Server IP : 66.29.132.122 / Your IP : 18.116.40.28 Web Server : LiteSpeed System : Linux business142.web-hosting.com 4.18.0-553.lve.el8.x86_64 #1 SMP Mon May 27 15:27:34 UTC 2024 x86_64 User : admazpex ( 531) PHP Version : 7.2.34 Disable Function : NONE MySQL : OFF | cURL : ON | WGET : ON | Perl : ON | Python : ON | Sudo : OFF | Pkexec : OFF Directory : /proc/self/root/proc/self/root/proc/thread-self/root/proc/thread-self/root/proc/self/root/proc/thread-self/root/proc/self/root/proc/self/root/proc/self/root/proc/self/root/opt/cloudlinux/venv/lib64/python3.11/site-packages/lvestats/snapshots/ |
Upload File : |
# 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 import datetime import sys import prettytable from sqlalchemy import or_ from sqlalchemy.orm import sessionmaker import lvestats.lib.commons.decorators from lvestats.lib import dbengine, uidconverter from lvestats.lib.commons import dateutil from lvestats.lib.jsonhandler import prepare_data_json from lvestats.lib.parsers.lve_read_snapshot_argparse import lve_read_snapshot_parser from lvestats.lib.snapshot import Snapshot from lvestats.orm.incident import incident REPORT_HEADER = 'Snapshots collected starting from %s to %s for lve id %d @ %s:\n' REPORT_FOOTER = 'Done..\n' DEFAULT_SERVER_ID = 'localhost' # used when nothing is configured or specified in cmdline def _calculate_period(opts): if opts.period: start, end = opts.period elif opts.timestamp: start = opts.timestamp end = start + 0.999999 else: try: start = dateutil.parse_date(" ".join(opts.ffrom)) end = dateutil.parse_date(" ".join(opts.to)) except ValueError: print('please use [YY]YY-MM-DD[ HH:MM] format for --from and --to') return None, None return start, end @lvestats.lib.commons.decorators.no_sigpipe def snapshot_reader_main(config, argv_=None): parser = lve_read_snapshot_parser() opts = parser.parse_args(argv_) if not opts.id and not opts.user: parser.print_help() print('One of -u --user or -i --id should be specified') return 1 try: engine = dbengine.make_db_engine(config) except dbengine.MakeDbException as e: print(e) return 1 server_id = config.get('server_id', DEFAULT_SERVER_ID) if opts.user: uid = uidconverter.username_to_uid(opts.user, server_id, server_id, engine) if uid is None: print(f'User {opts.user}@{server_id} not found') return 1 else: uid = opts.id start, end = _calculate_period(opts) if start is None and end is None: return 1 lve_read_snapshot = LVEReadSnaphot( engine, start, end, uid, server_id, opts.output, opts.json, ) if opts.list: lve_read_snapshot.list() # show snapshots timestamps list elif opts.stats: lve_read_snapshot.stats(opts.unit) else: lve_read_snapshot.run() def _try_convert_to_timestamp(o): """ Convert local datetime to unix timestamp, or just passes unix timestamp as output if specified. :param o: :return: """ if isinstance(o, datetime.datetime): return dateutil.gm_datetime_to_unixtimestamp(dateutil.local_to_gm(o)) return o class LVEReadSnaphot(object): def __init__(self, engine, start, end, uid, server_id, output_file, do_json): """ :param start: datetime.datetime | int (unix timestamp) :param end: datetime.datetime | int (unix timestamp) :param uid: :param server_id: :param output_file: filename :param do_json: boolean """ self.engine = engine self.do_json = do_json self.output_file = output_file self.uid = uid self.start = _try_convert_to_timestamp(start) self.end = _try_convert_to_timestamp(end) self.server_id = server_id # Faults names dictionary self.fault_names = { 'cpu_fault': 'CPU', 'mem_fault': 'Virtual memory', 'mep_fault': 'EP', 'memphy_fault': 'Physical memory', 'nproc_fault': 'NPROC', 'io_fault': 'IO', 'iops_fault': 'IOPS' } def get_incidents(self, session): return session.query(incident).filter(incident.uid == self.uid, incident.server_id == self.server_id, or_(incident.incident_start_time.between(self.start, self.end), incident.incident_end_time.between(self.start, self.end)) ).order_by(incident.incident_start_time).all() def stats_by_incident(self, session): result = [] for i in self.get_incidents(session): result.append({'from': i.incident_start_time, 'to': max(i.dump_time, i.incident_end_time or 0), 'incidents': 1, 'snapshots': i.snapshot_count, 'duration': self.get_duration(i) }) return result @staticmethod def get_duration(i, from_=0, to_=sys.maxsize): from_ = max(i.incident_start_time, from_) to_ = min(max(i.dump_time, i.incident_end_time or 0), to_) return to_ - from_ @staticmethod def get_incident_count(incidents, pos, from_ts, to_ts): count = 0 duration = 0 while pos < len(incidents): i = incidents[pos] if i.dump_time < from_ts: pos += 1 continue if i.incident_start_time > to_ts: break # we are done count += 1 pos += 1 duration += LVEReadSnaphot.get_duration(i, from_ts, to_ts) if count == 0: return 0, 0, pos else: return count, duration, pos - 1 def stats_by_time_unit(self, session, time_unit): incidents = self.get_incidents(session) snapshot_files = Snapshot({'uid': self.uid}).get_file_list() result = [] from_ts = self.start pos = 0 while from_ts < self.end: to_ts = min(from_ts + time_unit, self.end) incident_count, duration, pos = self.get_incident_count(incidents, pos, from_ts, to_ts) if incident_count == 0: # skip this one, we have nothing here from_ts = to_ts continue snapshots = Snapshot.snapshot_filter(snapshot_files, from_ts, to_ts) if len(snapshots) == 0: snapshots.append(0) # always show like there is at least one snapshot result.append({'from': from_ts, 'to': to_ts, 'incidents': incident_count, 'snapshots': len(snapshots), 'duration': duration }) from_ts = to_ts return result def print_stats_json(self, stats): data = { 'from': self.start, 'to': self.end, 'stats': stats } out = self.open() out.write(prepare_data_json(data)) out.write('\n') out.flush() def print_stats(self, stats): out = self.open() out.write(f'Stats from {dateutil.ts_to_iso(self.start)} to {dateutil.ts_to_iso(self.end)}\n') for stat in stats: out.write('---\n') out.write(f'\tfrom: {dateutil.ts_to_iso(stat["from"])}\n') out.write(f'\tto: {dateutil.ts_to_iso(stat["to"])}\n') out.write(f'\tincidents: {stat["incidents"]}\n') out.write(f'\tsnapshots: {stat["snapshots"]}\n') out.write(f'\tduration: {stat["duration"]} sec.\n') out.flush() def stats(self, stats_unit_str): try: time_unit = dateutil.parse_period2(stats_unit_str) if self.end - self.start < time_unit: # this prevents situations when we get stats for last 10 minutes, but group it by 1 day self.start = self.end - time_unit group_by_incident = False except ValueError: time_unit = None group_by_incident = stats_unit_str == 'auto' if not group_by_incident: raise session = sessionmaker(bind=self.engine)() try: if group_by_incident: stats = self.stats_by_incident(session) else: stats = self.stats_by_time_unit(session, time_unit) session.expunge_all() if self.do_json: self.print_stats_json(stats) else: self.print_stats(stats) finally: session.close() def run(self): snapshots = Snapshot({'uid': self.uid}) self.report(snapshots.get_snapshots(self.start, self.end)) def list(self): snapshots = Snapshot({'uid': self.uid}) snapshots_list = snapshots.get_ts_list(self.start, self.end) out = self.open() if self.do_json: out.write(prepare_data_json(snapshots_list)) else: out.write( f'Snapshots timestamp list; from {dateutil.ts_to_iso(self.start)} ' f'to {dateutil.ts_to_iso(self.end)} for lve id {self.uid}\n' ) for ts in snapshots_list: out.write(dateutil.ts_to_iso(ts)) out.write('\n') out.write(REPORT_FOOTER) out.flush() def report(self, snapshots): out = self.open() if self.do_json: out.write(prepare_data_json({ 'snapshots': snapshots, })) out.write('\n') out.flush() return out.write(REPORT_HEADER % (dateutil.ts_to_iso(self.start), dateutil.ts_to_iso(self.end), self.uid, self.server_id)) for snapshot_data in snapshots: self.format_snapshot(out, snapshot_data) out.write(REPORT_FOOTER) out.flush() def open(self): if self.output_file: try: return open(self.output_file, "w", encoding='utf-8') except IOError: pass # maybe need error message # fixme --> if we are trying to write to a file, and cannot, # this is an error, we shouldn't write to stdout return sys.stdout @staticmethod def _process_data_aggregate(process_data): """ Aggregates process data by PID by summing CPU % and MEM for same PIDs :param process_data: input data. Dictionary: { u'151048': {u'MEM': u'1', u'CMD': u'bash', u'PID': u'151048', u'CPU': u'0%'}, u'151047': {u'MEM': u'1', u'CMD': u'su cltest1', u'PID': u'151047', u'CPU': u'0%'}, u'153642': {u'MEM': u'1', u'CMD': u'./threads', u'PID': u'153640', u'CPU': u'0%'}, u'153641': {u'MEM': u'1', u'CMD': u'./threads', u'PID': u'153640', u'CPU': u'0%'}, u'153640': {u'MEM': u'1', u'CMD': u'./threads', u'PID': u'153640', u'CPU': u'5%'} } :return: Output data - List of dictionaries: [ {u'MEM': u'1', u'CMD': u'bash', u'PID': u'151048', u'CPU': u'0%'}, {u'MEM': u'1', u'CMD': u'su cltest1', u'PID': u'151047', u'CPU': u'0%'}, {u'MEM': u'3', u'CMD': u'./threads', u'PID': u'153640', u'CPU': u'5%'}, ] """ # 1. Build thread dictionary as # pid: {'PID', 'CMD', 'MEM', 'CPU'} # and aggregate data thread_dict = {} for _, proc_data in process_data.items(): if 'PID' not in proc_data: # old format snapshot, do not aggregate it # Example of old format snapshot: # {u'31228': {u'MEM': u'1', u'CMD': u'31228', u'IOPS': u'N/A', u'CPU': u'1%', u'IO': u'N/A'}} pid = proc_data['CMD'] process_data_new = {} process_data_new['PID'] = pid process_data_new['MEM'] = proc_data['MEM'] process_data_new['CMD'] = pid process_data_new['CPU'] = proc_data['CPU'] thread_dict[pid] = process_data_new continue pid = proc_data['PID'] # remove '%' from CPU value and convert CPU/MEM to integers if proc_data['CPU'] != 'N/A': proc_data['CPU'] = int(proc_data['CPU'].replace('%', '')) if proc_data['MEM'] != 'N/A': proc_data['MEM'] = int(proc_data['MEM']) if pid in thread_dict: # PID already present, add new data to it if proc_data['CPU'] != 'N/A': if thread_dict[pid]['CPU'] != 'N/A': thread_dict[pid]['CPU'] += proc_data['CPU'] else: thread_dict[pid]['CPU'] = proc_data['CPU'] if proc_data['MEM'] != 'N/A': if thread_dict[pid]['MEM'] != 'N/A': thread_dict[pid]['MEM'] += proc_data['MEM'] else: thread_dict[pid]['MEM'] = proc_data['MEM'] else: # PID absent, add it thread_dict[pid] = proc_data # 2. Build output list out_data = list(thread_dict.values()) return out_data def format_snapshot(self, out, snapshot_data): out.write(f'>>> {dateutil.ts_to_iso(snapshot_data["dump_time"])}, UID {snapshot_data["uid"]}\n') out.write('\nFaults:\n') for k, v in snapshot_data['snap_faults'].items(): out.write(f'\t* {self.fault_names.get(k, k)}: {v}\n') if snapshot_data['snap_sql']: out.write('\nSQL Queries:\n') sql_table = prettytable.PrettyTable(['CMD', 'Time', 'SQL-query']) list(map(sql_table.add_row, snapshot_data['snap_sql'])) out.write(sql_table.get_string()) out.write('\nProcesses:\n') # fields = ('PID', 'COM', 'SPEED', 'MEM', 'IO', 'IOPS') # table = prettytable.PrettyTable(fields=fields) fields = set() for data in list(snapshot_data['snap_proc'].values()): for key in list(data.keys()): fields.add(key) # Keys list for data extacting data_keys = list(['PID']) # Form table header: PID, CMD, Memory (Mb), CPU (%) table_columns = ['PID'] if 'MEM' in fields: table_columns.append('Memory (Mb)') data_keys.append('MEM') if 'CPU' in fields: table_columns.append('CPU (%)') data_keys.append('CPU') if 'CMD' in fields: table_columns.append('CMD') data_keys.append('CMD') table = prettytable.PrettyTable(table_columns) # Left align for CMD column, if it present if 'CMD' in table_columns: table.align['CMD'] = 'l' # Do process data aggregation (CPU/MEM summing for all threads of same process) snap_proc_aggr = self._process_data_aggregate(snapshot_data['snap_proc']) for data in snap_proc_aggr: table.add_row([data.get(k, 'N/A') for k in data_keys]) out.write(str(table)) out.write('\n\n') if snapshot_data['snap_http']: out.write('Http requests:\n') http_table = prettytable.PrettyTable(['Pid', 'Domain', 'Http type', 'Path', 'Http version', 'Time']) list(map(http_table.add_row, snapshot_data['snap_http'])) out.write(str(http_table)) out.write('\n\n')