abstract-block-dumper 0.0.7__tar.gz → 0.0.8__tar.gz

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/CHANGELOG.md

@@ -9,6 +9,11 @@ upcoming release can be found in [changelog.d](changelog.d).
 
 <!-- towncrier release notes start -->
 
+## [0.0.8](https://github.com/bactensor/abstract-block-dumper/releases/tag/v0.0.8) - 2025-12-04
+
+No significant changes.
+
+
 ## [0.0.7](https://github.com/bactensor/abstract-block-dumper/releases/tag/v0.0.7) - 2025-12-01
 
 No significant changes.

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: abstract-block-dumper
-Version: 0.0.7
+Version: 0.0.8
 Project-URL: Source, https://github.com/bactensor/abstract-block-dumper
 Project-URL: Issue Tracker, https://github.com/bactensor/abstract-block-dumper/issues
 Author-email: Reef Technologies <opensource@reef.pl>
@@ -18,6 +18,8 @@ Requires-Dist: bittensor>=9.10.1
 Requires-Dist: celery>=5.3
 Requires-Dist: django<6.0,>=3.2
 Requires-Dist: structlog>=25.4.0
+Provides-Extra: prometheus
+Requires-Dist: prometheus-client>=0.17.0; extra == 'prometheus'
 Description-Content-Type: text/markdown
 
 # Abstract Block Dumper

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/pyproject.toml

@@ -27,6 +27,9 @@ dependencies = [
     "structlog>=25.4.0",
 ]
 
+[project.optional-dependencies]
+prometheus = ["prometheus-client>=0.17.0"]
+
 [project.urls]
 "Source" = "https://github.com/bactensor/abstract-block-dumper"
 "Issue Tracker" = "https://github.com/bactensor/abstract-block-dumper/issues"
@@ -160,6 +163,8 @@ module = [
     "cloudpickle",
     "cloudpickle.*",
     "nox",
+    "prometheus_client",
+    "prometheus_client.*",
     "pytest",
     "tests.*",
 ]

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/src/abstract_block_dumper/_internal/dal/django_dal.py

@@ -21,13 +21,19 @@ def get_ready_to_retry_attempts() -> QuerySet[abd_models.TaskAttempt]:
 
 
 def executed_block_numbers(executable_path: str, args_json: str, from_block: int, to_block: int) -> set[int]:
-    block_numbers = abd_models.TaskAttempt.objects.filter(
-        executable_path=executable_path,
-        args_json=args_json,
-        block_number__gte=from_block,
-        block_number__lt=to_block,
-        status=abd_models.TaskAttempt.Status.SUCCESS,
-    ).values_list("block_number", flat=True)
+    # Use iterator() to avoid Django's QuerySet caching which causes memory leaks
+    # during long-running backfill operations
+    block_numbers = (
+        abd_models.TaskAttempt.objects.filter(
+            executable_path=executable_path,
+            args_json=args_json,
+            block_number__gte=from_block,
+            block_number__lt=to_block,
+            status=abd_models.TaskAttempt.Status.SUCCESS,
+        )
+        .values_list("block_number", flat=True)
+        .iterator()
+    )
     return set(block_numbers)
 
 

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/src/abstract_block_dumper/_internal/dal/memory_registry.py

@@ -23,15 +23,14 @@ class RegistryItem:
         """Check if condition matches for given block and arguments."""
         try:
             return self.condition(block_number, **kwargs)
-        except Exception as e:
-            logger.error(
+        except Exception as exc:
+            logger.exception(
                 "Condition evaluation failed",
                 condition=self.function.__name__,
                 block_number=block_number,
-                exc_info=True,
             )
-            msg = f"Failed to evaluate condition: {e}"
-            raise ConditionEvaluationError(msg) from e
+            msg = "Failed to evaluate condition"
+            raise ConditionEvaluationError(msg) from exc
 
     def get_execution_args(self) -> list[dict[str, Any]]:
         """Get list of argument sets for execution."""

abstract_block_dumper-0.0.8/src/abstract_block_dumper/_internal/services/backfill_scheduler.py

@@ -0,0 +1,438 @@
+"""
+Backfill scheduler for historical block processing.
+
+This module provides a dedicated scheduler for backfilling historical blocks
+with rate limiting and automatic archive network switching.
+"""
+
+from __future__ import annotations
+
+import time
+from dataclasses import dataclass
+from typing import TYPE_CHECKING
+
+import structlog
+
+import abstract_block_dumper._internal.dal.django_dal as abd_dal
+import abstract_block_dumper._internal.services.utils as abd_utils
+from abstract_block_dumper._internal.services.block_processor import BlockProcessor, block_processor_factory
+from abstract_block_dumper._internal.services.metrics import (
+    BlockProcessingTimer,
+    increment_blocks_processed,
+    set_backfill_progress,
+    set_current_block,
+)
+from abstract_block_dumper._internal.services.utils import serialize_args
+
+if TYPE_CHECKING:
+    import bittensor as bt
+
+    from abstract_block_dumper._internal.dal.memory_registry import RegistryItem
+
+logger = structlog.get_logger(__name__)
+
+# Blocks older than this threshold from current head require archive network
+ARCHIVE_BLOCK_THRESHOLD = 300
+
+# Progress logging interval
+PROGRESS_LOG_INTERVAL = 100
+ARCHIVE_NETWORK = "archive"
+
+# Memory cleanup interval (every N blocks)
+MEMORY_CLEANUP_INTERVAL = 1000
+
+
+@dataclass
+class DryRunStats:
+    """Statistics for dry-run mode."""
+
+    total_blocks: int = 0
+    already_processed: int = 0
+    blocks_needing_tasks: int = 0
+    estimated_tasks: int = 0
+
+
+class BackfillScheduler:
+    """Scheduler for backfilling historical blocks with rate limiting."""
+
+    def __init__(
+        self,
+        block_processor: BlockProcessor,
+        network: str,
+        from_block: int,
+        to_block: int,
+        rate_limit: float = 1.0,
+        dry_run: bool = False,
+    ) -> None:
+        """
+        Initialize the backfill scheduler.
+
+        Args:
+            block_processor: The block processor to use for task execution.
+            network: The bittensor network name (e.g., 'finney').
+            from_block: Starting block number (inclusive).
+            to_block: Ending block number (inclusive).
+            rate_limit: Seconds to sleep between processing each block.
+            dry_run: If True, preview what would be processed without executing.
+
+        """
+        self.block_processor = block_processor
+        self.network = network
+        self.from_block = from_block
+        self.to_block = to_block
+        self.rate_limit = rate_limit
+        self.dry_run = dry_run
+        self.is_running = False
+        self._subtensor: bt.Subtensor | None = None
+        self._archive_subtensor: bt.Subtensor | None = None
+        self._current_head_cache: int | None = None
+
+    @property
+    def subtensor(self) -> bt.Subtensor:
+        """Get the regular subtensor connection, creating it if needed."""
+        if self._subtensor is None:
+            self._subtensor = abd_utils.get_bittensor_client(self.network)
+        return self._subtensor
+
+    @property
+    def archive_subtensor(self) -> bt.Subtensor:
+        """Get the archive subtensor connection, creating it if needed."""
+        if self._archive_subtensor is None:
+            self._archive_subtensor = abd_utils.get_bittensor_client("archive")
+        return self._archive_subtensor
+
+    def get_subtensor_for_block(self, block_number: int) -> bt.Subtensor:
+        """
+        Get the appropriate subtensor for the given block number.
+
+        Uses archive network for blocks older than ARCHIVE_BLOCK_THRESHOLD
+        from the current head.
+        """
+        if self._current_head_cache is None:
+            self._current_head_cache = self.subtensor.get_current_block()
+
+        blocks_behind = self._current_head_cache - block_number
+
+        if blocks_behind > ARCHIVE_BLOCK_THRESHOLD:
+            logger.debug(
+                "Using archive network for old block",
+                block_number=block_number,
+                blocks_behind=blocks_behind,
+            )
+            return self.archive_subtensor
+        return self.subtensor
+
+    def _get_network_type_for_block(self, block_number: int) -> str:
+        """Get the network type string for a block (for display purposes)."""
+        if self._current_head_cache is None:
+            self._current_head_cache = self.subtensor.get_current_block()
+
+        blocks_behind = self._current_head_cache - block_number
+        return ARCHIVE_NETWORK if blocks_behind > ARCHIVE_BLOCK_THRESHOLD else self.network
+
+    def start(self) -> DryRunStats | None:
+        """
+        Start processing blocks from from_block to to_block.
+
+        Returns:
+            DryRunStats if dry_run is True, None otherwise.
+
+        """
+        self.is_running = True
+
+        # Refresh current head for accurate archive network decisions
+        self._current_head_cache = self.subtensor.get_current_block()
+
+        total_blocks = self.to_block - self.from_block + 1
+        network_type = self._get_network_type_for_block(self.from_block)
+
+        logger.info(
+            "BackfillScheduler starting",
+            from_block=self.from_block,
+            to_block=self.to_block,
+            total_blocks=total_blocks,
+            rate_limit=self.rate_limit,
+            dry_run=self.dry_run,
+            network_type=network_type,
+            current_head=self._current_head_cache,
+        )
+
+        if self.dry_run:
+            return self._run_dry_run()
+
+        self._run_backfill()
+        return None
+
+    def _run_dry_run(self) -> DryRunStats:
+        """
+        Run in dry-run mode to preview what would be processed.
+
+        Optimized to fetch all executed blocks in one query per registry item,
+        instead of querying for each block individually.
+        """
+        stats = DryRunStats(total_blocks=self.to_block - self.from_block + 1)
+
+        registry_items = self.block_processor.registry.get_functions()
+
+        # Pre-fetch all executed blocks for each registry item + args combination
+        # This reduces N queries (one per block) to M queries (one per registry item + args)
+        executed_blocks_cache: dict[tuple[str, str], set[int]] = {}
+
+        logger.info(
+            "Dry run: pre-fetching executed blocks",
+            from_block=self.from_block,
+            to_block=self.to_block,
+            registry_items_count=len(registry_items),
+        )
+
+        for registry_item in registry_items:
+            for args in registry_item.get_execution_args():
+                args_json = serialize_args(args)
+                cache_key = (registry_item.executable_path, args_json)
+
+                # Fetch all executed blocks in the range with a single query
+                executed_blocks_cache[cache_key] = abd_dal.executed_block_numbers(
+                    registry_item.executable_path,
+                    args_json,
+                    self.from_block,
+                    self.to_block + 1,
+                )
+
+        logger.info(
+            "Dry run: analyzing blocks",
+            cache_entries=len(executed_blocks_cache),
+        )
+
+        # Track which blocks have at least one task
+        blocks_with_tasks: set[int] = set()
+
+        for registry_item in registry_items:
+            for args in registry_item.get_execution_args():
+                args_json = serialize_args(args)
+                cache_key = (registry_item.executable_path, args_json)
+                executed_blocks = executed_blocks_cache[cache_key]
+
+                for block_number in range(self.from_block, self.to_block + 1):
+                    if not self.is_running:
+                        break
+
+                    if block_number in executed_blocks:
+                        continue
+
+                    # Check if condition matches
+                    try:
+                        if registry_item.match_condition(block_number, **args):
+                            stats.estimated_tasks += 1
+                            blocks_with_tasks.add(block_number)
+                    except Exception as exc:
+                        logger.debug(
+                            "Error evaluating match condition during dry run",
+                            function_name=registry_item.function.__name__,
+                            block_number=block_number,
+                            args=args,
+                            error=str(exc),
+                        )
+
+        stats.blocks_needing_tasks = len(blocks_with_tasks)
+        stats.already_processed = stats.total_blocks - stats.blocks_needing_tasks
+
+        return stats
+
+    def _run_backfill(self) -> None:
+        """Run the actual backfill process."""
+        processed_count = 0
+        total_blocks = self.to_block - self.from_block + 1
+
+        # Set initial metrics
+        set_backfill_progress(self.from_block, self.to_block, self.from_block)
+
+        # Pre-fetch all executed blocks to avoid per-block DB queries
+        logger.info(
+            "Pre-fetching executed blocks",
+            from_block=self.from_block,
+            to_block=self.to_block,
+        )
+        executed_blocks_cache = self._prefetch_executed_blocks()
+        logger.info(
+            "Pre-fetch complete",
+            cache_entries=len(executed_blocks_cache),
+        )
+
+        try:
+            for block_number in range(self.from_block, self.to_block + 1):
+                if not self.is_running:
+                    logger.info("BackfillScheduler stopping early", processed_count=processed_count)
+                    break
+
+                try:
+                    with BlockProcessingTimer(mode="backfill"):
+                        self._process_block(block_number, executed_blocks_cache)
+
+                    processed_count += 1
+
+                    # Update metrics
+                    set_current_block("backfill", block_number)
+                    set_backfill_progress(self.from_block, self.to_block, block_number)
+                    increment_blocks_processed("backfill")
+
+                    # Log progress periodically
+                    if processed_count % PROGRESS_LOG_INTERVAL == 0:
+                        progress_pct = (processed_count / total_blocks) * 100
+                        logger.info(
+                            "Backfill progress",
+                            processed=processed_count,
+                            total=total_blocks,
+                            progress_percent=f"{progress_pct:.1f}%",
+                            current_block=block_number,
+                        )
+
+                    # Rate limiting between block submissions
+                    if block_number < self.to_block and self.rate_limit > 0:
+                        time.sleep(self.rate_limit)
+
+                except KeyboardInterrupt:
+                    raise
+                except Exception:
+                    logger.exception(
+                        "Error processing block during backfill",
+                        block_number=block_number,
+                    )
+                    # Continue with next block
+                    time.sleep(self.rate_limit)
+
+        except KeyboardInterrupt:
+            logger.info(
+                "BackfillScheduler interrupted",
+                processed_count=processed_count,
+                last_block=self.from_block + processed_count - 1 if processed_count > 0 else self.from_block,
+            )
+            self.stop()
+
+        logger.info(
+            "BackfillScheduler completed",
+            processed_count=processed_count,
+            total_blocks=total_blocks,
+        )
+
+    def _prefetch_executed_blocks(self) -> dict[tuple[str, str], set[int]]:
+        """Pre-fetch all executed blocks for all registry items in the range."""
+        cache: dict[tuple[str, str], set[int]] = {}
+
+        for registry_item in self.block_processor.registry.get_functions():
+            for args in registry_item.get_execution_args():
+                args_json = serialize_args(args)
+                cache_key = (registry_item.executable_path, args_json)
+
+                cache[cache_key] = abd_dal.executed_block_numbers(
+                    registry_item.executable_path,
+                    args_json,
+                    self.from_block,
+                    self.to_block + 1,
+                )
+
+        return cache
+
+    def _process_block(
+        self,
+        block_number: int,
+        executed_blocks_cache: dict[tuple[str, str], set[int]],
+    ) -> None:
+        """Process a single block during backfill."""
+        for registry_item in self.block_processor.registry.get_functions():
+            try:
+                self._process_registry_item_for_backfill(
+                    registry_item,
+                    block_number,
+                    executed_blocks_cache,
+                )
+            except Exception:
+                logger.exception(
+                    "Error processing registry item during backfill",
+                    function_name=registry_item.function.__name__,
+                    block_number=block_number,
+                )
+
+    def _requires_archive_network(self, block_number: int) -> bool:
+        """Check if a block requires archive network based on age."""
+        if self._current_head_cache is None:
+            return False
+        blocks_behind = self._current_head_cache - block_number
+        return blocks_behind > ARCHIVE_BLOCK_THRESHOLD
+
+    def _process_registry_item_for_backfill(
+        self,
+        registry_item: RegistryItem,
+        block_number: int,
+        executed_blocks_cache: dict[tuple[str, str], set[int]],
+    ) -> None:
+        """Process a registry item for backfill - only submits if not already executed."""
+        for args in registry_item.get_execution_args():
+            args_json = serialize_args(args)
+            cache_key = (registry_item.executable_path, args_json)
+
+            # Check if already executed using pre-fetched cache
+            executed_blocks = executed_blocks_cache.get(cache_key, set())
+
+            if block_number in executed_blocks:
+                continue
+
+            # Check condition and execute
+            try:
+                if registry_item.match_condition(block_number, **args):
+                    use_archive = self._requires_archive_network(block_number)
+                    logger.debug(
+                        "Backfilling block",
+                        function_name=registry_item.function.__name__,
+                        block_number=block_number,
+                        args=args,
+                        use_archive=use_archive,
+                    )
+                    self.block_processor.executor.execute(
+                        registry_item,
+                        block_number,
+                        args,
+                        use_archive=use_archive,
+                    )
+            except Exception:
+                logger.exception(
+                    "Error during backfill task execution",
+                    function_name=registry_item.function.__name__,
+                    block_number=block_number,
+                    args=args,
+                )
+
+    def stop(self) -> None:
+        """Stop the backfill scheduler."""
+        self.is_running = False
+        logger.info("BackfillScheduler stopped")
+
+
+def backfill_scheduler_factory(
+    from_block: int,
+    to_block: int,
+    network: str = "finney",
+    rate_limit: float = 1.0,
+    dry_run: bool = False,
+) -> BackfillScheduler:
+    """
+    Factory for BackfillScheduler.
+
+    Args:
+        from_block: Starting block number (inclusive).
+        to_block: Ending block number (inclusive).
+        network: Bittensor network name. Defaults to "finney".
+        rate_limit: Seconds to sleep between blocks. Defaults to 1.0.
+        dry_run: If True, preview without executing. Defaults to False.
+
+    Returns:
+        Configured BackfillScheduler instance.
+
+    """
+    return BackfillScheduler(
+        block_processor=block_processor_factory(),
+        network=network,
+        from_block=from_block,
+        to_block=to_block,
+        rate_limit=rate_limit,
+        dry_run=dry_run,
+    )

{abstract_block_dumper-0.0.7 → abstract_block_dumper-0.0.8}/src/abstract_block_dumper/_internal/services/block_processor.py

@@ -1,3 +1,5 @@
+import time
+
 import structlog
 from django.db import transaction
 
@@ -5,7 +7,6 @@ import abstract_block_dumper._internal.dal.django_dal as abd_dal
 from abstract_block_dumper._internal.dal.memory_registry import BaseRegistry, RegistryItem, task_registry
 from abstract_block_dumper._internal.exceptions import ConditionEvaluationError
 from abstract_block_dumper._internal.services.executor import CeleryExecutor
-from abstract_block_dumper._internal.services.utils import serialize_args
 from abstract_block_dumper.models import TaskAttempt
 
 logger = structlog.get_logger(__name__)
@@ -18,16 +19,15 @@ class BlockProcessor:
         self._cleanup_phantom_tasks()
 
     def process_block(self, block_number: int) -> None:
+        """Process a single block - executes registered tasks for this block only."""
         for registry_item in self.registry.get_functions():
             try:
-                self.process_backfill(registry_item, block_number)
                 self.process_registry_item(registry_item, block_number)
             except Exception:
-                logger.error(
+                logger.exception(
                     "Error processing registry item",
                     function_name=registry_item.function.__name__,
                     block_number=block_number,
-                    exc_info=True,
                 )
 
     def process_registry_item(self, registry_item: RegistryItem, block_number: int) -> None:
@@ -43,64 +43,28 @@ class BlockProcessor:
                 )
                 # Continue with other tasks
             except Exception:
-                logger.error("Unexpected error processing task", exc_info=True)
-
-    def process_backfill(self, registry_item: RegistryItem, current_block: int) -> None:
-        if not registry_item.backfilling_lookback:
-            return None
-
-        start_block = max(0, current_block - registry_item.backfilling_lookback)
-
-        logger.info(
-            "Processing backfill",
-            function_name=registry_item.function.__name__,
-            start_block=start_block,
-            current_block=current_block,
-            lookback=registry_item.backfilling_lookback,
-        )
-
-        execution_args_list = registry_item.get_execution_args()
-
-        for args in execution_args_list:
-            args_json = serialize_args(args)
-
-            executed_blocks = abd_dal.executed_block_numbers(
-                registry_item.executable_path,
-                args_json,
-                start_block,
-                current_block,
-            )
-
-            for block_number in range(start_block, current_block):
-                if block_number in executed_blocks:
-                    continue
+                logger.exception("Unexpected error processing task")
 
-                try:
-                    if registry_item.match_condition(block_number, **args):
-                        logger.debug(
-                            "Backfilling block",
-                            function_name=registry_item.function.__name__,
-                            block_number=block_number,
-                            args=args,
-                        )
-                        self.executor.execute(registry_item, block_number, args)
-                except Exception:
-                    logger.error(
-                        "Error during backfill",
-                        function_name=registry_item.function.__name__,
-                        block_number=block_number,
-                        args=args,
-                        exc_info=True,
-                    )
-
-    def recover_failed_retries(self) -> None:
+    def recover_failed_retries(self, poll_interval: int, batch_size: int | None = None) -> None:
         """
         Recover failed tasks that are ready to be retried.
 
         This handles tasks that may have been lost due to scheduler restarts.
+
+        Args:
+            poll_interval: Seconds to sleep between processing each retry.
+            batch_size: Maximum number of retries to process. If None, process all.
+
         """
         retry_count = 0
-        for retry_attempt in abd_dal.get_ready_to_retry_attempts():
+        retry_attempts = abd_dal.get_ready_to_retry_attempts()
+
+        # Apply batch size limit if specified
+        if batch_size is not None:
+            retry_attempts = retry_attempts[:batch_size]
+
+        for retry_attempt in retry_attempts:
+            time.sleep(poll_interval)
             try:
                 # Find the registry item to get celery_kwargs
                 registry_item = self.registry.get_by_executable_path(retry_attempt.executable_path)
@@ -148,10 +112,9 @@ class BlockProcessor:
                     attempt_count=task_attempt.attempt_count,
                 )
             except Exception:
-                logger.error(
+                logger.exception(
                     "Failed to recover retry",
                     task_id=retry_attempt.id,
-                    exc_info=True,
                 )
                 # Reload task to see current state after potential execution failure
                 try: