#acumos-meeting: Architecture Committee
Meeting started by farheen_att at 14:04:59 UTC
(full logs).
Meeting summary
- Agenda (farheen_att, 14:06:31)
- we want to spend time on release status, sprint
4 blockers, kafka pub/sub, planning for Demeter, new feature
ideas. (farheen_att,
14:07:26)
- Add the self-release of artifacts for mavin or
docker feature for gerrit. (farheen_att,
14:08:09)
- Chris Lott has done some self-release process.
This is a good topic to cover. (farheen_att,
14:08:40)
- Chris - Self release is only good for jar
files (farheen_att,
14:08:53)
- Justin - LF fixed the bugs and we are releasing
docker also. (farheen_att,
14:09:11)
- Release Status (farheen_att, 14:10:02)
- We want to focus on blockers for the pending
items in sprint 4. Are we on target? (farheen_att,
14:10:37)
- Federation We ware on track Murali is having a
separate discussion with Ting. (farheen_att,
14:11:06)
- Guy - we had a pass thru feature for ONAP that
has been implemented. More work regarding adding meta data. For
that we need to synchronize with Tommy. It is not blocking the demo
but it should be finished. It's work that has to be done by
onboarding or common services team. (farheen_att,
14:12:40)
- we will have a demo that will benefit the
Acumos community. (farheen_att,
14:13:18)
- O-RAN October 8th demo (farheen_att, 14:13:33)
- Guy will have it ready next week. (farheen_att,
14:13:50)
- ONS demo (farheen_att, 14:14:01)
- Guy - There was a resource that was stalled.
Nat said they have gotten someone but I have not heard from
TechM. (farheen_att,
14:14:25)
- Nat - I got the input from Amol
offshore. (farheen_att,
14:14:55)
- ACTION: Nat will
align the resource to join the next onboarding scrum call
(farheen_att,
14:16:31)
- ACTION: Murali put it
as a blocker waiting for a resource. One week task for a skilled
developer. (farheen_att,
14:17:54)
- Jenkins Model Deployment (farheen_att, 14:18:48)
- Bryan - Was released yesterday. Mark it as
deliver to testing team. It is available on the TechM K8
server. (farheen_att,
14:19:38)
- Ken - no issues with alignment on
deployment. (farheen_att,
14:21:00)
- when can we have the demo on that? (farheen_att,
14:21:07)
- when the portal is integrated with portal
3.0.15. (farheen_att,
14:21:43)
- Ken - we have to wait for the portal team to
complete? (farheen_att,
14:22:00)
- yes (farheen_att,
14:22:02)
- in your absense is there anyone who can back
you up? (farheen_att,
14:22:23)
- No (farheen_att,
14:22:26)
- open issue in deployment (farheen_att,
14:23:00)
- ACTION: Vasu talk to
Santosh and come up with an ETA. (farheen_att,
14:23:20)
- is there a portal user story? (farheen_att,
14:23:29)
- ACTION: Murali update
jira epic to 3433. (farheen_att,
14:24:03)
- ACTION: Murali check
ETA for deployment. This is important to track at that level to set
expectations. (farheen_att,
14:24:39)
- MLWB (farheen_att, 14:27:42)
- Sayee - working through issues no
blockers (farheen_att,
14:27:53)
- Licensing (farheen_att, 14:28:00)
- - no blockers (farheen_att,
14:28:11)
- Justin we are dependent on SV 1.10 to get our
SV checking on platform done. We are tracking down the abandoned
commit. (farheen_att,
14:28:53)
- 3024 is the issue to track. (farheen_att,
14:29:15)
- Acumos-3005 should be updated. (farheen_att,
14:29:43)
- ACTION: Murali update
Acumos-3005 (farheen_att,
14:29:57)
- Acumos-3065 is open, why? (farheen_att,
14:30:10)
- there may be documentation user stories.
(farheen_att,
14:30:23)
- Upgrade to java 11. (farheen_att, 14:30:51)
- who are the teams working on this? (farheen_att,
14:30:59)
- ACTION: Murali follow
up on the upgrade to java 11. (farheen_att,
14:31:14)
- Ken do you see testing related concerns. Have
you started on the delivered cycle. (farheen_att,
14:31:38)
- Yes, Bryans release does not have portal
integration. (farheen_att,
14:32:18)
- keep the progress on the delivered items to
test. (farheen_att,
14:32:53)
- Pub/Sub mechanism support thru Kafka. (farheen_att, 14:33:53)
- Bryan - do homework to understand what is meant
and done in other open source projects in the same space.
(farheen_att,
14:34:21)
- we change component address to connect to a
message bus and know how to subscribe on that bus. Incoming API
subjects on the bus. It is a more flexible and cloud native
approach. (farheen_att,
14:35:15)
- People need to understand the technology to
understand the value of this approach to the Acumos project. People
need to come to the table with an understanding of the
technology. (farheen_att,
14:35:59)
- Is there a performance hit? and do we become
dependent on the bus. If the bus goes down does everything
stop? (farheen_att,
14:36:24)
- latency of transactions being filled need to be
considered. Openstack would not be based on the same concept for
real production clouds. (farheen_att,
14:37:15)
- There is a spring-cloud to help us with this.
There are many message systems such as Kafka nanakube. (farheen_att,
14:37:52)
- To what extent will we go? (farheen_att,
14:38:02)
- this enables a service discovery a lot
smoother. Cost is you have to replace a jar. (farheen_att,
14:38:25)
- wouldn't we put it out as a platform service
and allow those who want to use it? (farheen_att,
14:38:49)
- There is a roadmap to adoption. Everything
will have to be changed from APIs to message bus. (farheen_att,
14:39:41)
- suggestion Initially build one component that
listens to the bus and reroutes that to the appropriate web API
calls. (farheen_att,
14:40:11)
- over time as components switch over they can
stop the central bus. (farheen_att,
14:40:47)
- Kafka are for continuous message. Such as
knowing the status. Or save something on a database. We don't have
to go through Kafka (farheen_att,
14:41:31)
- Yes, CDS is like that. It is well served by the
rest interface no need to apply. (farheen_att,
14:42:03)
- we do need to solve the service discovery
issue. When you bring up a system things aren't connected
properly. (farheen_att,
14:42:30)
- Service discovery is not the primary use
case. (farheen_att,
14:42:54)
- primary benefit is asynchronous use cases. Any
component that wants this mechanism can switch. Those who do not
don't have to. (farheen_att,
14:43:33)
- you are not going to solve this complex
spaghetti code system by taking a half step. You want every API to
report to the bus. (farheen_att,
14:44:25)
- what would benefit non lazy components?
(farheen_att,
14:45:07)
- Coupling . I can start a service and no other
component cares. Project needs to adopt. (farheen_att,
14:45:46)
- wouldn't bringing down a service change the
other components? (farheen_att,
14:46:26)
- remote procedure calls. (farheen_att,
14:46:49)
- To accommodate a-synchronous would require
recoding in a lot of places. (farheen_att,
14:47:29)
- java programmers typically don't program that
way. (farheen_att,
14:47:45)
- it removes dependencies of trigger
events. (farheen_att,
14:48:10)
- If you have a piece of code that says get the
list of models from this user. Put a call back function that
updates something. that will be tricky. (farheen_att,
14:49:00)
- luckily that paradigm familiar to web
programmers. (farheen_att,
14:49:17)
- Good discussion. Let's do our homework on
this. Once we have the intial module the licensing team is building
we will adopt pub/sub (farheen_att,
14:50:23)
- Demeter ideas (farheen_att, 14:51:16)
- enter ideas into the etherpad. (farheen_att,
14:51:28)
- Self Serve release (farheen_att, 14:52:09)
- https://docs.releng.linuxfoundation.org/projects/global-jjb/en/latest/jjb/lf-release-jobs.html
(farheen_att,
14:52:17)
- This will remove the long wait from LF to grant
permission. (farheen_att,
14:52:57)
- you do one commit for the mavin and one commit
in docker. Stage the release as you normally do. pre-req you must
have the release job added to your jenkins job configuration. These
are the two jobs in jenkins (farheen_att,
14:54:13)
- two configs in jenkins some projects have
multiple jars and single. on a side note we are ways to release
each jar independenly. If you have all of these sub modules in
maven defined. (farheen_att,
14:55:08)
- we want to see the releases we have in the
past. (farheen_att,
14:59:29)
- in ONAP we are not creating multiple version
files. (farheen_att,
14:59:54)
- ONAP history and release artifacts will address
that concern. (farheen_att,
15:00:15)
- use acumos/ and update the tags. (farheen_att,
15:03:01)
- Thank you for stepping up and sharing with the
PTLs. Expect Eric Ball from LF. LF turn around time is 24 hours.
If you follow this process you can release your artifacts
yourself. (farheen_att,
15:04:51)
- 3433 3430 are in a delivered state. (farheen_att, 15:05:25)
- verify with the portal team that the expected
items for SV and jenkins workflow are completed. (farheen_att,
15:05:46)
- which commit were they delivered. (farheen_att,
15:06:07)
- ACTION: Vasu double
check commits to Murali and Bryan. (farheen_att,
15:06:41)
Meeting ended at 15:06:46 UTC
(full logs).
Action items
- Nat will align the resource to join the next onboarding scrum call
- Murali put it as a blocker waiting for a resource. One week task for a skilled developer.
- Vasu talk to Santosh and come up with an ETA.
- Murali update jira epic to 3433.
- Murali check ETA for deployment. This is important to track at that level to set expectations.
- Murali update Acumos-3005
- Murali follow up on the upgrade to java 11.
- Vasu double check commits to Murali and Bryan.
People present (lines said)
- farheen_att (95)
- collabot` (3)
Generated by MeetBot 0.1.4.