Make tofu runs support multiple years on single machine (#143)

Author: KristijanFaust-OET

.pre-commit-config.yaml

@@ -25,3 +25,4 @@ repos:
         types_or: [javascript, jsx, ts, tsx]
         additional_dependencies:
           - prettier@3.1.0
+        language_version: 18.18.0

infrastructure/README.md

@@ -24,7 +24,7 @@ Place your benchmark YAML files in the `benchmarks` directory. Example:
 
 ```yaml
 machine-type: c4-standard-2
-year: 2024
+years: [2023, 2024]
 benchmarks:
     Short description: Sector-coupled PyPSA-Eur infrastructure run for Italy considering 2050 as single planning horizon (LP, lot of variables, strongly intermeshed constraints)
     Model name: PyPSA-Eur

infrastructure/main.tf

@@ -117,7 +117,7 @@ resource "google_compute_instance" "benchmark_instances" {
   metadata = {
     ssh-keys = var.ssh_user != "" && var.ssh_key_path != "" ? "${var.ssh_user}:${file(var.ssh_key_path)}" : null
     benchmark_file = each.value.filename
-    benchmark_year = lookup(each.value.content, "year", "2024")
+    benchmark_years = jsonencode(lookup(each.value.content, "years", ["2024"]))
     benchmark_content = file("${path.module}/benchmarks/${each.value.filename}")
     enable_gcs_upload = tostring(var.enable_gcs_upload)
     gcs_bucket_name = var.gcs_bucket_name

infrastructure/startup-script.sh

@@ -24,29 +24,13 @@ echo "Setting up conda environment..."
 echo "source ~/miniconda3/bin/activate" >> ~/.bashrc
 ~/miniconda3/bin/conda init bash
 
-# Get benchmark year from instance metadata
-BENCHMARK_YEAR=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/benchmark_year")
-echo "Using benchmark year: ${BENCHMARK_YEAR}"
+# Get benchmark years from instance metadata
+BENCHMARK_YEARS_JSON=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/benchmark_years")
+echo "Retrieved benchmark years: ${BENCHMARK_YEARS_JSON}"
 
-# Create conda environment with the appropriate year
-cd /solver-benchmark/
-~/miniconda3/bin/conda env create -f runner/envs/benchmark-${BENCHMARK_YEAR}-fixed.yaml
-
-# Activate the conda environment
-echo "Activating conda environment benchmark-${BENCHMARK_YEAR}..."
-source ~/miniconda3/bin/activate
-conda activate benchmark-${BENCHMARK_YEAR}
-
-# Add auto-activation to .bashrc
-echo -e "\n# Automatically activate benchmark environment" >> ~/.bashrc
-echo "conda activate benchmark-${BENCHMARK_YEAR}" >> ~/.bashrc
-
-# Verify environment is active
-echo "Current conda environment:"
-conda info --envs | grep "*"
-
-echo "Setup completed at $(date)"
-echo "Conda environment benchmark-${BENCHMARK_YEAR} is now active and will be activated on login"
+# Parse the JSON array into a space-separated string for benchmark_all.sh
+BENCHMARK_YEARS_STR=$(echo "${BENCHMARK_YEARS_JSON}" | jq -r 'join(" ")')
+echo "Parsed benchmark years: ${BENCHMARK_YEARS_STR}"
 
 # Get benchmark filename from instance metadata
 BENCHMARK_FILE=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/benchmark_file")
@@ -58,28 +42,53 @@ BENCHMARK_CONTENT=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.in
 # Write the benchmark file - preserve the exact content
 echo "${BENCHMARK_CONTENT}" > /solver-benchmark/benchmarks/${BENCHMARK_FILE}
 
-# Run the benchmarks
-echo "Starting benchmarks..."
-python /solver-benchmark/runner/run_benchmarks.py /solver-benchmark/benchmarks/${BENCHMARK_FILE} ${BENCHMARK_YEAR}
+# Make benchmark_all.sh executable
+cd /solver-benchmark/
+chmod +x ./runner/benchmark_all.sh
+
+# Run the benchmark_all.sh script with our years
+echo "Starting benchmarks for years: ${BENCHMARK_YEARS_STR}"
+source ~/miniconda3/bin/activate
+./runner/benchmark_all.sh -y "${BENCHMARK_YEARS_STR}" ./benchmarks/${BENCHMARK_FILE}
+BENCHMARK_EXIT_CODE=$?
+
+if [ $BENCHMARK_EXIT_CODE -ne 0 ]; then
+    echo "ERROR: Benchmark failed with exit code $BENCHMARK_EXIT_CODE at $(date)"
+    exit $BENCHMARK_EXIT_CODE
+fi
+
+echo "All benchmarks completed at $(date)"
+
+# Create timestamp for the results
+TIMESTAMP=$(date +%Y%m%d_%H%M%S)
+echo "Using timestamp: ${TIMESTAMP}"
+
+# Create a copy of results with timestamp
+CLEAN_FILENAME=$(basename "${BENCHMARK_FILE}" .yaml)
+RESULTS_COPY="/tmp/${CLEAN_FILENAME}_${TIMESTAMP}.csv"
+echo "Creating copy of results as: ${RESULTS_COPY}"
+
+cp /solver-benchmark/results/benchmark_results.csv "${RESULTS_COPY}"
+COPY_EXIT_CODE=$?
+
+if [ $COPY_EXIT_CODE -ne 0 ]; then
+    echo "ERROR: Failed to copy benchmark results at $(date). Exit code: $COPY_EXIT_CODE"
+    echo "Check if file exists: /solver-benchmark/results/benchmark_results.csv"
+    ls -la /solver-benchmark/results/
+    exit $COPY_EXIT_CODE
+fi
+
+echo "Benchmark results successfully copied at $(date)"
 
 # ----- GCS UPLOAD CONFIGURATION -----
+# Only proceed if the benchmark and copy operations were successful
 # Check if GCS upload is enabled
 ENABLE_GCS_UPLOAD=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/enable_gcs_upload")
 if [ "${ENABLE_GCS_UPLOAD}" == "true" ]; then
     # Get the GCS bucket name
     GCS_BUCKET_NAME=$(curl -H "Metadata-Flavor: Google" "http://metadata.google.internal/computeMetadata/v1/instance/attributes/gcs_bucket_name")
     echo "Using GCS bucket: ${GCS_BUCKET_NAME}"
 
-    # Create timestamp for the file
-    TIMESTAMP=$(date +%Y%m%d_%H%M%S)
-    echo "Using timestamp: ${TIMESTAMP}"
-
-    # Create the properly named copy of results
-    CLEAN_FILENAME=$(basename "${BENCHMARK_FILE}" .yaml)
-    RESULTS_COPY="/tmp/${CLEAN_FILENAME}_${TIMESTAMP}.csv"
-    echo "Creating copy of results as: ${RESULTS_COPY}"
-    cp /solver-benchmark/results/benchmark_results.csv "${RESULTS_COPY}"
-
     # Ensure gsutil is available (should be on GCP instances by default)
     if ! command -v gsutil &> /dev/null; then
         echo "Installing Google Cloud SDK..."