anvil-rstudio-* images do not have WORKSPACE_* environment variables defined

[mtmorgan]

terra-jupyter-* images have useful environment variables WORKSPACE_BUCKET, WORKSPACE_NAME, WORKSPACE_NAMESPACE.

Can these be made available on the rstudio images? WORKSPACE_NAME and _BUCKET are from https://github.com/DataBiosphere/terra-docker/blob/b849abb7e9a4a5afd5dbfc67dad1b63b7ea1bc6a/terra-jupyter-base/scripts/kernel/kernel_bootstrap.sh#L16 ; I don't know where WORKSPACE_NAMESPACE comes from

[rtitle]

@mtmorgan yes this is a known limitation -- we have a ticket to address it: https://broadworkbench.atlassian.net/browse/IA-1849

We currently have these env vars defined in the RStudio image: https://github.com/DataBiosphere/leonardo/blob/develop/http/src/main/resources/init-resources/rstudio-docker-compose.yaml#L12

WORKSPACE_NAME and WORKSPACE_BUCKET are a bit trickier because the same runtime can be used across multiple workspaces. So it's not sufficient to set them once at container creation time; we need to switch them when the workspace context changes. Jupyter does this with a hook in the kernel startup script; but we need to find a different solution for RStudio. @kyuksel had worked on a technical design for this, but we still need to implement it.

[mtmorgan]

Thanks; WORKSPACE_NAMESPACE doesn't seem to make it through to our image, but maybe that's our problem.

> system2("whoami")
rstudio
> sysenv = Sys.getenv()
> sort(names(sysenv))
 [1] "CLICOLOR_FORCE"                  "DISPLAY"                         "EDITOR"                          "GIT_ASKPASS"                    
 [5] "HOME"                            "LANG"                            "LD_LIBRARY_PATH"                 "LN_S"                           
 [9] "LOGNAME"                         "MAKE"                            "MPLENGINE"                       "PAGER"                          
[13] "PATH"                            "R_BROWSER"                       "R_BZIPCMD"                       "R_DOC_DIR"                      
[17] "R_GZIPCMD"                       "R_HOME"                          "R_INCLUDE_DIR"                   "R_LIBS"                         
[21] "R_LIBS_SITE"                     "R_LIBS_USER"                     "R_PAPERSIZE"                     "R_PDFVIEWER"                    
[25] "R_PLATFORM"                      "R_PRINTCMD"                      "R_RD4PDF"                        "R_SESSION_TMPDIR"               
[29] "R_SHARE_DIR"                     "R_STRIP_SHARED_LIB"              "R_STRIP_STATIC_LIB"              "R_SYSTEM_ABI"                   
[33] "R_TEXI2DVICMD"                   "R_UNZIPCMD"                      "R_ZIPCMD"                        "RMARKDOWN_MATHJAX_PATH"         
[37] "RS_RPOSTBACK_PATH"               "RSTUDIO"                         "RSTUDIO_CONSOLE_COLOR"           "RSTUDIO_CONSOLE_WIDTH"          
[41] "RSTUDIO_HTTP_REFERER"            "RSTUDIO_PANDOC"                  "RSTUDIO_R_MODULE"                "RSTUDIO_R_PRELAUNCH_SCRIPT"     
[45] "RSTUDIO_R_VERSION_LABEL"         "RSTUDIO_SESSION_STREAM"          "RSTUDIO_USER_IDENTITY"           "RSTUDIO_WINUTILS"               
[49] "SED"                             "SSH_ASKPASS"                     "TAR"                             "TERM"                           
[53] "TERRA_R_PLATFORM"                "TERRA_R_PLATFORM_BINARY_VERSION" "USER"           
> sysenv[["TERRA_R_PLATFORM"]]
[1] "anvil-rstudio-bioconductor"

I don't think the WORKSPACE_NAMESPACE is enough to discover the WORKSPACE_NAME through the terra or leonardo APIs?

[rtitle]

No - WORKSPACE_NAMESPACE is the same as GOOGLE_PROJECT. Looks like you don't have that set either 🤔

All these should be set currently: https://github.com/DataBiosphere/leonardo/blob/develop/http/src/main/resources/init-resources/rstudio-docker-compose.yaml#L12

[mtmorgan]

@rtitle any insight on solving this? it would help enable use of RStudio during the Galaxy hiatus