#opendaylight-ovsdb: ovsdb_weekly_call

Meeting started by tbachman at 20:00:26 UTC (full logs).

Meeting summary

  1. agenda (tbachman, 20:00:45)
    1. https://meetings.opendaylight.org/opendaylight-ovsdb/2015/ovsdb_weekly_call/opendaylight-ovsdb-ovsdb_weekly_call.2015-01-13-20.01.html Minutes from last week’s meeting (tbachman, 20:01:43)

  2. trello board (tbachman, 20:05:04)
    1. shague says that OVSDB has joined the Lithium simultaneous release (tbachman, 20:06:15)
    2. shague has been looking into pulling out the neutron services and putting it in it’s own project; will submit as project proposal to TSC (tbachman, 20:06:49)
    3. users of neutron should be able to point to the new bundle (tbachman, 20:07:03)
    4. edwarnicke has been working on creating a yang model to put in OVSDB for MD-SAL (tbachman, 20:07:42)
    5. michal_rehak has merged the nicira extension work in the openflowplugin (tbachman, 20:08:11)
    6. tbachman and flaviof are working on getting this set up in OVSDB (tbachman, 20:08:20)
    7. plaurin_ is working with mestery on the neutron side; something strange with the unit tests (tbachman, 20:10:45)
    8. sharad is working on DPDK, and has a few issues that he’s trying to resolve (tbachman, 20:12:18)
    9. sharad asks how to expose the capabilities of the switch and the host in ODL (tbachman, 20:12:40)
    10. sharad is looking into how to expose this to openstack (tbachman, 20:12:53)
    11. sharad says the OVS folks are proposing to change the OVSDB schema to add a capabilities field to bridges to support different port types, which could be provided to openstack (tbachman, 20:14:32)
    12. edwarnicke says we could look into adding to the model, as that’s how you can get this information represented in ODL so that it can be used for muiltiple apps (tbachman, 20:15:47)

  3. MD-SAL model for OVSDB (tbachman, 20:18:50)
    1. edwarnicke says his understanding that there are 3 kinds of things that users of OVSDB want to do (tbachman, 20:22:08)
    2. the first one is that people might want to create a switch-like thing (tbachman, 20:22:22)
    3. the second is to CRUD a port for the datapath (tbachman, 20:22:33)
    4. the third is to CRUD a tunnel for a datapath (tbachman, 20:22:42)
    5. edwarnicke says that to do this, he’s created two models: an OVSDB model, and an OVSDB switch model (tbachman, 20:23:13)
    6. https://git.opendaylight.org/gerrit/#/c/14261/ Gerrit that edwarnicke is referencing (tbachman, 20:24:07)
    7. the ovsdb model creates augmentations to the nodes in topology (tbachman, 20:26:21)
    8. one of the model elements is to augment the node with an OVSDB node (tbachman, 20:26:44)
    9. another one of the model elements is to augment the node with bridges (tbachman, 20:26:59)
    10. edwarnicke says a lot of this was taken from SFC’s models (tbachman, 20:27:12)
    11. shague asks why we can’t use the OVSDB schema to generate the model (tbachman, 20:31:07)
    12. edwarnicke says there are two things to look at; you have to dynamic translation of schema to yang, which is doable, but a lot more work (tbachman, 20:31:27)
    13. edwarnicke says we’re describing high-level information about what we’re doing (tbachman, 20:31:42)
    14. anil asks if this nb would talk to the AD-SAL plugin (tbachman, 20:34:35)
    15. edwarnicke says this NB uses the library but does not talk to the AD-SAL lib (tbachman, 20:34:47)
    16. shague says that there’s a layer above the library, and now there’d be two different paths into the library (tbachman, 20:35:30)
    17. edwarnicke says the plugin currently pulls in the entire AD-SAL (tbachman, 20:35:40)
    18. shague says there are other places in OVSDB that use the AD-SAL (tbachman, 20:35:52)
    19. edwarnicke says there’s not maintenance for the AD-SAL going forward (tbachman, 20:36:05)
    20. shague says that you can’t just use the library today, as there are other pieces invovled (tbachman, 20:36:23)
    21. shague says that you have to include a couple of other pieces (tbachman, 20:36:36)
    22. edwarnicke says this already functions, asks why it required pulling in the entire AD-SAL (tbachman, 20:36:51)
    23. shague says you could extract out just the library, but you’d break the other users (tbachman, 20:37:03)
    24. there’s some discrepency between shague and edwarnicke about what the VTN project is going to be using for Lithium (tbachman, 20:37:48)
    25. LuisGomez asks about the API to use the v3 APIs (tbachman, 20:42:22)
    26. edwarnicke says he’s not opposed to that, but there are requests for the use cases that he’s created the yang models for (tbachman, 20:42:43)
    27. this is also driven by the M2 deadline (tbachman, 20:43:10)
    28. edwarnicke says that it’s possible that both ends of the tunnel might be on an OVSDB managed node, and another might be on a non-OVSDB managed node (tbachman, 20:44:05)
    29. https://lists.opendaylight.org/pipermail/ovsdb-dev/2015-January/000980.html email thread that talks about overlay APIs being looked into (tbachman, 20:47:15)
    30. flaviof points out that there’s a bunch of OVS-like things that might be represented (e.g. trunked links) (tbachman, 20:49:59)
    31. edwarnicke says he doesn’t mind binding things together that are intrinsicly bound together, but is against artificial glomming together of things in a NB model (tbachman, 20:51:11)
    32. the tunnel.yang model creates a bunch of tunnel types, which are used to augment a node with the tunnel type (tbachman, 20:53:16)
    33. anil asks what the link signifies (tbachman, 20:53:34)
    34. it depends on the topology that you’re looking at. (tbachman, 20:54:07)
    35. anil says if you’re talking about physical links, there could be many tunnels on that link (tbachman, 20:54:19)
    36. edwarnicke says these are tunnel links, not physical links (tbachman, 20:54:28)
    37. Eric Multanen asks where things like the DPDK port types would fit in the model (tbachman, 20:57:03)
    38. Eric Multanen says you set the type when you create the port (tbachman, 20:57:57)
    39. edwarnicke says he thinks this would be an attribute of the port that you’d augment (tbachman, 20:58:16)
    40. edwarnicke asks if port types are a capability of the bridge or the OVSDB instance (tbachman, 20:58:59)
    41. Eric Multanen says maybe both - it’s part of an instance of a bridge that you create (tbachman, 20:59:15)
    42. edwarnicke says in the bridge you could get a list of supported port types, which is extensible (tbachman, 20:59:28)
    43. edwarnicke asks if port type is really an attribute of the port, or of two port types put together (tbachman, 21:00:02)
    44. sharad asks if we should be looking at using the MD-SAL model for the Lithium time frame (tbachman, 21:04:18)
    45. shague says the real issue is what is our flow (tbachman, 21:04:39)
    46. edwarnicke asks if Eric is looking to use OVSDB to query the supported port types on the bridge (tbachman, 21:05:56)
    47. Eric says that’s the case, but the ports are actually created by openstack, via libvirt (tbachman, 21:06:32)
    48. shague says openstack queries the OVSDB database now, so why replicate this in OVSDB? (tbachman, 21:07:52)
    49. edwarnicke asks if this is a good first pass at the model (tbachman, 21:08:26)
    50. shague says this is a good start (tbachman, 21:08:31)


Meeting ended at 21:15:09 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. tbachman (83)
  2. odl_meetbot (4)
  3. vishnoianil (4)
  4. plaurin_ (1)
  5. shague (0)
  6. flaviof (0)


Generated by MeetBot 0.1.4.