In the VM Environment setting for RStudio, I don’t see docker, or samtools installed on the Application configuration. Does it mean I have to install those packages myself? That would mean I have to use the startup script, which is another step users have to learn (or relyon AI to give answers).
Is my understanding correct? I wonder the reason why docker is not installed? How do users use dsub or workflow if AnVIL does not provide docker in the base environment?
Hi @truckload,
I don’t see docker, or samtools installed on the Application configuration. Does it mean I have to install those packages myself?
Yes, if it is not available, you will have to install it, use a startup script, or use a custom environment based on the AnVIL RStudio image.
How do users use dsub or workflow if AnVIL does not provide docker in the base environment?
If you are hoping to run workflows from your RStudio environment, we recommend checking out the Bioconductor AnVIL Packages, as shown in this vignette: Running an AnVIL workflow within R
Please let us know if this helps!
Ava
Thank you! That is good to know. Currently the default setting for RStudio (R4.5.1, Bioconductor 3.21, Python 2.12.3) does not have AnVILGCP installed. To install the AnVILGCP, a substantial number of package need to be updated. Once a user installs the updated packages, it seems one shall not delete this RStudio application, unless they want to re-install/update all the packages.
please give me the advice on:
- If I successfully install the AnVILGCP, I can delete the VM environment, but keep the persistent disk. I can start a RStudio app again, because all the R code is store on my persistent disk. Or whenever I delete the Rstudio, I have to re-install AnVILGCP.
- To use R in workflow, which way is better? Use my own startup script or keep Rstudio app / Workspace persistent disk in order to have AnVILGCP available.
I also review the link “Running an AnVIL workflow within R”. It seems like that AnVILGCP provide wrapper function to run a workflow. It still need the workflow file to run. When running a workflow, to be able to execute the R command, the docker image need to be provided (with the required R packages). It seems to me that with AnVILGCP, I can avoid the workflow web interface, instead, I will use R interface to manage the workflow. However, I still need to have a docker image.
You have reference AnVIL RStudio image, it seems that I can use those docker file on workflow. However, I need my analysis-specific R packages, such as qqman(). That seems to be missing from the pre-built docker image. So I need to modify the Dockerfile to build my custom docker image.
For a batch job, a docker with RStudio and bioconductor packages seems to be a little overkill, especially I just need a couple of R packages for my job… So RStudio environment is really more fit for interactive job …
If you were saying that I can use this as RStudio application configuration, this is not an option in the Application configuration drop down menu.
To install the AnVILGCP, a substantial number of package need to be updated.
When installing AnVILGCP, are you selecting y rather than n? You may not need to update all the related packages. Selecting n may be substantially faster.
If I successfully install the AnVILGCP, I can delete the VM environment, but keep the persistent disk. I can start a RStudio app again, because all the R code is store on my persistent disk.
Keeping the persistent disk should also keep the installed packages. They should be available once you restart the computing environment without having to reinstall.
To use R in workflow, which way is better? Use my own startup script or keep Rstudio app / Workspace persistent disk in order to have AnVILGCP available.
This is up to you. There are some costs associated with retaining a persistent disk, but there is also the time it would take to create a custom startup script or Docker image.
You have reference AnVIL RStudio image, it seems that I can use those docker file on workflow. However, I need my analysis-specific R packages, such as qqman(). That seems to be missing from the pre-built docker image. So I need to modify the Dockerfile to build my custom docker image.
Yes, any custom image you provide should be based on AnVIL RStudio image in order for it to work on Terra. You will need to modify it if you want other software/packages.
If you were saying that I can use this as RStudio application configuration, this is not an option in the Application configuration drop down menu.
You will need to select “Custom Environment” and link to your custom image. This custom image should be based on AnVIL RStudio image.
1 Like