Initial bisect command

Author: hwikle-lanl

lib/pavilion/commands/__init__.py

@@ -16,6 +16,7 @@
     '_log_results': ('_log_results', 'LogResults'),
     '_run':       ('_run', '_RunCommand'),
     '_series':    ('_series', 'AutoSeries'),
+    'bisect':     ('bisect', 'BisectCommand'),
     'build':      ('build', 'BuildCommand'),
     'cancel':     ('cancel', 'CancelCommand'),
     'cat':        ('cat', 'CatCommand'),

lib/pavilion/commands/bisect.py

@@ -0,0 +1,120 @@
+from argparse import ArgumentParser
+
+from pavilion import output
+from pavilion.errors import TestSeriesError
+from pavilion.config import PavConfig
+from pavilion.series.series import TestSeries
+from pavilion.schedulers.config import parse_node_range
+from .base_classes import Command
+
+
+class BisectCommand(Command):
+    """Identify bad nodes by using a test to perform a binary search."""
+
+    def __init__(self):
+        super().__init__(
+            'bisect',
+            'Perform a bisection search on a given set of nodes using a particular test.',
+            short_help="Perform a bisection search."
+        )
+
+    def _setup_arguments(self, parser: ArgumentParser) -> None:
+        """Set up the parser arguments."""
+
+        parser.add_argument("test_name", type=str,
+                help="The name of the test to use to bisect the nodes.")
+        parser.add_argument("nodes", type=str, default="",
+                help="The set of nodes with which to start the search.")
+         parser.add_argument(
+            '-p', '--platform', action='store',
+            help='The platform to configure this test for. If not '
+            'specified, the current platform as denoted by the sys '
+            'plugin \'platform\' is used.')
+        parser.add_argument(
+            '-H', '--host', action='store',
+            help='The host to configure this test for. If not specified, the '
+                 'current host as denoted by the sys plugin \'sys_host\' is '
+                 'used. Host configurations are overlayed on operating system '
+                 'configurations.')
+        parser.add_argument(
+            '-n', '--name', action='store', default=''
+        )
+        parser.add_argument(
+            '-m', '--mode', action='append', dest='modes', default=[],
+            help='Mode configurations to overlay on the host configuration for '
+                 'each test. These are overlayed in the order given.')
+        parser.add_argument(
+            '-c', dest='overrides', action='append', default=[],
+            help='Overrides for specific configuration options. These are '
+                 'gathered used as a final set of overrides before the '
+                 'configs are resolved. They should take the form '
+                 '\'key=value\', where key is the dot separated key name, '
+                 'and value is a json object. Example: `-c schedule.nodes=23`')
+
+    def run(pav_cfg: PavConfig, args: Namespace) -> int:
+        """Run the bisection search."""
+
+        try:
+            nodes = self._parse_nodes(args.nodes)
+        except ValueError:
+            output.fprint(self.errfile, f"Error parsing node list {args.nodes}.")
+
+            return 1
+
+        # 1. Split the set of nodes in half.
+        num_nodes = len(nodes)
+        first_half = nodes[:num_nodes]
+        second_half = nodes[num_nodes:]
+
+        # 2. Run the test on each set of nodes
+        series_cfg = generate_series_config(
+            name="bisect",
+            modes=args.modes,
+            platform=args.platform,
+            host=args.host,
+            overrides=args.overrides,
+            ignore_errors=args.ignore_errors,
+        )
+
+        series_obj = TestSeries(pav_cfg, series_cfg=series_cfg, outfile=self.outfile)
+        testset_name = cmd_utils.get_testset_name(pav_cfg, [args.test_name], [])
+
+        series_obj.add_test_set_config(
+            testset_name,
+            [args.test_name],
+            modes=args.modes,
+        )
+
+        self.last_series = series_obj
+
+        try:
+            series_obj.run(
+                rebuild=args.rebuild,
+                local_builds_only=local_builds_only,
+                log_results=log_results)
+            self.last_tests = list(series_obj.tests.values())
+        except TestSeriesError as err:
+            self.last_tests = list(series_obj.tests.values())
+            output.fprint(self.errfile, err, color=output.RED)
+
+            return 2
+
+        # 3. Wait for tests to finish
+        series_obj.wait()
+
+        # 4. Choose test with failed nodes and repeat
+
+        return 0
+
+    @staticmethod
+    def _parse_nodes(nodes: str) -> List[str]:
+        """Parse a list of nodes into individual nodes."""
+
+        nodes = []
+
+        ranges = nodes.split(",")
+
+        for rng in ranges:
+            nodes.append(parse_node_range(rng))
+
+        return nodes