apdbCassandraReplica.py

Go to the documentation of this file.

# This file is part of dax_apdb.
#
# Developed for the LSST Data Management System.
# This product includes software developed by the LSST Project
# (http://www.lsst.org).
# See the COPYRIGHT file at the top-level directory of this distribution
# for details of code ownership.
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
from __future__ import annotations
 
__all__ = ["ApdbCassandraReplica"]
 
import logging
from collections.abc import Iterable, Mapping
from typing import TYPE_CHECKING, cast
 
import astropy.time
import felis.datamodel
 
from ..apdbReplica import ApdbReplica, ApdbTableData, ReplicaChunk
from ..apdbSchema import ApdbTables
from ..monitor import MonAgent
from ..schema_model import ExtraDataTypes
from ..timer import Timer
from ..versionTuple import VersionTuple
from .apdbCassandraSchema import ExtraTables
from .cassandra_utils import (
    ApdbCassandraTableData,
    execute_concurrent,
    select_concurrent,
)
 
if TYPE_CHECKING:
    from .apdbCassandra import ApdbCassandra
 
_LOG = logging.getLogger(__name__)
 
_MON = MonAgent(__name__)
 
VERSION = VersionTuple(1, 1, 0)
"""Version for the code controlling replication tables. This needs to be
updated following compatibility rules when schema produced by this code
changes.
"""
 
 
class ApdbCassandraReplica(ApdbReplica):
    """Implementation of `ApdbReplica` for Cassandra backend.
 
    Parameters
    ----------
    apdb : `ApdbCassandra`
        Instance of ApbdCassandra for database.
    """
 
    def __init__(self, apdb: ApdbCassandra):
        # Note that ApdbCassandra instance must stay alive while this object
        # exists, so we keep reference to it.
        self._apdb = apdb
 
    def _timer(self, name: str, *, tags: Mapping[str, str | int] | None = None) -> Timer:
        """Create `Timer` instance given its name."""
        return Timer(name, _MON, tags=tags)
 
    def schemaVersion(self) -> VersionTuple:
        # Docstring inherited from base class.
        context = self._apdb._context
        return context.db_versions.schema_version
 
    @classmethod
    def apdbReplicaImplementationVersion(cls) -> VersionTuple:
        # Docstring inherited from base class.
        return VERSION
 
    @classmethod
    def hasChunkSubPartitions(cls, version: VersionTuple) -> bool:
        """Return True if replica chunk tables have sub-partitions."""
        return version >= VersionTuple(1, 1, 0)
 
    def getReplicaChunks(self) -> list[ReplicaChunk] | None:
        # docstring is inherited from a base class
        context = self._apdb._context
        config = context.config
 
        if not context.schema.replication_enabled:
            return None
 
        # everything goes into a single partition
        partition = 0
 
        table_name = context.schema.tableName(ExtraTables.ApdbReplicaChunks)
        # We want to avoid timezone mess so return timestamps as milliseconds.
        query = (
            "SELECT toUnixTimestamp(last_update_time), apdb_replica_chunk, unique_id "
            f'FROM "{config.keyspace}"."{table_name}" WHERE partition = %s'
        )
 
        with self._timer("chunks_select_time") as timer:
            result = context.session.execute(
                query,
                (partition,),
                timeout=config.connection_config.read_timeout,
                execution_profile="read_tuples",
            )
            # order by last_update_time
            rows = sorted(result)
            timer.add_values(row_count=len(rows))
        return [
            ReplicaChunk(
                id=row[1],
                last_update_time=astropy.time.Time(row[0] / 1000, format="unix_tai"),
                unique_id=row[2],
            )
            for row in rows
        ]
 
    def deleteReplicaChunks(self, chunks: Iterable[int]) -> None:
        # docstring is inherited from a base class
        context = self._apdb._context
        config = context.config
 
        if not context.schema.replication_enabled:
            raise ValueError("APDB is not configured for replication")
 
        # everything goes into a single partition
        partition = 0
 
        # Iterable can be single pass, make everything that we need from it
        # in a single loop.
        repl_table_params = []
        chunk_table_params: list[tuple] = []
        for chunk in chunks:
            repl_table_params.append((partition, chunk))
            if context.schema.has_chunk_sub_partitions:
                for subchunk in range(config.replica_sub_chunk_count):
                    chunk_table_params.append((chunk, subchunk))
            else:
                chunk_table_params.append((chunk,))
        # Anything to do att all?
        if not repl_table_params:
            return
 
        table_name = context.schema.tableName(ExtraTables.ApdbReplicaChunks)
        query = (
            f'DELETE FROM "{config.keyspace}"."{table_name}" WHERE partition = ? AND apdb_replica_chunk = ?'
        )
        statement = context.preparer.prepare(query)
 
        queries = [(statement, param) for param in repl_table_params]
        with self._timer("chunks_delete_time") as timer:
            execute_concurrent(context.session, queries)
            timer.add_values(row_count=len(queries))
 
        # Also remove those chunk_ids from Dia*Chunks tables.
        tables = list(ExtraTables.replica_chunk_tables(context.schema.has_chunk_sub_partitions).values())
        for table in tables:
            table_name = context.schema.tableName(table)
            query = f'DELETE FROM "{config.keyspace}"."{table_name}" WHERE apdb_replica_chunk = ?'
            if context.schema.has_chunk_sub_partitions:
                query += " AND apdb_replica_subchunk = ?"
            statement = context.preparer.prepare(query)
 
            queries = [(statement, param) for param in chunk_table_params]
            with self._timer("table_chunk_detele_time", tags={"table": table_name}) as timer:
                execute_concurrent(context.session, queries)
                timer.add_values(row_count=len(queries))
 
    def getTableDataChunks(self, table: ApdbTables, chunks: Iterable[int]) -> ApdbTableData:
        # docstring is inherited from a base class
        context = self._apdb._context
        config = context.config
 
        if not context.schema.replication_enabled:
            raise ValueError("APDB is not configured for replication")
        if table not in ExtraTables.replica_chunk_tables(False):
            raise ValueError(f"Table {table} does not support replica chunks.")
 
        # We need to iterate few times.
        chunks = list(chunks)
 
        # If schema was migrated then a chunk can appear in either old or new
        # chunk table (e.g. DiaObjectChunks or DiaObjectChunks2). Chunk table
        # has a column which will be set to true for new table.
        has_chunk_sub_partitions: dict[int, bool] = {}
        if context.schema.has_chunk_sub_partitions:
            table_name = context.schema.tableName(ExtraTables.ApdbReplicaChunks)
            chunks_str = ",".join(str(chunk_id) for chunk_id in chunks)
            query = (
                f'SELECT apdb_replica_chunk, has_subchunks FROM "{config.keyspace}"."{table_name}" '
                f"WHERE partition = %s and apdb_replica_chunk IN ({chunks_str})"
            )
            partition = 0
            result = context.session.execute(
                query,
                (partition,),
                timeout=config.connection_config.read_timeout,
                execution_profile="read_tuples",
            )
            has_chunk_sub_partitions = dict(result)
        else:
            has_chunk_sub_partitions = dict.fromkeys(chunks, False)
 
        # Check what kind of tables we want to query, if chunk list is empty
        # then use tbales which should exist in the schema.
        if has_chunk_sub_partitions:
            have_subchunks = any(has_chunk_sub_partitions.values())
            have_non_subchunks = not all(has_chunk_sub_partitions.values())
        else:
            have_subchunks = context.schema.has_chunk_sub_partitions
            have_non_subchunks = not have_subchunks
 
        # NOTE: if an existing database is migrated and has both types of chunk
        # tables (e.g. DiaObjectChunks and DiaObjectChunks2) it is possible
        # that the same chunk can appear in both tables. In reality schema
        # migration should only happen during the downtime, so there will be
        # suffient gap and a different chunk ID will be used for new chunks.
 
        table_data: ApdbCassandraTableData | None = None
        table_data_subchunk: ApdbCassandraTableData | None = None
 
        table_name = context.schema.tableName(ExtraTables.replica_chunk_tables(False)[table])
        with self._timer("table_chunk_select_time", tags={"table": table_name}) as timer:
            if have_subchunks:
                replica_table = ExtraTables.replica_chunk_tables(True)[table]
                table_name = context.schema.tableName(replica_table)
                query = (
                    f'SELECT * FROM "{config.keyspace}"."{table_name}" '
                    "WHERE apdb_replica_chunk = ? AND apdb_replica_subchunk = ?"
                )
                statement = context.preparer.prepare(query)
 
                queries: list[tuple] = []
                for chunk in chunks:
                    if has_chunk_sub_partitions.get(chunk, False):
                        for subchunk in range(config.replica_sub_chunk_count):
                            queries.append((statement, (chunk, subchunk)))
                if not queries and not have_non_subchunks:
                    # Add a dummy query to return correct set of columns.
                    queries.append((statement, (-1, -1)))
 
                if queries:
                    table_data_subchunk = cast(
                        ApdbCassandraTableData,
                        select_concurrent(
                            context.session,
                            queries,
                            "read_raw_multi",
                            config.connection_config.read_concurrency,
                        ),
                    )
 
            if have_non_subchunks:
                replica_table = ExtraTables.replica_chunk_tables(False)[table]
                table_name = context.schema.tableName(replica_table)
                query = f'SELECT * FROM "{config.keyspace}"."{table_name}" WHERE apdb_replica_chunk = ?'
                statement = context.preparer.prepare(query)
 
                queries = []
                for chunk in chunks:
                    if not has_chunk_sub_partitions.get(chunk, True):
                        queries.append((statement, (chunk,)))
                if not queries and not table_data_subchunk:
                    # Add a dummy query to return correct set of columns.
                    queries.append((statement, (-1,)))
 
                if queries:
                    table_data = cast(
                        ApdbCassandraTableData,
                        select_concurrent(
                            context.session,
                            queries,
                            "read_raw_multi",
                            config.connection_config.read_concurrency,
                        ),
                    )
 
            # Merge if both are non-empty.
            if table_data and table_data_subchunk:
                table_data_subchunk.project(drop=["apdb_replica_subchunk"])
                table_data.append(table_data_subchunk)
            elif table_data_subchunk:
                table_data = table_data_subchunk
            elif not table_data:
                raise AssertionError("above logic is incorrect")
 
            timer.add_values(row_count=len(table_data.rows()))
 
            table_schema = self._apdb._schema.tableSchemas[table]
            # Regular tables should never have columns of ExtraDataTypes, this
            # is just to make mypy happy.
            column_types = {
                column.name: column.datatype
                for column in table_schema.columns
                if not isinstance(column.datatype, ExtraDataTypes)
            }
            column_types["apdb_replica_chunk"] = felis.datamodel.DataType.long
            # It may also have subchunk column, we do not always drop it, and
            # clients should not need it, but we need to provide type for it.
            column_types["apdb_replica_subchunk"] = felis.datamodel.DataType.int
            table_data.set_column_types(column_types)
 
        return table_data
 
    def getTableUpdateChunks(self, table: ApdbTables, chunks: Iterable[int]) -> ApdbTableData:
        # docstring is inherited from a base class
        raise NotImplementedError()