[FR] Enable best effort parital analysis and verbose mode for trace printing (#139853)

Based on user feedback, we want to enable two things for FR analysis script:
1. Print out more information when verbose is specified.
2. Perform best effort based analysis when not all ranks have FR trace dumped.

Differential Revision: [D65516081](https://our.internmc.facebook.com/intern/diff/D65516081/)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/139853
Approved by: https://github.com/c-p-i-o

test/distributed/flight_recorder/test_fr_analysis.py

@@ -37,6 +37,7 @@ def create_one_event(
         "output_dtypes": output_dtypes,
         "collective_seq_id": str(collective_seq_id),
         "p2p_seq_id": str(p2p_seq_id),
+        "time_created_ns": 0,
     }
 
 

tools/flight_recorder/components/builder.py

@@ -6,6 +6,7 @@
 
 import argparse
 import ast
+import os
 import sys
 from typing import Any, Dict, List, Set, Tuple  # type: ignore[attr-defined]
 
@@ -186,6 +187,9 @@ def build_collectives(
     # instead, just record the remaining ops as NCCLCalls
     mismatch = {_groups[g].id: 0 for g in _groups}
     MISMATCH_TAIL = 10
+
+    # For best effort partial analysis.
+    dumps_ranks = {int(key) for key in all_entries.keys()}
     """
     - it doesn't matter what order I put collectives/ncclops into their table. we can later on re-sort it by start time
     - there could be multiple options for the "first" collective to pair up (rank 0,1 might do a bcast while rank 2,3 do a bcast)
@@ -238,7 +242,7 @@ def build_collectives(
                     else []
                 )
                 all_coalesced_entries[curr] = grp
-                for index, entry in grp:
+                for _, entry in grp:
                     op = Op(entry, _memberships, pg_name)
                     peer = None
                     if op.type == "send":
@@ -314,7 +318,9 @@ def build_collectives(
                         break
 
             # case one: not every rank join the collective or in the flight recorder.
-            if (candidate_ranks | found_ranks) != expected_ranks:
+            if (candidate_ranks | found_ranks) != expected_ranks and expected_ranks - (
+                candidate_ranks | found_ranks
+            ) <= dumps_ranks:
                 mismatch[pg_name] += 1
                 logger.info(
                     "Not all ranks joining collective %s at entry %s",
@@ -334,7 +340,7 @@ def build_collectives(
                 candidate_idx.update(found_idx)
                 found_idx.clear()
                 found_ranks.clear()
-            elif len(candidate_ranks) == 1:
+            elif len(candidate_ranks) == 1 and dumps_ranks == expected_ranks:
                 # case two: alltoall or alltoall_base case.
                 if has_undecided_case:
                     alltoall_cases = [entries[0]] + [
@@ -398,6 +404,19 @@ def build_collectives(
                 candidate_idx.update(found_idx)
                 found_idx.clear()
                 found_ranks.clear()
+            # partial analysis case when we cannot decide what's wrong with this collective entry.
+            else:
+                candidate_ranks.update(found_ranks)
+                candidate_idx.update(found_idx)
+                found_idx.clear()
+                found_ranks.clear()
+                mismatch[pg_name] += 1
+                logger.info(
+                    "We cannot decide what's wrong with this collective entry "
+                    "because we missed FR dumps from ranks (%s) so we don't have enough "
+                    "information. If you want to debug further use -j to dump all raw trace",
+                    str(expected_ranks - dumps_ranks),
+                )
 
             # at this point there are 3 possibilities
             # 1. we found a match on all the ranks that are members of the group
@@ -450,6 +469,8 @@ def build_collectives(
 def build_db(
     details: Dict[str, Dict[str, Any]], args: argparse.Namespace, version: str
 ) -> Database:
+    if args.verbose:
+        os.environ["FR_TRACE_VERBOSE_OUTPUT"] = "1"
     # temporary state used for building database
     entries = {}
     pg_config = {}
@@ -470,12 +491,13 @@ def build_db(
     )
     logger.debug("built groups, memberships")
 
+    if not args.allow_incomplete_ranks:
+        check_no_missing_dump_files(entries, memberships)
+
     if args.just_print_entries:
         just_print_entries(entries, _groups, _memberships, _pg_guids, args)
         sys.exit(0)
 
-    check_no_missing_dump_files(entries, memberships)
-
     tracebacks, collectives, nccl_calls = build_collectives(
         entries, _groups, _memberships, _pg_guids, version
     )

tools/flight_recorder/components/config_manager.py

@@ -35,6 +35,15 @@ class JobConfig:
             type=int,
             help="List of ranks we want to show traces for.",
         )
+        self.parser.add_argument(
+            "--allow-incomplete-ranks",
+            action="store_true",
+            help=(
+                "FR trace require all ranks to have dumps for analysis. "
+                "This flag allows best-effort partial analysis of results "
+                "and printing of collected data."
+            ),
+        )
         self.parser.add_argument(
             "--pg-filters",
             default=None,

tools/flight_recorder/components/loader.py

@@ -4,6 +4,7 @@
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import argparse
 import gc
 import os
 import pickle
@@ -11,7 +12,7 @@ import re
 import time
 import typing
 from collections import defaultdict
-from typing import Any, Dict, List, Optional, Set, Tuple, Union
+from typing import Any, Dict, List, Set, Tuple, Union
 
 from tools.flight_recorder.components.fr_logger import FlightRecorderLogger
 
@@ -66,16 +67,15 @@ def _determine_prefix(files: List[str]) -> str:
         )
 
 
-def read_dir(
-    prefix: Optional[str], folder: str
-) -> Tuple[Dict[str, Dict[str, Any]], str]:
+def read_dir(args: argparse.Namespace) -> Tuple[Dict[str, Dict[str, Any]], str]:
     gc.disable()
+    prefix = args.prefix
     details = {}
     t0 = time.time()
     version = ""
     filecount = 0
-    assert os.path.isdir(folder), f"folder {folder} does not exist"
-    for root, _, files in os.walk(folder):
+    assert os.path.isdir(args.folder), f"folder {args.folder} does not exist"
+    for root, _, files in os.walk(args.folder):
         if prefix is None:
             prefix = _determine_prefix(files)
         for f in files:
@@ -86,6 +86,6 @@ def read_dir(
             if not version:
                 version = str(details[f]["version"])
     tb = time.time()
-    assert len(details) > 0, f"no files loaded from {folder} with prefix {prefix}"
+    assert len(details) > 0, f"no files loaded from {args.folder} with prefix {prefix}"
     logger.debug("loaded %s files in %ss", filecount, tb - t0)
     return details, version

tools/flight_recorder/components/types.py

@@ -5,6 +5,7 @@
 # LICENSE file in the root directory of this source tree.
 
 import math
+import os
 from enum import auto, Enum
 from typing import (  # type: ignore[attr-defined]
     _eval_type,
@@ -199,7 +200,7 @@ class Op:
         type = parts[0]
         meta = parts[1] if len(parts) == 2 else None
         self.state = event["state"]
-        self.pg_name, _ = event["process_group"]
+        self.pg_name, self.pg_desc = event["process_group"]
         assert type in COLLECTIVES | P2P | {
             "coalesced"
         }, f"{type} is not a supported operation"
@@ -212,7 +213,6 @@ class Op:
             self._dst, self._src = int(d), int(s)
         else:
             self._src, self._dst = -1, -1
-        _, pg_desc = event["process_group"]
         self._init_global_src_dst(memberships[pg_name])
         self.pg_size = len(memberships[pg_name])
         if type in P2P | COLLECTIVES:
@@ -224,6 +224,8 @@ class Op:
         self.p2p_seq_id = event["p2p_seq_id"]
         self.input_dtypes = event["input_dtypes"]
         self.output_dtypes = event["output_dtypes"]
+        self.time_created_ns = event["time_created_ns"]
+        self.is_verbose = os.getenv("FR_TRACE_VERBOSE_OUTPUT", "0") == "1"
 
     def _init_global_src_dst(self, pg_ranks: Set[Any]) -> None:
         pg_ranks = sorted(pg_ranks)
@@ -241,9 +243,31 @@ class Op:
         return self._dst
 
     def __repr__(self) -> str:
+        p2p_info = ""
         if self.type in P2P:
-            return f"{self.type}(s={self._src_g} d={self._dst_g}, sz={self.input_sizes}, state={self.state})"
-        return f"{self.type}(input_sizes={self.input_sizes}, state={self.state})"
+            p2p_info = f"s={self._src_g} d={self._dst_g}"
+        if self.is_verbose:
+            verbose_info = (
+                f"timestamp_created={self.time_created_ns}",
+                p2p_info,
+                f"input_sizes={self.input_sizes}",
+                f"output_sizes={self.output_sizes}",
+                f"input_dtypes={self.input_dtypes}",
+                f"output_dtypes={self.output_dtypes}",
+                "collective_seq_id | p2p_seq_id="
+                f"{self.p2p_seq_id if self.type in P2P else self.collective_seq_id}",
+                f"pg_name={self.pg_name}",
+                f"pg_description={self.pg_desc}",
+                f"pg_size={self.pg_size}",
+                f"state={self.state}",
+            )
+            return f"{self.type}(%s)" % ", ".join(s for s in verbose_info if s)
+        return (
+            f"{self.type}(%sinput_sizes={self.input_sizes}, state={self.state})"
+            % f"{p2p_info}, "
+            if p2p_info
+            else ""
+        )
 
     def match(self, other: "Op") -> MatchState:
         # TODO: I think this can validly not match,

tools/flight_recorder/fr_trace.py

@@ -41,7 +41,7 @@ def main(args: Optional[Sequence[str]] = None) -> None:
     config = JobConfig()
     args = config.parse_args(args)
     assert args.trace_dir, "Trace directory trace_dir is required"
-    details, version = read_dir(args.prefix, args.trace_dir)
+    details, version = read_dir(args)
     db = build_db(details, args, version)
     if args.output:
         with open(args.output, "wb") as f: