Allow users to specify percentiles for compare subcommand (opensearch…

…-project#503) Signed-off-by: Ian Hoang <[email protected]>
IanHoang · Apr 4, 2024 · bee133d · bee133d
1 parent 5f19515
commit bee133d
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 2 deletions.
diff --git a/osbenchmark/benchmark.py b/osbenchmark/benchmark.py
@@ -197,6 +197,11 @@ def add_workload_source(subparser):
         "--contender",
         required=True,
         help=f"TestExecution ID of the contender (see {PROGRAM_NAME} list test_executions).")
+    compare_parser.add_argument(
+        "--percentiles",
+        help=f"A comma-separated list of percentiles to report for latency and service time"
+             f"(default: {metrics.GlobalStatsCalculator.DEFAULT_LATENCY_PERCENTILES}).",
+        default=metrics.GlobalStatsCalculator.DEFAULT_LATENCY_PERCENTILES)
     compare_parser.add_argument(
         "--results-format",
         help="Define the output format for the command line results (default: markdown).",
@@ -820,6 +825,7 @@ def configure_results_publishing_params(args, cfg):
     cfg.add(config.Scope.applicationOverride, "results_publishing", "values", args.show_in_results)
     cfg.add(config.Scope.applicationOverride, "results_publishing", "output.path", args.results_file)
     cfg.add(config.Scope.applicationOverride, "results_publishing", "numbers.align", args.results_numbers_align)
+    cfg.add(config.Scope.applicationOverride, "results_publishing", "percentiles", args.percentiles)
 
 
 def print_test_execution_id(args):

diff --git a/osbenchmark/results_publisher.py b/osbenchmark/results_publisher.py
@@ -337,14 +337,15 @@ def _line(self, k, task, v, unit, converter=lambda x: x, force=False):
 
 class ComparisonResultsPublisher:
     def __init__(self, config):
+        self.logger = logging.getLogger(__name__)
         self.results_file = config.opts("results_publishing", "output.path")
         self.results_format = config.opts("results_publishing", "format")
         self.numbers_align = config.opts("results_publishing", "numbers.align",
                                          mandatory=False, default_value="right")
         self.cwd = config.opts("node", "benchmark.cwd")
         self.show_processing_time = convert.to_bool(config.opts("results_publishing", "output.processingtime",
                                                                 mandatory=False, default_value=False))
-        self.latency_percentiles = comma_separated_string_to_number_list(config.opts("workload", "latency.percentiles", mandatory=False))
+        self.percentiles = comma_separated_string_to_number_list(config.opts("results_publishing", "percentiles", mandatory=False))
         self.plain = False
 
     def publish(self, r1, r2):
@@ -442,7 +443,7 @@ def _publish_processing_time(self, baseline_stats, contender_stats, task):
 
     def _publish_percentiles(self, name, task, baseline_values, contender_values):
         lines = []
-        for percentile in metrics.percentiles_for_sample_size(sys.maxsize, percentiles_list=self.latency_percentiles):
+        for percentile in metrics.percentiles_for_sample_size(sys.maxsize, percentiles_list=self.percentiles):
             baseline_value = baseline_values.get(metrics.encode_float_key(percentile))
             contender_value = contender_values.get(metrics.encode_float_key(percentile))
             self._append_non_empty(lines, self._line("%sth percentile %s" % (percentile, name),