HistomicsTK integration to HistomicsUI error

[elizaan]

I am trying to integrate histomicsTK with my histomisui deployment but being failed multiple times

docker-compose.yaml:
services:

____________________________________________________________________________

histomicsui-dev:
image: magicscan/histomicsui
build: histomicsui
profiles: ["dev"]
network_mode: "host"
volumes:
- ./girder-dev.cfg:/etc/girder.cfg
- ./homepage.md:/root/homepage.md
- ./README.md:/root/README.md
- ./setup.py:/root/setup.py
- ./docs:/root/docs
- ./scripts:/root/scripts
- ./yaml:/root/yaml
- ./large_image_source_openvisus/init.py:/usr/local/lib/python3.10/site-packages/large_image_source_openvisus/init.py
- ./large_image_source_openvisus/girder_source.py:/usr/local/lib/python3.10/site-packages/large_image_source_openvisus/girder_source.py
..........

---

Message queue for communication between Girder and the Worker

rabbitmq:
image: rabbitmq:3-management
profiles: ["dev", "production"]
restart: unless-stopped
network_mode: "host"

environment:
  - RABBITMQ_DEFAULT_USER=guest
  - RABBITMQ_DEFAULT_PASS=guest
volumes:
  - rabbitmq_data:/var/lib/rabbitmq

____________________________________________________________________________

Worker to run background analysis jobs like Slicer CLIs

girder-worker:
image: magicscan/histomicsui
build: histomicsui
profiles: ["dev"]
restart: unless-stopped
depends_on:
- rabbitmq
network_mode: "host"
volumes:
- /var/run/docker.sock:/var/run/docker.sock
- ./girder-dev.cfg:/etc/girder.cfg # ADDED: Same config as main service

environment:
  - GW_DIRECT_PATHS=true
  - C_FORCE_ROOT=1 # Needed for running as root in Docker
  # Use container name instead of localhost
  - CELERY_BROKER_URL=amqp://guest:guest@localhost:5672//
  - CELERY_BROKER_CONNECTION_RETRY_ON_STARTUP=True
  - CELERY_BROKER_CONNECTION_RETRY=True
  - CELERY_BROKER_CONNECTION_MAX_RETRIES=10

command: girder-worker -l info --concurrency=2

volumes:
rabbitmq_data:

girder-dev.cfg:

[global]
server.socket_host = "0.0.0.0"
server.socket_port = 8081
server.max_request_body_size = 1073741824

[database]
uri = "mongodb+srv://........"

[worker]

Tells Girder to use RabbitMQ to send and receive jobs

backend = "rabbitmq"
amqp_url = "amqp://guest:guest@localhost:5672/"
broker_connection_retry_on_startup = True
broker_connection_retry = True
broker_connection_max_retries = 10
broker_heartbeat = 30

[slicer_cli_web]

Tells the Slicer CLI plugin to send tasks to the Girder Worker

task_mode = "girder_worker"
task_timeout = 3600
girder_api_url = "http://localhost:8081/api/v1"
girder_token = ""

can not fix the error, help will be appreciated!

[manthey]

It looks like there is some of authentication error going on. From your screenshot, you are clearly logging in as an administrator but the failure message implies that the user failed to get access to the target folder where tasks are created.

Can you explicitly log out and log back in and try it again?

If that doesn't work, that error message mentions a user database ID and then says the user is None. Does that user exist by id?

[elizaan]

Hi David,
Yes, the user ID is flagged as 'admin' in the MongoDB database we have. I tried logging out and logging in as admin and re-running the job, and it kept failing (screenshot 1)

Then I tried another approach, from another existing user. In the system settings, I deliberately made the user admin and logged in as that user, and the job status is success now! (screenshot 2)

Any idea why this might be happening? I want to check how these two users differ in terms of privilege level. How can I check?

[manthey]

I don't have any idea what is occurring -- admin users should all be fully-permissioned for everything. When we add a slicer_cli_web module, it gets a folder and item structure in the default tasks collection unless you go out of your way to do otherwise. The Tasks collection and folders are usually created as publicly visible but owned by an admin user.

In the original error message there is a user id mentioned. Is that your first admin user's id?

[elizaan]

yes,that is the admin id.

also, I a. getting another error that I can not solve. When i ran the nuclear detection algorithm, it started the job but failed because it says file not found. But, I applied it on our files saved in mogodb and visualized in histomicsUI . I have mounted the physical location in docker-compose.yaml under the girder worker container.

I would ask for any simple demo video or a step-by-step documentation that will explain each step, from integrating HistomicsTK to running an analysis algorithm, or examples of docker-compose.yaml / Dockerfile will be great to understand the full setup. Our end goal is to add our custom algorithm, i found that is possible, but how that is not written anywhere.

[manthey]

What version are you on? There was a change in a component library (girder/slicer_cli_web#250) that may have relevance.

The whole system should work with the default docker-compose in https://github.com/DigitalSlideArchive/digital_slide_archive/tree/master/devops/dsa (it gets tested through CI). There is a document here: https://github.com/DigitalSlideArchive/HistomicsUI/blob/master/docs/slicer_cli_plugins.rst describing the slicer-cli-web packaging of algorithms, and there are a lot of examples that range in complexity from HistomicsTK itself, to https://github.com/DigitalSlideArchive/histomicstk-similarity, to https://github.com/girder/slicer_cli_web/tree/master/small-docker.

[elizaan]

All the versions are: