#opendaylight-group-policy: GBP Policy Repo

Meeting summary

1. 1. Rob Adams has been blocked due to issues with how YANG/MD-SAL works, and the code generation.. net-net it took more time than he anticipated (alagalah, 16:09:51)
   2. readams is sharing on ghangout his preliminary yang model. This has not been pushed to the repo (we are still getting our Jenkins environment up etc etc... so COMMITs not yet possible) (alagalah_, 16:11:25)
   3. filename is endpoint.yang (alagalah_, 16:11:54)
   4. readams says his yang is a simple database of all the endpoints stored in the registry, and replicating APIC, there are two copies of every endpoint, 1 keyed my L2 namespace and MAC and 1 keyed by L3 namespace and IP (alagalah_, 16:12:53)
   5. Renderers/Or anything that knows about endpoints, are responsible for writing endpoint information to the repository. Or anything exernal like OpenStack that knows endpoints that could write into it. (alagalah_, 16:13:58)
   6. using an RPC call that understands it has to write to both places with a timestamp (for consistency) (alagalah_, 16:14:15)
   7. L2/3 namespace id, tenant id, group id, condition id... L3 address = namespace+IP addr tuple (alagalah_, 16:14:58)
   8. implemented as a LIST (alagalah_, 16:15:14)
   9. keyed by L2 address (ie namespace + MAC) and similarly for L3 address (ie namespace + IP ADDR) (alagalah_, 16:15:43)
   10. has RPCs "registerEndpoint" and "unregisterEndpoint" (alagalah_, 16:15:57)
   11. this is a preliminary strawman (alagalah, 16:17:15)
   12. some of the MD-SAL examples on the wiki (ie Toaster) do not work as advertised, which the MD-SAL team is aware of and is working towards fixing (alagalah, 16:17:22)
   13. will implement sharding once that code appears (alagalah, 16:17:40)
   14. hemanth.... (alagalah, 16:17:47)
   15. does the generated code have methods to store/access from the md-sal data-store (hemanthravi, 16:18:25)
   16. or do you need to need to fill in the methods (hemanthravi, 16:18:38)
   17. readams answers hemanthravi question (couldn't pick it up) but is talking about instanceIDs as key into tree (alagalah, 16:18:38)
   18. readams not sure if there is enforcement of endpoint containing key data.... may need validation code when renderer writes? Or ensure that renderers are always part of GBP and hence part of a controlled architecture (alagalah, 16:19:17)
   19. readams feels that we may need a wrapper API around the generated API ... (alagalah, 16:19:50)
   20. alagalah feels we should work with the MD-SAL team to ensure that the generated APIs are useful, as ideally we shouldn't use wrappers. (alagalah_, 16:20:56)
   21. alagalah feels we should work with the MD-SAL team to ensure that the generated APIs are useful, as ideally we shouldn't use wrappers. (alagalah_, 16:20:57)
   22. i'll get back on this later today (hemanthravi, 16:21:57)
   23. s3wong can contribute code to this group, and around 20% time a week (s3wong, 16:22:28)
   24. do you mean just for this subgroup or for GBP as a whole? (readams, 16:22:47)
   25. alagalah clarifies this is for this sub-group (alagalah_, 16:23:04)
   26. I'll contribute 100% of my time until this is done later this week (readams, 16:23:29)
   27. readams points out that this isn't really onerous and we don't have a lot more to talk about at this point until we have a working prototype (alagalah_, 16:25:30)
   28. https://plus.google.com/hangouts/_/calendar/ZHZvcmtpbkBub2lyb25ldHdvcmtzLmNvbQ.ilarq9u79ea3u3a5q0hhg1gd8s (alagalah_, 16:27:04)
   29. readams says instanceID == URI ... (alagalah_, 16:28:36)
   30. regXboi still has concerns... alagalah may have mis-worded minutes (alagalah_, 16:28:53)
   31. Point has been clarified, regXboi is happy... and thats a good thing (alagalah_, 16:29:11)

Action items

1. (none)

People present (lines said)

1. alagalah_ (34)
2. regXboi (23)
3. alagalah (12)
4. readams (4)
5. hemanthravi (4)
6. dvorkinista (3)
7. odl_meetbot (3)
8. s3wong (2)
9. banix (1)
10. aboch (1)