Feature/custom package (#272)

* Added custom package script

* Added feature custom download

* Fixed typo

* Fixed directory for installation

* Fixed full folder directory

* Add dependencies and fix pattern

* Fix pattern not found

* Added repo

* Switching to devtools

* Fixing devtools install with directory

* Fix in for merger.R

* Working cluster custom packages

* Removed printed statements

* Working on custom docs

* Custom packages sample docs

* Fixed typo in azure files typo

* Fixed typos based on PR

Author: brnleehng

R/cluster.R

@@ -151,6 +151,10 @@ makeCluster <-
         "wget https://raw.githubusercontent.com/Azure/doAzureParallel/",
         "master/inst/startup/install_bioconductor.R"
       ),
+      paste0(
+        "wget https://raw.githubusercontent.com/Azure/doAzureParallel/",
+        "master/inst/startup/install_custom.R"
+      ),
       "chmod u+x install_bioconductor.R",
       installAndStartContainerCommand
     )

R/commandLineUtilities.R

@@ -123,6 +123,7 @@ dockerRunCommand <-
       dockerOptions <-
         paste(
           dockerOptions,
+          "-e AZ_BATCH_NODE_SHARED_DIR=$AZ_BATCH_NODE_SHARED_DIR",
           "-e AZ_BATCH_TASK_ID=$AZ_BATCH_TASK_ID",
           "-e AZ_BATCH_JOB_ID=$AZ_BATCH_JOB_ID",
           "-e AZ_BATCH_TASK_WORKING_DIR=$AZ_BATCH_TASK_WORKING_DIR",

docs/20-package-management.md

@@ -38,29 +38,37 @@ You can install packages by specifying the package(s) in your JSON pool configur
 }
 ```
 
+## Installing Packages per-*foreach* Loop
+
+You can also install cran packages by using the **.packages** option in the *foreach* loop. You can also install github/bioconductor packages by using the **github** and **bioconductor" option in the *foreach* loop. Instead of installing packages during pool creation, packages (and its dependencies) can be installed before each iteration in the loop is run on your Azure cluster.
+
+### Installing a Github Package
+
+doAzureParallel supports github package with the **github** option. 
+
+Please do not use "https://github.com/" as prefix for the github package name above.
+
 ## Installing packages from a private GitHub repository
 
-Clusters can be configured to install packages from a private GitHub repository by setting the __githubAuthenticationToken__ property. If this property is blank only public repositories can be used. If a token is added then public and the private github repo can be used together.
+Clusters can be configured to install packages from a private GitHub repository by setting the __githubAuthenticationToken__ property in the credentials file. If this property is blank only public repositories can be used. If a token is added then public and the private github repo can be used together.
 
 When the cluster is created the token is passed in as an environment variable called GITHUB\_PAT on start-up which lasts the life of the cluster and is looked up whenever devtools::install_github is called.
 
+Credentials File for github authentication token
+``` json
+{
+  ...
+  "githubAuthenticationToken": "",
+  ...
+}
+
+```
+
+Cluster File
 ```json
 {
     {
-    "name": <your pool name>,
-    "vmSize": <your pool VM size name>,
-    "maxTasksPerNode": <num tasks to allocate to each node>,
-    "poolSize": {
-        "dedicatedNodes": {
-            "min": 2,
-            "max": 2
-        },
-        "lowPriorityNodes": {
-            "min": 1,
-            "max": 10
-        },
-        "autoscaleFormula": "QUEUE"
-    },
+    ...
     "rPackages": {
         "cran": [],
         "github": ["<project/some_private_repository>"],
@@ -71,10 +79,18 @@ When the cluster is created the token is passed in as an environment variable ca
 }
 ```
 
-_More information regarding github authentication tokens can be found [here](https://help.github.com/articles/creating-a-personal-access-token-for-the-command-line/)_
+_More information regarding github authentication tokens can be found [here](https://help.github.com/articles/creating-a-personal-access-token-for-the-command-line/)
 
-## Installing Packages per-*foreach* Loop
-You can also install cran packages by using the **.packages** option in the *foreach* loop. You can also install github/bioconductor packages by using the **github** and **bioconductor" option in the *foreach* loop. Instead of installing packages during pool creation, packages (and its dependencies) can be installed before each iteration in the loop is run on your Azure cluster.
+### Installing Multiple Packages
+By using character vectors of the packages, 
+
+```R
+number_of_iterations <- 10
+results <- foreach(i = 1:number_of_iterations,
+                  .packages=c('package_1', 'package_2'),
+                  github = c('Azure/rAzureBatch', 'Azure/doAzureParallel'),
+                  bioconductor = c('IRanges', 'Biobase')) %dopar% { ... }
+```
 
 To install a single cran package:
 ```R
@@ -94,7 +110,6 @@ number_of_iterations <- 10
 results <- foreach(i = 1:number_of_iterations, github='azure/rAzureBatch') %dopar% { ... }
 ```
 
-Please do not use "https://github.com/" as prefix for the github package name above.
 
 To install multiple github packages:
 ```R
@@ -114,7 +129,7 @@ number_of_iterations <- 10
 results <- foreach(i = 1:number_of_iterations, bioconductor=c('package_1', 'package_2')) %dopar% { ... }
 ```
 
-## Installing Packages from BioConductor
+## Installing a BioConductor Package
 The default deployment of R used in the cluster (see [Customizing the cluster](./30-customize-cluster.md) for more information) includes the Bioconductor installer by default. Simply add packages to the cluster by adding packages in the array.
 
 ```json
@@ -134,17 +149,27 @@ The default deployment of R used in the cluster (see [Customizing the cluster](.
         },
         "autoscaleFormula": "QUEUE"
     },
+    "containerImage:" "rocker/tidyverse:latest",
     "rPackages": {
         "cran": [],
         "github": [],
         "bioconductor": ["IRanges"]
     },
-    "commandLine": []
+    "commandLine": [],
+    "subnetId": ""
     }
 }
 ```
 
-Note: Container references that are not provided by tidyverse do not support Bioconductor installs. If you choose another container, you must make sure that Biocondunctor is installed.
+Note: Container references that are not provided by tidyverse do not support Bioconductor installs. If you choose another container, you must make sure that Bioconductor is installed.
+
+## Installing Custom Packages
+doAzureParallel supports custom package installation in the cluster. Custom packages installation on the per-*foreach* loop level is not supported. 
+
+For steps on installing custom packages, it can be found [here](../samples/package_management/custom/README.md).
+
+Note: If the package requires a compilation such as apt-get installations, users will be required
+to build their own containers.
 
-## Uninstalling packages
+## Uninstalling a Package
 Uninstalling packages from your pool is not supported. However, you may consider rebuilding your pool.

inst/startup/install_custom.R

@@ -0,0 +1,49 @@
+args <- commandArgs(trailingOnly = TRUE)
+
+sharedPackageDirectory <- file.path(
+  Sys.getenv("AZ_BATCH_NODE_SHARED_DIR"),
+  "R",
+  "packages")
+
+tempDir <- file.path(
+  Sys.getenv("AZ_BATCH_NODE_STARTUP_DIR"),
+  "tmp")
+
+.libPaths(c(sharedPackageDirectory, .libPaths()))
+
+pattern <- NULL
+if (length(args) > 1) {
+  if (!is.null(args[2])) {
+    pattern <- args[2]
+  }
+}
+
+devtoolsPackage <- "devtools"
+if (!require(devtoolsPackage, character.only = TRUE)) {
+  install.packages(devtoolsPackage)
+  require(devtoolsPackage, character.only = TRUE)
+}
+
+packageDirs <- list.files(
+  path = tempDir,
+  full.names = TRUE,
+  recursive = FALSE)
+
+for (i in 1:length(packageDirs)) {
+  print("Package Directories")
+  print(packageDirs[i])
+
+  devtools::install(packageDirs[i],
+                    args = c(
+                      paste0(
+                        "--library=",
+                        "'",
+                        sharedPackageDirectory,
+                        "'")))
+
+  print("Package Directories Completed")
+}
+
+unlink(
+  tempDir,
+  recursive = TRUE)

inst/startup/merger.R

@@ -18,7 +18,9 @@ batchJobPreparationDirectory <-
   Sys.getenv("AZ_BATCH_JOB_PREP_WORKING_DIR")
 batchTaskWorkingDirectory <- Sys.getenv("AZ_BATCH_TASK_WORKING_DIR")
 taskPackageDirectory <- paste0(batchTaskWorkingDirectory)
-clusterPackageDirectory <- paste0(Sys.getenv("AZ_BATCH_NODE_SHARED_DIR", "/R/packages"))
+clusterPackageDirectory <- file.path(Sys.getenv("AZ_BATCH_NODE_SHARED_DIR"),
+                                     "R",
+                                     "packages")
 
 libPaths <- c(
   taskPackageDirectory,

samples/azure_files/azure_files_cluster.json

@@ -20,5 +20,8 @@
   },
   "commandLine": [
     "mkdir /mnt/batch/tasks/shared/data",
-    "mount -t cifs //<STORAGE_ACCOUNT_NAME>.file.core.windows.net/<FILE_SHARE_NAME> /mnt/batch/tasks/shared/data -o vers=3.0,username=<STORAGE_ACCOUNT_NAME>,password=<STORAGE_ACCOUNT_KEY>==,dir_mode=0777,file_mode=0777,sec=ntlmssp"]
+    "mount -t cifs //<STORAGE_ACCOUNT_NAME>.file.core.windows.net/<FILE_SHARE_NAME> /mnt/batch/tasks/shared/data -o vers=3.0,username=<STORAGE_ACCOUNT_NAME>,password=<STORAGE_ACCOUNT_KEY>,dir_mode=0777,file_mode=0777,sec=ntlmssp",
+    "wget https://raw.githubusercontent.com/Azure/doAzureParallel/feature/custom-package/inst/startup/install_custom.R",
+    "docker run --rm -v $AZ_BATCH_NODE_ROOT_DIR:$AZ_BATCH_NODE_ROOT_DIR  -e AZ_BATCH_NODE_ROOT_DIR=$AZ_BATCH_NODE_ROOT_DIR -e AZ_BATCH_NODE_STARTUP_DIR=$AZ_BATCH_NODE_STARTUP_DIR --rm -v $AZ_BATCH_NODE_ROOT_DIR:$AZ_BATCH_NODE_ROOT_DIR -e AZ_BATCH_NODE_SHARED_DIR=$AZ_BATCH_NODE_SHARED_DIR	-e AZ_BATCH_NODE_ROOT_DIR=$AZ_BATCH_NODE_ROOT_DIR  -e AZ_BATCH_NODE_STARTUP_DIR=$AZ_BATCH_NODE_STARTUP_DIR rocker/tidyverse:latest Rscript --no-save --no-environ --no-restore --no-site-file --verbose $AZ_BATCH_NODE_STARTUP_DIR/wd/install_custom.R /mnt/batch/tasks/shared/data"
+    ]
 }

samples/azure_files/readme.md

@@ -12,4 +12,4 @@ This samples shows how to update the cluster configuration to create a new mount
 
 For large data sets or large traffic applications be sure to review the Azure Files [scalability and performance targets](https://docs.microsoft.com/en-us/azure/storage/common/storage-scalability-targets#scalability-targets-for-blobs-queues-tables-and-files).
 
-For very large data sets we recommend using Azure Blobs. You can learn more in the [persistent storage](../../docs/23-persistent-storage.md) and [distrubuted data](../../docs/21-distributing-data.md) docs.
+For very large data sets we recommend using Azure Blobs. You can learn more in the [persistent storage](../../docs/23-persistent-storage.md) and [distributing data](../../docs/21-distributing-data.md) docs.

samples/package_management/bioconductor.r renamed to samples/package_management/bioconductor/bioconductor.r

@@ -1,4 +1,4 @@
-#Please see documentation at docs/20-package-management.md for more details on packagement management.
+#Please see documentation at docs/20-package-management.md for more details on package management.
 
 # import the doAzureParallel library and its dependencies
 library(doAzureParallel)

samples/package_management/package_management_cluster.json renamed to samples/package_management/bioconductor/package_management_cluster.json

@@ -0,0 +1,32 @@
+## Installing Custom Packages
+doAzureParallel supports custom package installation in the cluster. Custom packages are R packages that cannot be hosted on Github or be built on a docker image. The recommended approach for custom packages is building them from source and uploading them to an Azure File Share.
+
+Note: If the package requires a compilation such as apt-get installations, users will be required
+to build their own containers.
+
+### Building Package from Source in RStudio
+1. Open *RStudio*
+2. Go to *Build* on the navigation bar
+3. Go to *Build From Source*
+
+### Uploading Custom Package to Azure Files
+For detailed steps on uploading files to Azure Files in the Portal can be found
+[here](https://docs.microsoft.com/en-us/azure/storage/files/storage-how-to-use-files-portal)
+
+### Notes
+1) In order to build the custom packages' dependencies, we need to untar the R packages and build them within their directories. By default, we will build custom packages in the *$AZ_BATCH_NODE_SHARED_DIR/tmp* directory. 
+2) By default, the custom package cluster configuration file will install any packages that are a *.tar.gz file in the file share. If users want to specify R packages, they must change this line in the cluster configuration file.
+
+Finds files that end with *.tar.gz in the current Azure File Share directory 
+``` json
+{
+  ...
+  "commandLine": [
+    ...
+    "mkdir $AZ_BATCH_NODE_STARTUP_DIR/tmp | for i in `ls $AZ_BATCH_NODE_SHARED_DIR/data/*.tar.gz | awk '{print $NF}'`; do tar -xvf $i -C $AZ_BATCH_NODE_STARTUP_DIR/tmp; done",
+    ...
+    ]
+}
+```
+3) For more information on using Azure Files on Batch, follow our other [sample](./azure_files/readme.md) of using Azure Files
+4) Replace your Storage Account name, endpoint and key in the cluster configuration file