Make the worker batch timeout configurable via LIDAR_BATCH_TIMEOUT

The pool of tile workers used to cancel every remaining tile after a
hardcoded 2-hour wall clock, silently truncating large batches (a
670-tile completion run lost its last 348 tiles that way). The timeout
now defaults to unlimited and can be capped per deployment with the
LIDAR_BATCH_TIMEOUT environment variable (seconds); the local worker
compose sets it to 6 hours.

💘 Generated with Crush

Assisted-by: Crush:glm-5.2

lidar_pipeline/pipeline.py

@@ -47,6 +47,22 @@ def resolve_workers(value):
         return 1
 
 
+def _batch_timeout_s():
+    """Wall-clock budget for one batch, from LIDAR_BATCH_TIMEOUT (seconds).
+
+    Unset, empty or 0 means unlimited (default). A negative or non-numeric
+    value is ignored with a warning.
+    """
+    raw = (os.environ.get("LIDAR_BATCH_TIMEOUT") or "").strip()
+    if not raw:
+        return 0.0
+    try:
+        return max(0.0, float(raw))
+    except ValueError:
+        logger.warning(f"LIDAR_BATCH_TIMEOUT={raw!r} is not a number — ignored (unlimited)")
+        return 0.0
+
+
 def _file_basename(path):
     """Extract base name from a LAZ/LAS file, removing all known extensions.
 
@@ -852,15 +868,17 @@ class LidarArchaeoPipeline:
                     for laz_file in files
                 }
                 done = 0
-                t_deadline = time.time() + 7200
+                budget_s = _batch_timeout_s()
+                deadline = time.time() + budget_s if budget_s > 0 else None
                 try:
-                    # timeout=: without it, as_completed blocks between two
-                    # completions and the 2 h deadline is never evaluated if
-                    # no worker returns (run stuck forever).
+                    # timeout= lets the deadline be checked between two
+                    # completions; without it a stuck worker would block
+                    # as_completed forever. No deadline (LIDAR_BATCH_TIMEOUT
+                    # unset or 0) → wait indefinitely.
+                    as_completed_kwargs = ({"timeout": max(1.0, deadline - time.time())}
+                                           if deadline is not None else {})
                     try:
-                        futures_iter = as_completed(future_to_file,
-                                                    timeout=max(1.0, t_deadline - time.time()))
-                        for future in futures_iter:
+                        for future in as_completed(future_to_file, **as_completed_kwargs):
                             laz_file = future_to_file[future]
                             done += 1
                             try:
@@ -877,7 +895,8 @@ class LidarArchaeoPipeline:
                                              "tile", "fail", detail=str(e))
                                 results[laz_file.name] = False
                     except FuturesTimeoutError:
-                        logger.error("Timeout exceeded (2 h) — cancelling remaining workers")
+                        logger.error(f"Batch exceeded LIDAR_BATCH_TIMEOUT "
+                                     f"({budget_s:g} s) — cancelling remaining workers")
                         for f in future_to_file:
                             f.cancel()
                 except KeyboardInterrupt: