feat(config): provide function to dynamically assign resources based …

…on number of attempts (#380)

* feat(config):  provide function to dynamically assign resources based on number of attempts

* fix(quickstart): seperate lines between git clone and cd

* fix(config): fix typos in cpus and memory assignment function

docs/developer_guidelines.md

@@ -152,6 +152,25 @@ tuple env(FILE_ID), val("${output}"), val(params.LOG_LEVELS.INFO), file(".comman
         file(".command.out"), file(".command.err"), file(".command.log"), emit: logs
 ```
 
+
+### Resources
+
+Usually all processes get a label that is linked to the flavor provided in the resources section of the configuration file (e.g. `small`, `medium`, `large`).
+In cases where we can foresee that a processes might need more RAM depending on the size of the input data, it is also possible to use
+the getMemoryResources and getCPUsResources methods provided by the Utils class:
+
+Example:
+
+```
+memory { Utils.getMemoryResources(params.resources.small, "${sample}", task.attempt, params.resources) }
+
+cpus { Utils.getCPUsResources(params.resources.tiny.small, "${sample}", task.attempt, params.resources) }
+```
+
+One example where it might make sense to use this option is when a process uses `csvtk join` and uses per default a label with
+little RAM assigned.
+
+
 ### Time Limit
 
 Every process must define a time limit which will never be reached on "normal" execution. This limit is only useful for errors in the execution environment  

lib/Utils.groovy

@@ -147,4 +147,77 @@ class Utils {
       return 90
     }
   }
+
+  /*
+  *
+  * This method returns the number of vCPUs according to the output of the getResources method.
+  *
+  */
+  static Integer getCPUsResources(defaults, sample, attempt, resources){
+     return getResources(defaults, sample, attempt, resources)["cpus"]
+  }
+
+
+  /*
+  *
+  * This method returns RAM value according to the output of the getResources method.
+  *
+  */
+  static String getMemoryResources(defaults, sample, attempt, resources){
+     return getResources(defaults, sample, attempt, resources)["memory"] + ' GB'
+  }
+
+  /*
+  *
+  * This method returns the flavor with more RAM provided in the resources section 
+  * of the configuration file.
+  *
+  * defaults: This variable contains the resources label object provided to the process (e.g. small, medium, ...).
+  * sample: This variable contains the sample name.
+  * attempt: This variable contains the number of attempts to run the process.
+  * resources: This variable represents all resources provied in the configuration file.
+  *
+  */
+  static Map<String, Integer> getResources(defaults, sample, attempt, resources){
+
+     // get map of memory values and flavor names (e.g. key: 256, value: large)
+     def memoryLabelMap = resources.findAll().collectEntries( { [it.value.memory, it.key] });
+
+     // Get map of resources with label as key
+     def labelMap = resources.findAll()\
+		.collectEntries( { [it.key, ["memory" : it.value.memory, "cpus" : it.value.cpus ] ] })
+
+     // get memory values as list
+     def memorySet = memoryLabelMap.keySet().sort()
+
+     // If the memory value is not predicted do the following
+     def defaultCpus = defaults.cpus
+
+     def defaultMemory = defaults.memory
+
+     def defaultMemoryIndex = memorySet.findIndexOf({ it == defaultMemory })
+
+     def nextHigherMemoryIndex = defaultMemoryIndex + attempt - 1 
+
+
+     if(nextHigherMemoryIndex >= memorySet.size()){
+        // In case it is already the highest possible memory setting
+        // then try the label with the highest memory
+        println("Warning: Highest possible memory setting " \
+		+ " of the dataset " + sample + " is already reached." \
+                + " Retry will be done using a flavor with the highest possible RAM configuration")
+
+        def label = memoryLabelMap[memorySet[memorySet.size()-1]]
+        def cpus =  labelMap[label]["cpus"]
+        def ram =  labelMap[label]["memory"]
+
+        return ["cpus": cpus, "memory": ram];
+     } else {
+   	def label = memoryLabelMap[memorySet[nextHigherMemoryIndex]];
+        def cpus = labelMap[label]["cpus"];
+        def memory = labelMap[label]["memory"];
+
+	return ["cpus": cpus, "memory": memory];
+     }
+  }
 }

modules/annotation/module.nf

@@ -647,7 +647,9 @@ process pCount {
 **/
 process pCollectFile {
 
-    label 'highmemMedium'
+    memory { Utils.getMemoryResources(params.resources.small, "${sample}", task.attempt, params.resources) }
+
+    cpus { Utils.getCPUsResources(params.resources.small, "${sample}", task.attempt, params.resources) }
 
     tag "Sample: $sample, Database: $dbType"
 

modules/binning/processes.nf

@@ -26,7 +26,9 @@ process pGetBinStatistics {
 
     publishDir params.output, mode: "${params.publishDirMode}", saveAs: { filename -> getOutput("${sample}", params.runid, "${module}", "${binner}", filename) }
 
-    label 'tiny'
+    memory { Utils.getMemoryResources(params.resources.tiny, "${sample}", task.attempt, params.resources) }
+
+    cpus { Utils.getCPUsResources(params.resources.tiny, "${sample}", task.attempt, params.resources) }
 
     input:
     val(module)

scripts/test_quickstart.sh

@@ -1,6 +1,6 @@
 # This file will be referenced on the online wiki
 
-git clone [email protected]:metagenomics/metagenomics-tk.git \
+git clone [email protected]:metagenomics/metagenomics-tk.git
 cd metagenomics-tk
 
 make nextflow