Topic: Smart Data Transaction for CPS Weekly Meeting
Weekly Meeting
Progress report
Image analysis function is implemented, now testing
Can probably move on to automatic actions based on image analysis in next sprint
Colin will communicate basic spec for automatic actions function to the team later today
Weekly Meeting
Meeting canceled
Weekly Meeting
Progress report
Compile of images based on r32.6.1 is working
There is an intermittent issue where CUDA APIs stop responding after 1.5 or so hours of use
The problem above will be investigated in the background while we continue with the coding and tests
If coding completes next week, move on to integration testing of the completed modules (rule-based actuation can wait)
Colin is on vacation next week so the weekly meeting will be canceled
Weekly Meeting
Progress report
Have had success building containers on l4t-base:r32.4.3
Want to upgrade to r32.6.1 to match the version on the Jetson devices
Once that is done, move on to quickly try out the OpenCV and/or CUDA interfaces we're going to use
Check status next week and decide how to progress, especially while Colin is on vacation
Weekly Meeting
Progress report
CUDA-enabled programs work when directly compiled on Jetson, but not in containers.
Seems to be a version mismatch problem. The CUDA version in the containers we are using is 11.7, and the Jetson version is 10.2.
Unfortunately, there doesn't seem to be ARM support for CUDA base containers from NVidia at version 10.2 of CUDA.
Looking for alternatives. Worst case, we may need to build a CPU-only OpenCV container and leave GPU support for later.
Keep trying to get the GPU version working until sprint end , and if we can't get it going, switch to CPU-only for next sprint.
Weekly Meeting
Progress report
Coding and unit test of data sharing complete
Currently examining
How to build containers that contain the CUDA and OpenCV libraries plus drivers for GPU access
How to enable GPU access for containers on the NVidia Jetson
Design of a simple shape-counting app using Go-CV in progress
We are having trouble calling CUDA-enabled Go-CV in the container, with what looks like a driver and library version mismatch. Investigation is ongoing.
Investigation and design of the app scheduled to take the next sprint, until
Weekly Meeting
Progress report
Design of data sharing pretty much complete, maybe a few changes from review comments.
Coding and testing should be complete today, depending on effect of design review comments.
About 1wk ahead of schedule, we should be able to look into image analysis. Issues include:
What, exactly, are we going to do for analysis?
How to use the GPU in the system (can it be done from a container on K8s?)
Will need to sync up on those issues and any others that come up.
Implementation of modified device-camera-go for SDT is done (will review)
Test case for snapshot reading from camera devices is done (will review)
Next sprint (ends ) combines design and implementation/test for sharing snapshot data between nodes using the sync-app
Colin will be on summer vacation from to . We'll work on a plan for work done in that period.
Weekly Meeting
Progress report
Investigations of device-camera-go and device-onvif-go are done. We have confirmed we can get snapshots from our camera devices with either device service.
device-onvif-go has a problem with AuthMethod "digest"
Will open a separate issue and put that aside for the moment
Since we have slightly more confidence with device-camera-go, we will continue with that service for now
Hoping to finish the service integration and CT steps by the end of this sprint ()
Weekly Meeting
Progress report
Retrying image pulls worked well enough to build local device-camera-go
Not clear if there is any good solution for overriding the camera IP at run time
Had trouble getting device-camera-go to retrieve images
DeviceResources does not appear to be populated?
Having trouble debugging (e.g. because of image rebuild problems)
Colin will look into the problem as well to see if he can offer support
Release 6 PR is on the marketing page (Smart Data Transactions for CPS section)
Weekly Meeting
Progress report
Mid-sprint 2, focus on design of image acquisition
Some trouble with Docker rate-limiting of anonymous image pulls
Will investigate possible solutions offline
It should be possible to override the camera IP address with an environment variable setting, possibly, which would eliminate the need to rebuild the image
The Python algorithms for using OpenCV don't seem to translate well to Go, producing less acceptable results
Share updates in the ticket with details
IP cameras installed in test environment
Will be moving files in the shared directories between 11:45 and 13:15 JST today
Tina requests if we can add something for SDT to the Release 6 Marketing, hopefully before the NAOSS (6/20)
Weekly Meeting
Progress report
Reviewed function block definitions. Looks ok.
Next sprint (until 6/21) focuses on design of camera device and image reading, sprint after that (until 7/5) will be coding/CT for same.
Install IP cameras on Thursday
Weekly Meeting
Meeting was held via Fujitsu's Teams account due to problems with Zoom
Progress report
Architecture diagram and document updated on Wiki
Working on functional block list
Check the image analysis example in issue #63 on the local GitLab
Weekly Meeting
Progress report
Initial sprint plan and backlog complete.
Architecture diagram and function block diagram updates basically complete. Review today.
Running ahead of schedule. Pick up new tasks from the backlog for next sprint as we go.
Cameras will be added to the lab setup as soon as we get the AC adaptors for the cameras.
Meeting with Tina tomorrow at 9AM JST to discuss new blueprint proposal.
From Tina: Reminder to update the R6 marketing case studies.
From Fukano-san: Have updated the LF Edge Commercial Ecosystem presentation.
Weekly Meeting
R7 Backlog list format ready. Will hold another meeting this week to discuss with JFTT and fill it in.
Sprint planning should be complete for next week's meeting.
We will be using the weekly meeting for progress reports from JFTT, and every two weeks a sprint review.
We will be adding camera devices in the lab setup soon. Will let the team know when they are ready.
Weekly Meeting
Zoom login problems; meeting held via Teams
Release 6 is apparently delayed as other blueprint teams are still updating their release requirements
SDT work for release 7 will kick off tomorrow. Shared access details for the Fujitsu internal lab environment.
Will be contacting Jim Xu from the Upstream & Downstream Sub-committee to discuss impact of EdgeX and Kong issues on the blueprint's inclusion in release 6.
Weekly Meeting
Progress report:
Phase 5:
K8s security: Confirmed API protected by certificate and user account
Docker: Enabled TLS security for private repository
EdgeX security measures have encountered some difficulty because of problems, mainly from using k8s instead of docker-compose (e.g. edgex-go #3852 and #3851)
Currently stuck on an issue with Kong reporting a DNS error looking up "localhost"
Will try contacting mailing list and/or TSC chairs to see if anyone has any experience with this kind of error
Prepare a demo, tentatively scheduled for 2/18
Weekly Meeting
Progress report:
Phase 4 complete, temperature sensor data relayed via LoRa to MQTT broker, tests implemented
Phase 5: Security measures for K8s, EdgeX, and Docker under investigation, planned implementation by 2/8
Device SDK Working Group meeting: LoRa device service to be submitted to edgex-examples
New Jetson devices (x4) arrived 1/20.
Phase 4 will complete today, moving on to Phase 5 (security, CD) and document preparation
Weekly Meeting
Progress report:
Phase 4, LoRa and sensor integration, in progress, planned completion 1/25
Tasks for release 6 listed in progress report (slide 9): Documents, API form, BluVal + security checks are main items
The full lab setup is configured and now we are working on device support and a pass-through app on the sensor nodes
We'd like to build a second lab setup but need to add two more Raspberry Pi devices for that, also LoRa dongles if possible (meet with Fukano-san on Wednesday to discuss further)
Device and Device SDK Working Group meeting was moved to 1/24