vhive-serverless / vSwarm

A suite of representative serverless cloud-agnostic (i.e., dockerized) benchmarks
MIT License
46 stars 20 forks source link

Error re-making Docker image: video-analytics benchmark #334

Open cg592 opened 1 year ago

cg592 commented 1 year ago

Describe the bug I am trying to (1) run the video-analytics benchmark on other videos that I have uploaded to the video_streaming/reference directory, and also (2) modify the source code to print out additional information. To do so, I believe I need to re-make the Docker image. However, the make all-image command fails when trying to make a new Docker image. The error message mentions needing the GOPRIVATE_KEY I am running the benchmarks locally following the instructions here: https://github.com/vhive-serverless/vSwarm/blob/main/docs/running_locally.md

To Reproduce

cd vSwarm/benchmarks/video-analytics/
make all-image

Expected behavior Docker image build to complete successfully.

Logs Output:

DOCKER_BUILDKIT=1 docker build \
-t vhiveease/video-analytics-decoder:latest \
--target decode \
--secret id=GOPRIVATE_KEY \
-f Dockerfile \
../../
could not parse secrets: [id=GOPRIVATE_KEY]: failed to stat GOPRIVATE_KEY: stat GOPRIVATE_KEY: no such file or directory
make: *** [Makefile:34: decoder-image] Error 1

On the instructions page linked above, it says to reach out to obtain the GOPRIVATE_KEY?

cg592 commented 1 year ago

Any updates on this?

ustiugov commented 1 year ago

@shyamjesal could you please take a look or give advice?

shyamjesal commented 1 year ago

Hi @cg592, Building video analytics benchmark requires access to a private repo. Hence the error. For a workaround, you can try replacing the file in the container. I have not tried it recently, but it should be possible.

ustiugov commented 1 year ago

@cg592 what you could do is to comment out or stub the xdt dependencies in the source code. it's not an ideal solution but should not be very difficult to do.

Another option is to actually upstream more verbose but configurable logging, e.g., by adding --debug option in the YAML files of the workload, this would be to the benefit of all the users. we would re-build & push containers with your changes