now we can rely on Terra itself for fast localization in minimap2 task

add output tracking how long the alignment step took

Author: SHuang-Broad

wdl/pipelines/ONT/Preprocessing/ONTFlowcellWGSuBAM.wdl

@@ -97,6 +97,7 @@ workflow ONTFlowcellWGSuBAM {
 
     output {
         String last_processing_date = today.yyyy_mm_dd
+        String aln_wallclock_time   = ALN.total_runtime
 
         File aligned_bam = FinalizeAlignedBam.gcs_path
         File aligned_bai = FinalizeAlignedBai.gcs_path

wdl/tasks/Alignment/AlignONTWGSuBAM.wdl

@@ -38,6 +38,8 @@ workflow AlignONTWGSuBAM {
     output {
         File aligned_bam = aBAM
         File aligned_bai = aBAI
+
+        String total_runtime = select_first([sumRuntimes.total_runtime, Minimap2.wallclocktime])
     }
 
     Map[String, String] ref_map = read_map(ref_map_file)
@@ -113,6 +115,7 @@ workflow AlignONTWGSuBAM {
             disk_type = if ('LOCAL' == aln_disk_type) then 'SSD' else 'LOCAL',
             timeout_hours = if (100 < size(uBAM, "GiB")) then 10 else 5  # heuristic: longer wait for bigger BAMs
         }
+        call sumRuntimes { input: runtimes = MapShard.wallclocktime }
     }
     if (emperical_bam_sz_threshold >= ceil(size(uBAM, "GiB"))) {
         call AR.Minimap2 {
@@ -212,4 +215,38 @@ task OneOffHandleSpacesInRGLine {
         disks: "local-disk 10 HDD"
         docker: "gcr.io/cloud-marketplace/google/ubuntu2004:latest"
     }
-}
+}
+
+task sumRuntimes {
+    input {
+        Array[String] runtimes
+    }
+
+    output {
+        String total_runtime = read_string("total.txt")
+    }
+    command <<<
+        total_minutes=0
+
+        # Read each runtime and sum up
+        while IFS= read -r line; do
+            # Extract hours and minutes from format "XX hours, YY minutes"
+            hours=$(echo "$line" | grep -oP '\d+(?=H)')
+            minutes=$(echo "$line" | grep -oP '\d+(?=M)')
+
+            # Convert to total minutes and add
+            total_minutes=$((total_minutes + hours * 60 + minutes))
+        done < ~{write_lines(runtimes)}
+
+        # Convert back to hours and minutes
+        final_hours=$((total_minutes / 60))
+        final_minutes=$((total_minutes % 60))
+
+        # Output in same format with padding
+        printf "%02dH%02dM\n" $final_hours $final_minutes > total.txt
+    >>>
+    runtime {
+        disks: "local-disk 10 HDD"
+        docker: "gcr.io/cloud-marketplace/google/ubuntu2004:latest"
+    }
+}

wdl/tasks/Alignment/AlignReads.wdl

@@ -24,10 +24,7 @@ task Minimap2 {
         descrpiton: "A wrapper to minimap2 for mapping & aligning (groups of) sequences to a reference. Note that this only works for reads belonging to a single readgroup."
     }
     parameter_meta {
-        reads: {
-            desciption: "query sequences to be mapped and aligned",
-            localization_optional: true
-        }
+        reads:                "query sequences to be mapped and aligned"
         reads_file_basenames: "basenames of the BAM files, note this includes the extention (e.g. .bam, .fasta.gz, etc) of files"
         longer_ont_read_hint: "hint that the input reads are longer (~>20kb N50) ONT reads; this is used to bump the memory to the task"
         ref_fasta:        "reference fasta"
@@ -57,14 +54,18 @@ task Minimap2 {
     Int mm2_threads = cpus - 2
 
     command <<<
-        set -euxo pipefail
+    set -euxo pipefail
+    start_time=$(date +%s)
+
+        results_dir=$(pwd)
 
         FILE="~{reads[0]}"
 
         ############
         # localize data (much faster than Cromwell)
         mkdir -p reads_dir
-        time gcloud storage cp ~{sep=' ' reads} /cromwell_root/reads_dir/
+        mv ~{sep=' ' reads} \
+           reads_dir/
         ls reads_dir
 
         cd reads_dir
@@ -169,20 +170,28 @@ task Minimap2 {
 
         ############
         # move results up
-        mv ~{prefix}.bam ~{prefix}.bam.bai /cromwell_root
+        mv ~{prefix}.bam ~{prefix}.bam.bai "${results_dir}/"
+
+    end_time=$(date +%s)
+    total_seconds=$((end_time - start_time))
+    hours=$((total_seconds / 3600))
+    minutes=$(((total_seconds % 3600) / 60))
+    printf "%02dH%02dM\n" $hours $minutes > "${results_dir}/wallclocktime.txt"
     >>>
 
     output {
         File aligned_bam = "~{prefix}.bam"
         File aligned_bai = "~{prefix}.bam.bai"
+
+        String wallclocktime = read_string("wallclocktime.txt")
     }
 
     #########################
     RuntimeAttr default_attr = object {
         cpu_cores:          cpus,
         mem_gb:             mem,
         disk_gb:            disk_size,
-        preemptible_tries:  3,
+        preemptible_tries:  1,
         max_retries:        0,
         docker:             "us.gcr.io/broad-dsp-lrma/lr-minimap2:2.26-gcloud"
     }

wdl/tasks/Utility/BAMutils.wdl

@@ -1645,7 +1645,6 @@ task SplitNameSortedUbam {
         read_cnt: "number of reads in the uBAM; providing this will reduce run time."
         n_reads: "desired number of reads per split; mutually exclusive with n_files"
         n_files: "desired number of split files; mutually exclusive with n_reads"
-        uBAM: { localization_optional: true }
     }
     input {
         File uBAM
@@ -1665,24 +1664,20 @@ task SplitNameSortedUbam {
     String split_arg = if defined(n_reads) then "--SPLIT_TO_N_READS ~{X}" else "--SPLIT_TO_N_FILES ~{X}"
     String helper_arg = if (defined(read_cnt)) then "--TOTAL_READS_IN_INPUT ~{read_cnt}" else " "
 
-    String base = basename(uBAM, ".bam")
-    String local_bam = "/cromwell_root/~{base}.bam"
-
     command <<<
         set -eux
 
         if ~{fail}; then echo "one and only one of [n_reads, n_files] must be specified" && exit 1; fi
 
         # prep
-        time gcloud storage cp ~{uBAM} ~{local_bam}
         mkdir -p split_outputs
 
         # higher memory, also lower # of reads in memory given ~100 longer reads (1.5E4 bp vs 1.5E2 bp)
         gatk SplitSamByNumberOfReads \
             --java-options "-Xmx28G -Xms24G" \
             -use_jdk_deflater -use_jdk_inflater \
             --MAX_RECORDS_IN_RAM 5000 \
-            -I ~{local_bam} \
+            -I ~{uBAM} \
             -O split_outputs \
             ~{split_arg} \
             ~{helper_arg}
@@ -1695,8 +1690,8 @@ task SplitNameSortedUbam {
         mem_gb:             32,
         disk_gb:            disk_size,
         boot_disk_gb:       10,
-        preemptible_tries:  2,
-        max_retries:        1,
+        preemptible_tries:  1,
+        max_retries:        0,
         docker:             "us.gcr.io/broad-gatk/gatk:4.4.0.0"
     }
     RuntimeAttr runtime_attr = select_first([runtime_attr_override, default_attr])