There are also some changes to help some
forwarding model constructs (tbachman,
18:16:10)
- abhijitkumbhare asks where the GBP openstack
fits in this model (tbachman,
18:16:47)
- alagalah_ says this is just passing straight
neutron to ODL (tbachman,
18:16:56)
- abhijitkumbhare asks if we’re using GBP in
openstack (tbachman,
18:17:06)
- alagalah_ says we’ve already done that
integration, but we want to implement a neutron interface in order
to support NFV use cases (tbachman,
18:17:32)
- alagalah_ says that automated testing setup is
needed — not just for this, but for all of ODL — in order to run
tempest testing against any neutron provider (tbachman,
18:18:05)
- daniel asks if the actions like LBaaS and
others are defined, or are these defined by what the renderer
supports (tbachman,
18:18:35)
- alagalah_ says this is more in the
renderer (tbachman,
18:18:41)
- daniel says he thinks it would be worthwhile to
have a minimal set of action enums defined (tbachman,
18:18:56)
- alagalah_ says there’s an enumerated type in
the yang model, which currently is only allow (tbachman,
18:19:16)
- daniel says there are at least 4 different
kinds of actions that they have use cases for (tbachman,
18:19:43)
- Louis asks if the redirect to a service chain
has been covered (tbachman,
18:20:18)
- alagalah_ says that’s what the project is
looking at (tbachman,
18:20:30)
- cdub asks how you map the VIF UUID to the
policy group (did I get that right?) (tbachman,
18:20:51)
- cdub asks if the ODL constructs are part of a
CRUD call, or are they policies (tbachman,
18:21:41)
- cdub says there is no service chain APIs in
neutron (tbachman,
18:21:51)
- This means openstack would have to inform ODL
with the unique policy ID (tbachman,
18:22:08)
- Rajeev asks how things are passed from neutron
— e.g. policies specific to SFC (tbachman,
18:23:15)
- edwarnicke says that neutron constructs must be
self contained and are passed thru (tbachman,
18:23:31)
- edwarnicke says that constructs that aren’t in
neutron have to be piggy-backed (tbachman,
18:23:46)
- Rajeev asks if an opaque container could be
passed through neutron and given to ODL’s SFC (tbachman,
18:24:37)
- edwarnicke says that’s certainly
possible (tbachman,
18:24:42)
- Rajeev says that this was brought up at the
OpenStack design summit and said that we needed to look at the use
cases; we may have a use case now (tbachman,
18:25:35)
- alagalah_ will update the release plan and
trello board with these tasks (tbachman,
18:25:55)
- Project status update (tbachman, 18:26:09)
- alagalah_ says that openstack GBP integration
is complete (tbachman,
18:26:37)
- tbachman has been working on the opflex
renderer — patch is outstanding (tbachman,
18:28:13)
Meeting ended at 18:28:23 UTC
(full logs).
Action items
- (none)
People present (lines said)
- tbachman (67)
- odl_meetbot (5)
- yapeng_ (2)
- alagalah_ (1)
Generated by MeetBot 0.1.4.