#opnfv-promise: promise
Meeting started by r-mibu at 14:01:31 UTC
(full logs).
Meeting summary
- roll call (r-mibu, 14:01:44)
- Ildiko Vancsa (ildikov,
14:01:57)
- Bertrand Souville (bertys,
14:02:02)
- Bryan Sullivan (bryan_att,
14:02:14)
- Ryota Mibu (r-mibu,
14:02:42)
- Gerald (GeraldK,
14:09:41)
- Way of working: there are Jira tickets, but the
way forward is not clear. should we organize into
short/mid/long-term issues? (GeraldK,
14:14:55)
- proposal to better utilize features of Jira to
do so (GeraldK,
14:15:43)
- inviting people from other projects to weekly
meetings is of general interest and should be kept (GeraldK,
14:18:30)
- requirements track should stay a couple of
month ahead of the implementation track (GeraldK,
14:21:14)
- Lab sharing with IPv6 project (GeraldK, 14:21:56)
- Peter Lee (corenova,
14:24:21)
- Bin: VM is used as IPv6 router; integration
efforts with ODL trying to solve gaps (GeraldK,
14:24:46)
- Bin: after this has been addressed,
infrastructure could be used for Promise (GeraldK,
14:25:39)
- IPv6 project is using Spirent community
lab (bertys,
14:26:53)
- Peter: would be great if Promise could send our
requirements (ODL version, number of nodes, ...) to Infra-Team and
they set it up (without Promise getting "empty" hardware
(GeraldK,
14:27:25)
- JIRA PROMISE-6 (r-mibu, 14:40:06)
- ACTION: Peter to
update Promise-6 with additional contributors (corenova,
14:40:55)
- team should review and provide +1/+2 to the
gerrit commit (corenova,
14:41:57)
- https://gerrit.opnfv.org/gerrit/#/c/1345/
(r-mibu,
14:42:04)
- JIRA PROMISE-19 (r-mibu, 14:42:37)
- Mirantis would be interested to re-open
Blazar (GeraldK,
14:43:12)
- related to Promise long term
architecture (GeraldK,
14:44:42)
- there is long-term effort to extract Nova
scheduler. we can start utilizing the latest status of this project
and update Blazar accordingly (GeraldK,
14:45:49)
- this extracting is *in the best case* done in
Mitaka-release, but most likely in later release (GeraldK,
14:46:49)
- shim-layer approach seems the only way we can
achieve our goals in few months. we should give high prio to solve
southbound I/F gaps. but long-term architecture should not be
neglected. (GeraldK,
14:51:57)
- Peter will make initial draft (GeraldK,
14:53:31)
- allocation part might be difficult. Peter is
assuming all allocation request go via shim layer to make this work.
we need to write down this big picture. (GeraldK,
14:54:06)
- promise software architecture diagram (shim layer) (GeraldK, 14:55:34)
- this is not part of weekly Promise meeting, but
additional technical meeting (GeraldK,
14:56:06)
- https://jira.opnfv.org/secure/attachment/10501/opnfv-promise.png
(GeraldK,
14:56:54)
- top-left: package "yan/opnfv-promise" has
dependency on other packages and schema files (GeraldK,
15:00:11)
- "yangforge" provides several internal
types (GeraldK,
15:01:11)
- ComputeElement, VMImage, etc are instances of
ResourceElement; ServerInstance is instance of
ResourceInstance (GeraldK,
15:03:20)
- Promise instance needs to be tied to OpenStack.
two ways to implement this: 1) tight integration between OpenStack
and Promise by direct mapping logic and 2) a new layer encapsulating
OpenStack (GeraldK,
15:05:35)
- OpenStack package can even capture different
versions of OpenStack, e.g. different API versions (GeraldK,
15:06:54)
- the OpenStack package can also be (re-)used for
other (OPNFV) projects (GeraldK,
15:08:31)
- there is not yet a basic demo for the OpenStack
package (GeraldK,
15:11:15)
- this is planned for this week (GeraldK,
15:12:20)
- i.e. initial transaction to add a
ResourceProvider and populate it with all related data (GeraldK,
15:13:09)
- idea is to capture all details, except things
done via API extensions. (GeraldK,
15:13:42)
- Ildiko: who stores the reservations? Peter:
idea is that Promise shim-layer stores all this information
required, encapsulating a lot of existing information available
downstream. (GeraldK,
15:15:05)
- Ildiko: are you planning to include "quotas"?
Peter: quota system would be "query"-based, but in the first step
not allow to change the quotas in OpenStack. (GeraldK,
15:16:33)
- shim-layer is kind of emulating OpenStack thus
allowing to implement Promise features in the shim-layer
(GeraldK,
15:17:27)
- this requires a feedback-loop to keep states
synchronized (best in real-time). some delay might be acceptable for
Promise (this has to be analyzed/discussed in more detail)
(GeraldK,
15:20:31)
- Gerald: Promise might be okay with a more
coarse-grained information. (GeraldK,
15:21:03)
- Peter: this also depends on the usage of the
reservation system, whether you do reservations/changes on a per
minute or per day frequency (GeraldK,
15:21:52)
- Peter could need help to model missing things
in the yangforge schemas; also the code logic for reservation is
missing. (GeraldK,
15:28:41)
- Ildiko: do we need all those models to evaluate
the approach as such? No, not all the models, but the basic ones,
e.g. storage/network could be done later. (GeraldK,
15:29:54)
- Ildiko: are the flows etc already captured,
e.g. in the requirement document? (GeraldK,
15:30:56)
- No, we don't have these explicit flows.
(GeraldK,
15:32:00)
- Peter is quite confident that this shim-layer
will allow for the basic reservation features among different
resource types; but if there are specific behaviours, we may not be
able to sufficiently model this (at least in the initial
version). (GeraldK,
15:36:05)
- Peter will prepare and share sequence diagrams
around the reservation flows (GeraldK,
15:40:41)
- team should then support Peter to continue from
these baseline diagrams to extend to other flows/logic (GeraldK,
15:43:14)
- discussion on documentation
priorities/needs. (GeraldK,
15:55:20)
Meeting ended at 15:55:56 UTC
(full logs).
Action items
- Peter to update Promise-6 with additional contributors
People present (lines said)
- GeraldK (51)
- r-mibu (12)
- collabot (11)
- corenova (4)
- bryan_att (2)
- bertys (2)
- ildikov (1)
Generated by MeetBot 0.1.4.