script: debug wl gen (python json = )

scripts-py/expe_energumen/m100_generate_batsim_workload.py

@@ -2,6 +2,7 @@
 import argparse
 from concurrent.futures import ProcessPoolExecutor
 import datetime
+from math import isfinite
 import json
 import os
 import pandas as pd
@@ -48,12 +49,13 @@ def generate_node_file(job_profile_dir, job_min_dt, job_max_dt, job_node, node_i
 def generate_job_power_time_series(job_profile_dir, job_min_dt, job_max_dt, job_nodes, min_power):
     fname = f'{job_profile_dir}/dynpower.csv'
     static_watts = len(job_nodes) * min_power
+    none_nan_to_idle = lambda x: static_watts if x is None or not isfinite(x) else x
 
     # assume that already existing dirs have valid content
     if os.path.exists(job_profile_dir):
         df = pd.read_csv(fname)
         df['job_total_dynamic_power'] = df['job_total_dynamic_power'] + static_watts
-        return float(df['job_total_dynamic_power'].mean()), float(df['job_total_dynamic_power'].max())
+        return none_nan_to_idle(float(df['job_total_dynamic_power'].mean())), none_nan_to_idle(float(df['job_total_dynamic_power'].max()))
 
     global POWER_DF
     time_mask = (POWER_DF['timestamp'] >= job_min_dt) & (POWER_DF['timestamp'] <= job_max_dt)
@@ -61,8 +63,8 @@ def generate_job_power_time_series(job_profile_dir, job_min_dt, job_max_dt, job_
     job_power_df = POWER_DF[time_mask & space_mask]
 
     job_summed_power_df = job_power_df.groupby('timestamp')['value'].sum()
-    job_real_power_mean = float(job_summed_power_df.mean())
-    job_real_power_max = float(job_summed_power_df.max())
+    job_real_power_mean = none_nan_to_idle(float(job_summed_power_df.mean()))
+    job_real_power_max = none_nan_to_idle(float(job_summed_power_df.max()))
 
     # remove static part
     job_summed_power_df = job_summed_power_df - static_watts
@@ -228,8 +230,8 @@ def main():
 "replay_begin_seconds_from_2022-01-01": {begin_seconds_since_m100_epoch},
 "replay_end_seconds_from_2022-01-01": {end_seconds_since_m100_epoch},
 "nb_res": 980,
-"jobs": {json.dumps(jobs, indent=4)},
-"profiles": {json.dumps(profiles, indent=4, sort_keys=True)}
+"jobs": {json.dumps(jobs, indent=4, allow_nan=False)},
+"profiles": {json.dumps(profiles, indent=4, sort_keys=True, allow_nan=False)}
 }}''')
 
     match args.profile_type: