#opnfv-sfc: OPNFV SFC weekly meeting: August 2, 2015
Meeting started by ebrjohn at 14:02:55 UTC
(full logs).
Meeting summary
-
- Chris (on IRC only) (ChrisPriceAB,
14:03:26)
- ebrjohn Brady Johnson (ebrjohn,
14:04:27)
- Bryan Sullivan (bryan_att,
14:04:44)
- Tim Rozet (trozet,
14:04:45)
- Serge Manning (SergeM,
14:04:48)
- sridhar_ram Sridhar Ramaswamy (sridhar_ram,
14:04:49)
- Andre Fredette (afredette,
14:04:59)
- Nicolas Bouthors (bouthors,
14:05:00)
- Dan Radez (radez,
14:05:05)
- Brady will update the wiki for the AIs - 3
ongoing (bryan_att,
14:05:59)
- Provide input to scripts is ongoing
(bryan_att,
14:06:30)
- 2nd is NSH-aware service functions (Paul);
products are in development, looking for an image the project can
use if possible (bryan_att,
14:07:24)
- 3rd (Uri) to investigate VNFM test driver (no
update right now) (bryan_att,
14:08:19)
- Sridhar from the Tacker project is here
(bryan_att,
14:09:13)
- Interested in bringing Tacker to this project,
and moving up to NVFO (bryan_att,
14:09:37)
- looking at bringing Tim's github work into the
Tacker repo, to instantiate VNFs and use the API to chain them; in
future all VNFs will be written in TOSCA (bryan_att,
14:10:50)
- current approach is fine given that the TOSCA
forwarding graph work is not completed (bryan_att,
14:11:18)
- ebrjohn: what about the dependencies between
projects (bryan_att,
14:11:50)
- sridhar_ram: (describes the stack) there is a
general top API with ODL as one mechanism; a clear path down
currently (bryan_att,
14:13:19)
- ebrjohn: concerned about any cyclic
dependencies (bryan_att,
14:13:40)
- sridhar_ram: will bring a diagram to
discuss (bryan_att,
14:14:05)
- trozet: can we have more than one option to
spec the chain (bryan_att,
14:15:39)
- https://github.com/openstack/networking-sfc/blob/master/doc/source/api.rst
(trozet,
14:16:18)
- neutron SFC spec (trozet,
14:16:26)
- rprakash: we should focus this on the neutron
APIs with passing parameters as needed; any MANO component could
drive this not just Tacker; a more appropriate way but left to the
team (bryan_att,
14:16:59)
- ebrjohn: should we move to bi-weekly
meetings (bryan_att,
14:18:22)
- rprakash: weekly is needed until the C
milestone (bryan_att,
14:18:49)
- bryan_att: agree (bryan_att,
14:18:57)
- ebrjohn: can create openstack vxlan ports
first (bryan_att,
14:20:10)
- (colleague explains the issue) (bryan_att,
14:21:17)
- ACTION: ferenc will
send a document explaining the results (ebrjohn,
14:23:14)
- trozet: you can't add a vxlan port in neutron -
it requires code changes (bryan_att,
14:24:09)
- iben: does this need a new feature?
(bryan_att,
14:24:25)
- ebrjohn: ferenc is working on that in the lab;
issue may be using vxlan ports for another purpose than tenant
separation (bryan_att,
14:24:58)
- edwarnicke: the problem is how ODL using
forwarding tags; we are looking for hackarounds; but could someone
address the root cause? (bryan_att,
14:25:47)
- ebrjohn: we are looing at patches to OVS to fix
the root cause (bryan_att,
14:26:06)
- edwarnicke: we need avoid a pile of
hackarounds (bryan_att,
14:26:27)
- edwarnicke: we need a push/pop of vxlan headers
via openflow; it's a pure OVS issue (bryan_att,
14:28:17)
- ebrjohn: add to slides that Sam has
started (bryan_att,
14:29:08)
- edwarnicke: the slides sent to the listr
describe the need for a push/pop in openflow for vxlan (bryan_att,
14:29:54)
- ebrjohn: we have an arch doc; we should add
this to it (bryan_att,
14:30:10)
- ebrjohn: we have this in a google doc
now (bryan_att,
14:30:42)
- bryan_att: I can help put this in git as soon
as its ready so we can use the OPNFV toolchain (which we should be
using) (bryan_att,
14:31:06)
- https://docs.google.com/document/d/1aoKYiExPtzyt3whETszeGh9neSKk2FpqxhgbGa0dkG0/edit
OPNFV SFC Arch doc (ebrjohn,
14:31:31)
- iben: we are looking to pass the virtual NIC
into the VM, bypassing OVS; would that be a workaround? (bryan_att,
14:31:36)
- (calls for more description of the idea)
(bryan_att,
14:31:56)
- ACTION: edwarnicke
will add the actual path a packet will take from SFF to SF
(ebrjohn,
14:32:01)
- iben: SR-IOV as an example passes traffic
straight to the VMs; OVS would be in the VM rather than the
hypervisor (bryan_att,
14:32:33)
- edwarnicke: various questions come to mind but
not looking at that as an option (bryan_att,
14:32:56)
- sriov for openstack kilo https://wiki.openstack.org/wiki/SR-IOV-Passthrough-For-Networking
(iben,
14:33:15)
- rprakash: that prevents control/data plane
separation (bryan_att,
14:33:16)
- trozet: can we get someone to work on the rool
cause (push/pop) in OVS? (bryan_att,
14:33:36)
- ebrjohn: that would be great; but need to
coordinate with NSH patch upstreaming (bryan_att,
14:34:14)
- paulq: existing patches are ready to go any
day; they are done (bryan_att,
14:34:28)
- ebrjohn: OVS is picky; can just make it easier
by coordinating? (bryan_att,
14:34:59)
- alagalah: that's not so easy (bryan_att,
14:35:18)
- rprakash: (describes an issue with OVS and
SFC) (bryan_att,
14:36:22)
- (comments about IETF drafts and
implementations) (bryan_att,
14:39:44)
- bryan_att: re drafts, we have one about MPLS
based SFC; we need to ensure a common model and support in OpenStack
for the various SFC methods OPNFV will support (bryan_att,
14:40:19)
- ebrjohn: need volunteers to help trozet in his
implementation in Tacker (bryan_att,
14:41:00)
- ebrjohn: need more members helping out
(bryan_att,
14:41:28)
- rprakash: will help (bryan_att,
14:41:56)
- bryan_att: is the work on github? (bryan_att,
14:42:22)
- trozet: need to define a spec first for input
to OpenStack (bryan_att,
14:42:37)
- trozet: looking for help with code and spec;
code is currently in github but will move to OpenStack as soon as
the spec is accepted (bryan_att,
14:43:36)
- https://github.com/trozet/tacker/tree/SFC
(trozet,
14:43:44)
- https://github.com/trozet/python-tackerclient/tree/SFC
(trozet,
14:43:59)
- bryan_att: we will also help to the extent we
can (bryan_att,
14:44:02)
- https://github.com/trozet/tacker-horizon/tree/SFC
(trozet,
14:44:14)
- https://github.com/trozet/sfc-random/blob/master/tacker_sfc_walkthrough.txt
(trozet,
14:44:28)
- ebrjohn: haven't looked deeply into the
installer discussion (bryan_att,
14:44:50)
- bryan_att: we have to ensure the max support
for SFC by not focusing all effort on a single installer
(bryan_att,
14:46:16)
- trozet: engage with genesis project and put
requirement on the installers (bryan_att,
14:46:59)
- bryan_att: that process is new to me but sounds
OK - we need to ensure that it's shared with all projects
(bryan_att,
14:48:19)
- iben: concerned about the impact to installers
having to support all projects (bryan_att,
14:48:50)
- trozet: if an installer is part of B release
then it needs to support all requirements agreed in genesis
(bryan_att,
14:49:17)
- iben: how do we address project-specific needs;
likely need to work in parallel in some cases (bryan_att,
14:50:11)
- ebrjohn: we need a special version of OVS with
NSH patches; will that cause a problem with genesis? (bryan_att,
14:50:40)
- trozet: there are other examples of similar
requjirements; it needs to be sent to genesis (bryan_att,
14:51:11)
- trozet: there is a generic guideline for now;
send an email to the genesis alias and a requirement in Jira;
genesis will then vote on it as a requirement (bryan_att,
14:53:05)
- ACTION: trozet to
send out genesis wiki guidelines (ebrjohn,
14:53:22)
- bryan_att: we need to know how this requirement
relates to the C milestone (bryan_att,
14:54:01)
- ebrjohn: how do we choose an installer if
generis can't support our requirements for B (bryan_att,
14:54:34)
- iben: you can use your own scripts for
now (bryan_att,
14:54:55)
- iben: need to show a working setup in a lab;
that will lead to a gap assessment in genesis (bryan_att,
14:55:28)
- ebrjohn: installation of SFC in the E/// lab
will be brought to the meeting next week as an example (bryan_att,
14:57:19)
- trozet: devstack just needs the right features
and tacker to openstack (bryan_att,
14:57:37)
- ebrjohn: will get action points and wiki
updated (bryan_att,
14:58:22)
Meeting ended at 15:48:51 UTC
(full logs).
Action items
- ferenc will send a document explaining the results
- edwarnicke will add the actual path a packet will take from SFF to SF
- trozet to send out genesis wiki guidelines
Action items, by person
- trozet
- trozet to send out genesis wiki guidelines
- UNASSIGNED
- ferenc will send a document explaining the results
- edwarnicke will add the actual path a packet will take from SFF to SF
People present (lines said)
- bryan_att (85)
- ethfci (12)
- ebrjohn (10)
- trozet (7)
- iben (6)
- fbl (6)
- ChrisPriceAB (5)
- collabot (4)
- alagalah (2)
- SergeM (1)
- rprakash (1)
- sridhar_ram (1)
- radez (1)
- bouthors (1)
- afredette (1)
Generated by MeetBot 0.1.4.