- Mais recentes
- Mais votos
- Mais comentários
First you have to decide do you want Microsoft R or Open R, Microsoft R I believe has some optimizations but is a pain to get to run in docker (they hard code the UIDs resulting in conflicts depending on the host node OS). I did Microsoft R for a client a few years back, it was a bit painful. Unfortunately I can't share that file/work.
If Open R works for your use case then why not just start from r-base? https://hub.docker.com/_/r-base/
I'd recommend a Dockerfile that is specific for your use case, installing data.table / other packages post batch startup isn't ideal, so FROM r-base generally we copied a install.R script to the docker image and run it in r-script to install everything. Then saved that image to ECR and referenced it in the batch jobs.
If you are running a lot of jobs you may want to also look into the ECS cluster to change the caching on the computer environment since the images will be quite large.
Conteúdo relevante
- AWS OFICIALAtualizada há um mês