docker

Copy data to RCP cluster

To copy data to the RCP cluster, need to have first copied the data on the NAS-RCP. Either the data is already in the raw data catalog and thus store on the NAS, or it needs to be copied. A folder exist for that purpose, it is located inside common/ForRCP. In this folder every user creates is own folder, for example clement and creates two sub-folders input and output. Resulting in something like common/ForRCP/clement/input and common/ForRCP/clement/output. In the input folder we will add all the data that we need to be copied to the cluster for processing, in the output folder, all the data produced during processing will be copied back (for example segmentation masks, model, text file). The cluster has a high performance scratch space of 2.5TB, but nothing is backup, and can be deleted when space is needed (for example when many TBs are sitting there for long time w/o accessing them). The only machine that has an access to both the lab NAS and the RCP scratch is the JumpHost. To connect to it, simply replace <USERNAME> with your gaspar and type in a terminal:

ssh <USERNAME>@haas001.rcp.epfl.ch

On this machine, our lab NAS is mounted as /mnt/upoates/collaborative and the scratch (accessible to the GPU cluster) as /mnt/upoates/scratch. If you want to list the files for RCP, do something like:

ls /mnt/upoates/collaborative/ForRCP

What we want at this point is to replicate all the files that we have inside /mnt/upoates/collaborative/ForRCP on the cluster's scratch /mnt/upoates/scratch. For that let's create the same structure.

id

uid=267988(helsens) gid=11349(UPOATES-StaffU)

docker buildx build --platform linux/amd64 --tag registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.3 --build-arg LDAP_GROUPNAME=UPOATES-StaffU --build-arg LDAP_GID=11349 --build-arg LDAP_USERNAME=helsens --build-arg LDAP_UID=267988 .

docker build . --tag registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1 --tag registry.rcp.epfl.ch/upoates-helsens/cellpose-env:latest
--build-arg LDAP_GROUPNAME=UPOATES-StaffU
--build-arg LDAP_GID=11349
--build-arg LDAP_USERNAME=helsens
--build-arg LDAP_UID=267988

docker build . --tag registry.rcp.epfl.ch/upoates-helsens/plantseg-env:v0.1
--build-arg LDAP_GROUPNAME=UPOATES-StaffU
--build-arg LDAP_GID=11349
--build-arg LDAP_USERNAME=helsens
--build-arg LDAP_UID=267988

try the container docker run -it registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1 bash docker run -it registry.rcp.epfl.ch/upoates-helsens/plantseg-env:v0.1 bash

docker run -it --mount type=bind,source=/Volumes/upoates,target=/mydata registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1 bash

to register to harbor docker login registry.rcp.epfl.ch

docker push registry.rcp.epfl.ch/upoates-helsens/cellpose-env:latest docker push registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1

get kube config the file curl https://wiki.rcp.epfl.ch/public/files/kube-config.yaml -o ~/.kube/config && chmod 600 ~/.kube/config

set the cluster runai config cluster rcp-caas-prod

Login to runAI runai login

list projects runai list project

runai config project upoates-helsens

PVC -> upoates-scratch kubectl get pvc -n runai-upoates-helsens NAME STATUS VOLUME CAPACITY ACCESS MODES STORAGECLASS AGE home Bound upoates-home-helsens 100Ti RWX 9d upoates-scratch Bound upoates-scratch-helsens 100Ti RWX 9d

kubectl get secrets for private images NAME TYPE DATA AGE my-registry-secret kubernetes.io/dockerconfigjson 1 25h

runai submit
--name cellpose-test
--image registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1
--gpu 0.5
--environment MY_ENV_VAR="A test ENV variable"
--existing-pvc claimname=upoates-scratch,path=/scratch
--existing-pvc claimname=home,path=/home/helsens
--command
-- /bin/bash -ic "sleep 6000"

connect to the job container when running runai bash my-demo-job

runai submit \
--name cellpose-test-6
--image registry.rcp.epfl.ch/upoates-helsens/cellpose-env:v0.1
--gpu 1.
--existing-pvc claimname=upoates-scratch,path=/scratch
--command
-- /usr/bin/python3 /home/helsens/3d_segmentation/3d_cellpose.py /scratch/data/feyza/pos1_Channel1_cropped /scratch/data/feyza/cellpose_training/2d/models/nuclei_h2b

runai submit \
--name plantseg-test-1
--image registry.rcp.epfl.ch/upoates-helsens/plantseg-env:v0.1
--gpu 1.
--existing-pvc claimname=upoates-scratch,path=/scratch
--command
-- conda init
-- bash

use this to login with different credentials --run-as-uid UID
--run-as-gid GID \