How to offboard researchers and labs:¶
Offboarding researchers:¶
Revoke SSH access:¶
To revoke SSH access, remove the researchers authorized_keys file:
# become the researcher
sudo su {researcher_name}
# remove the authorized_keys file
rm ~/.ssh/authorized_keys
Kill processes:¶
Kill any processes the researcher still has running:
# identify running processes:
sudo lsof /data/{researcher_name}
# kill running processes:
sudo lsof /data/{researcher_name} | awk '{print $2}' | tail -n +2 | xargs -r kill -9
Cancel Slurm jobs:¶
Cancel any jobs the researcher may still have running:
# to view all their jobs:
squeue --user={researcher_name}
# to cancel all their jobs:
scancel --user={researcher_name}
Archive the home directory:¶
Create an archive of the researchers home directory.
Note: That in the following step where we upload the tarball to a object store bucket we can only upload objects with a maximum size of 10TB. If the researchers home directory is greater than 10TB you will need to create multiple smaller tarballs that will be uploaded separately.
Upload data to bucket:¶
Use the upload.py script in the scripts directory to upload the tarball to object storage bucket:
Note: Update the bucket name in the script. Use the researchers name as the name,
# edit the script:
vim /home/opc/scripts/upload.py
# change the name of the bucket to the name of the researcher:
bucket_name = "{researcher_name}"
# upload the tarball
sudo /home/opc/scripts/upload.py /data/{researcher_name}.tar
After you've uploaded the researchers data to the bucket we can provide a download link for them to use. On the OCI Console, navigate to the bucket we just uploaded to and look for the Pre-Authenticated Requests tab on the left hand menu. On that tab, you can create a download link for the researcher to use.
Remove data from cluster:¶
# remove the tarball
sudo rm /data/{researcher_name}
# remove the home directory
sudo rm -r /data/{researcher_name}
Remove Slurm user:¶
Remove Cluster user:¶
Offboarding labs:¶
Offboard all researchers:¶
If the lab has any researcher that are still on the cluster follow the above steps for offboarding researchers for each one.