Mini Shell

# coding=utf-8
#
# Copyright © Cloud Linux GmbH & Cloud Linux Software, Inc 2010-2019 All Rights Reserved
#
# Licensed under CLOUD LINUX LICENSE AGREEMENT
# http://cloudlinux.com/docs/LICENSE.TXT

from math import ceil

from sqlalchemy import Integer, desc
from sqlalchemy import and_, select, func, case, cast
from sqlalchemy import or_

from lvestats.lib import uidconverter
from lvestats.lib.commons.dateutil import (
    gm_datetime_to_unixtimestamp,
    gm_to_local,
    unixtimestamp_to_gm_datetime,
    local_to_gm,
    round_1m,
)
from lvestats.lib.lveinfolib import OutputFormatter
from lvestats.lib.lveinfolib import convert_to_list, enumerate_duplicate_columns, add_labes_to_column
from lvestats.lib.lveinfolib import dyn_time_unit_groups
from lvestats.orm import history_gov


def get_bitwise_restrict(cause_of_restrict):
    """
    :param sqlalchemy.orm.attributes.InstrumentedAttribute cause_of_restrict:
    :return sqlalchemy.sql.elements.BinaryExpression:
    """
    return (
        func.max(cause_of_restrict.op('&')(4))
        + func.max(cause_of_restrict.op('&')(2))
        + func.max(cause_of_restrict.op('&')(1))
    )


def get_cause_of_restrict(cause_of_restrict):
    """
    :param sqlalchemy.sql.elements.BinaryExpression cause_of_restrict:
    :return sqlalchemy.sql.elements.Case:
    """
    return case(
        [
            (cause_of_restrict == 1, 'C'),
            (cause_of_restrict == 2, 'R'),
            (cause_of_restrict == 3, 'CR'),
            (cause_of_restrict == 4, 'W'),
            (cause_of_restrict == 5, 'CW'),
            (cause_of_restrict == 6, 'RW'),
            (cause_of_restrict == 7, 'CRW'),
        ],
        else_='',
    )


class HistoryShowDBGov(object):
    DEFAULT_ORDER = 'ts'

    def __init__(
        self,
        dbengine,
        period_from,
        period_to,
        uid=None,
        server_id='localhost',
        show_columns=None,
        order_by=None,
        limit=None,
        cfg=None,
        time_unit=None,
        reverse=False,
        by_usage=None,
        by_usage_percentage=0.9,
    ):
        """
        :param bool reverse:
        :param sqlalchemy.engine.base.Engine dbengine:
        :param datetime.datetime period_from:
        :param datetime.datetime period_to:
        :param int|None uid:
        :param str server_id:
        :param str|list|tuple|None show_columns:
        :param str order_by:
        :param int|None limit:
        :param dict|None cfg:
        :param int|None time_unit:
        :param bool reverse:
        :param list[str]|str|None by_usage:
        :param float by_usage_percentage: value in range [0..1]
        :return list:
        """
        self.by_usage_percentage = by_usage_percentage
        self.by_usage = by_usage
        self.dbengine = dbengine
        self.period_from = period_from
        self.period_to = period_to
        self.uid = uid
        self.server_id = server_id
        self.requested_columns = show_columns
        self.order_by = order_by
        self.limit = limit
        self.cfg = cfg or {'server_id': 'localhost'}
        if time_unit == -1:
            self.time_unit = None
        else:
            self.time_unit = time_unit
        self.reverse = reverse

    def history_dbgov_show_dynamic(self):
        """
        Show dbgov statistics with dynamic time-unit.
        :rtype: list
        """
        if self.uid is None:
            return []

        # get first and last record time
        max_ts, min_ts = self._get_min_max_ts(self.uid)
        period_from = max(gm_to_local(unixtimestamp_to_gm_datetime(min_ts)), gm_to_local(self.period_from))
        period_to = min(gm_to_local(unixtimestamp_to_gm_datetime(max_ts)), gm_to_local(self.period_to))

        # remove seconds from time
        period_from = round_1m(period_from)
        period_to = round_1m(period_to)

        rows = []
        for t_from, t_to, time_unit in reversed(dyn_time_unit_groups(period_from, period_to)):
            self.period_from = local_to_gm(t_from)
            self.period_to = local_to_gm(t_to)
            self.time_unit = time_unit
            rows += self.history_dbgov_show()
        return rows

    def history_dbgov_show(self):
        """
        Show statistics from history_gov table
        :rtype: list
        """
        # check and return some data without run sql query
        if self.uid == tuple() or self.uid == []:
            return []
        if self.uid is not None and not isinstance(self.uid, (list, tuple)) and self.uid <= 0:
            return []

        t_from = gm_datetime_to_unixtimestamp(self.period_from)
        t_to = gm_datetime_to_unixtimestamp(self.period_to)

        # convert 0 to None
        limit = self.limit or None

        # get information how to aggregate info in different columns
        # e.g. 'ts': sum(history_gov.ts) or 'cpu': avg(history_gov.cpu)
        table_aggregate_info = self._get_table_aggregate_info(t_from, t_to)

        # get list of columns to select from database
        columns_to_select = self._get_columns_for_select(table_aggregate_info)

        # order_by query
        order_by_query = self._get_order_by_query(table_aggregate_info)

        where_query = and_(history_gov.ts.between(t_from, t_to), history_gov.server_id == self.server_id)

        # query to group
        group_by_time_query = self.round_time(t_from)
        if isinstance(self.uid, int):
            query = (
                select(enumerate_duplicate_columns(columns_to_select))
                .where(and_(where_query, history_gov.uid == self.uid))
                .order_by(order_by_query)
            )
            if self.time_unit is not None:
                query = query.group_by(group_by_time_query, history_gov.uid).limit(limit)
            else:
                query = query.group_by(group_by_time_query, history_gov.uid, history_gov.cause_of_restrict).limit(limit)
        else:
            query = select(enumerate_duplicate_columns(columns_to_select)).where(where_query).order_by(order_by_query)
            # filter by user list if need
            if isinstance(self.uid, (list, tuple)):
                query = query.where(history_gov.uid.in_(list(self.uid)))
            # group and limit
            query = query.group_by(
                history_gov.server_id,
                history_gov.uid,
            ).limit(limit)

        if self.by_usage:
            query = query.having(self._get_by_usage_query(table_aggregate_info))

        rows = self.dbengine.execute(query)

        # replace uid with username and normalise types
        results = OutputFormatter(
            fields=self.requested_columns,
            rows=rows.fetchall(),
            orders=[
                (
                    ['USER'],
                    lambda uid_: uidconverter.uid_to_username(
                        int(uid_), self.cfg['server_id'], self.server_id, self.dbengine
                    ),
                ),
                (['CPU', 'READ', 'WRITE'], lambda x: x if x is None else round(float(x), 7)),
                (['TS', 'FROM', 'TO', 'lCPU', 'lREAD', 'lWRITE', 'CON'], lambda x: x if x is None else int(x)),
            ],
        )
        rows.close()
        return results.get_corrected_list()

    def _get_columns_for_select(self, table_aggregate_info):
        """
        Return list of columns that should be selected from table.
        :param table_aggregate_info: dict from _get_table_aggregate_info
        :return: query for each column
        :rtype: list
        """
        columns_to_select = []
        for column_name in convert_to_list(self.requested_columns):
            columns_to_select.append(table_aggregate_info[column_name.lower()])
        return columns_to_select

    def _get_table_aggregate_info(self, t_from, t_to):
        """
        Returns dictionary:
        {
         column_name: aggregate_query
        }
        Ignores t_from and t_to if self.uid is not None;

        :type t_from: int
        :type t_to: int
        :rtype: dict
        """
        if self.uid is None or isinstance(self.uid, list):
            columns_info = self.get_history_gov_all_users_func(t_to - t_from)
        else:
            columns_info = self.get_history_gov_all_users_func()
        return columns_info

    def _get_min_max_ts(self, uid):
        """
        Get min and max timestamp for given user id.
        :rtype: tuple(int, int)
        """
        query = select([func.min(history_gov.ts), func.max(history_gov.ts)]).where(history_gov.uid == uid)
        min_ts, max_ts = self.dbengine.execute(query).fetchall()[0]
        return max_ts, min_ts

    def _get_order_by_query(self, history_gov_aggregate_func):
        """
        Create order_by query
        :param history_gov_aggregate_func:
        :return: sqlalchemy query
        """
        # generate order by
        if self.order_by is None or self.order_by == 'con':  # con - deprecated; use default
            self.order_by = self.DEFAULT_ORDER
        if self.order_by == 'ts' and self.uid is None:
            self.order_by = 'id'
        order_by_query = history_gov_aggregate_func[self.order_by.lower()]

        if self.order_by not in ('id', 'con', 'ts') and self.reverse:
            order_by_query = desc(order_by_query)
        return order_by_query

    def _get_by_usage_query(self, table_aggregate_info):
        """
        Get by_usage query.
        :param table_aggregate_info: dict from _get_table_aggregate_info
        :return:
        """
        filter_list = []
        for by_usage_item in convert_to_list(self.by_usage):
            if by_usage_item.lower() == 'cpu':
                filter_list.append(
                    (table_aggregate_info['cpu'] / table_aggregate_info['lcpu']) > self.by_usage_percentage
                )
            elif by_usage_item.lower() == 'io':
                filter_list.append(
                    (table_aggregate_info['io'] / table_aggregate_info['lio']) > self.by_usage_percentage
                )
            else:
                raise NotImplementedError()
        return or_(*filter_list)

    def get_history_gov_all_users_func(self, time_period=None):
        """
        :type time_period: int
        :rtype: dict
        """
        t_from = gm_datetime_to_unixtimestamp(self.period_from)

        # round time depending on time_unit
        normalized_from = self.round_time(t_from, offset=0, ts_query=func.min(history_gov.ts))
        normalized_to = self.round_time(t_from, offset=1, ts_query=func.min(history_gov.ts))

        time_interval = time_period or self.time_unit
        if time_interval is None:
            column_labels = add_labes_to_column(
                {
                    'from': normalized_from,
                    'to': normalized_to,
                    'ts': history_gov.ts,
                    'id': history_gov.uid,
                    'user': history_gov.uid,
                    'cpu': func.max(history_gov.sum_cpu),
                    'read': func.sum(history_gov.sum_read),
                    'write': func.sum(history_gov.sum_write),
                    'io': func.avg(history_gov.sum_write + history_gov.sum_read),
                    'lio': func.max(history_gov.limit_write_on_period_end + history_gov.limit_read_on_period_end),
                    'con': 0,  # 'CON' deprecated every time 0
                    'lcpu': func.max(history_gov.limit_cpu_on_period_end),
                    'lread': func.max(history_gov.limit_read_on_period_end),
                    'lwrite': func.max(history_gov.limit_write_on_period_end),
                    'restrict': get_cause_of_restrict(history_gov.cause_of_restrict),
                }
            )
        else:
            minutes_passed = int(ceil(time_interval / 60.0))
            column_labels = add_labes_to_column(
                {
                    'from': normalized_from,
                    'to': normalized_to,
                    'ts': func.max(history_gov.ts),
                    'id': history_gov.uid,
                    'user': history_gov.uid,
                    'cpu': func.sum(history_gov.sum_cpu) / minutes_passed,
                    'read': func.sum(history_gov.sum_read) / minutes_passed,
                    'write': func.sum(history_gov.sum_write) / minutes_passed,
                    'con': 0,  # 'CON' deprecated every time 0
                    'io': func.sum(history_gov.sum_write + history_gov.sum_read) / minutes_passed,
                    'lio': func.max(history_gov.limit_write_on_period_end + history_gov.limit_read_on_period_end),
                    'lcpu': func.max(history_gov.limit_cpu_on_period_end),
                    'lread': func.max(history_gov.limit_read_on_period_end),
                    'lwrite': func.max(history_gov.limit_write_on_period_end),
                    'restrict': get_cause_of_restrict(get_bitwise_restrict(history_gov.cause_of_restrict)),
                }
            )
        return column_labels

    def round_time(self, time_from, offset=0, ts_query=None):
        """
        Round time depending on time_unit.
        Returns query, that returns nearest
         (N * time_unit + time_from) for current ts.
        :param ts_query: custom column
        :type time_from: int
        :type offset: int
        :return:
        """
        if ts_query is None:
            ts_query = history_gov.ts

        if self.time_unit:
            if self.dbengine.url.drivername == 'sqlite':
                # cast(..., Integer) using for compatibility with lve-stats-2.1-8 database; 'created' saved as float
                return cast(((ts_query - time_from) / self.time_unit + offset), Integer) * self.time_unit + time_from
            else:
                return func.floor((ts_query - time_from) / self.time_unit + offset) * self.time_unit + time_from
        else:
            return ts_query