技術めも

----------------------------------------
master
IP:192.168.35.11
----------------------------------------
     |                       |
     |(semi sync)            |(semi sync)
     |                       |
    --------------------    --------------------
    slave                   slave
    IP:192.168.35.12        IP:192.168.35.13
    --------------------    --------------------

# マルチソースをやる際に必要かもしれないので、レプリケーションクラスターで共通の数字
gtid_domain_id=11

# レプリケーションクラスター間で一意の数字（IPアドレスの下2桁など）
server_id={{ server_id }}

# 名前解決ができない環境向けに、自身のノードのIPアドレスを設定しておく
report_host={{ ip_addr }}

# とりあえず0だが、動的に変更する
read_only=0

# 安定のrow
binlog_format=row

# 毎コミットごとにバイナリログに書き出す
sync_binlog=1

# フェイルオーバー時にslaveが昇格しても大丈夫なように
log_slave_updates=1

# レプリケーション安全設定
relay_log_recovery=1

# 大きなリレーログへの対応
slave_max_allowed_packet=1G

# タイムアウト長めに
slave_net_timeout=3600

# slaveのSQLスレッドの並列化
slave_parallel_threads={{ cpu_num }}

# 厳格モード
gtid_strict_mode=1

# 以下、準同期設定（MariaDB10.3では、plugin_load_addは不要）
rpl_semi_sync_master_enabled=1
rpl_semi_sync_master_timeout=1500
rpl_semi_sync_slave_enabled=1
rpl_semi_sync_master_wait_point=AFTER_SYNC

{
  "Debug": false,
  "EnableSyslog": false,
  "ListenAddress": ":3000",
  "MySQLTopologyUser": "adminuser",
  "MySQLTopologyPassword": "adminpassword",
  "BackendDB": "sqlite",
  "SQLite3DataFile": "/usr/local/orchestrator/orchestrator.sqlite3",
  "DiscoverByShowSlaveHosts": true,
  "InstancePollSeconds": 5,
  "UnseenInstanceForgetHours": 240,
  "SnapshotTopologiesIntervalHours": 0,
  "InstanceBulkOperationsWaitTimeoutSeconds": 10,
  "HostnameResolveMethod": "default",
  "MySQLHostnameResolveMethod": "@@report_host",
  "SkipBinlogServerUnresolveCheck": true,
  "ExpiryHostnameResolvesMinutes": 60,
  "RejectHostnameResolvePattern": "",
  "ReasonableReplicationLagSeconds": 10,
  "ProblemIgnoreHostnameFilters": [],
  "VerifyReplicationFilters": false,
  "ReasonableMaintenanceReplicationLagSeconds": 20,
  "CandidateInstanceExpireMinutes": 60,
  "DetectClusterAliasQuery": "SELECT @@report_host",
  "PromotionIgnoreHostnameFilters": [],
  "DetectSemiSyncEnforcedQuery": "",
  "URLPrefix": "",
  "StatusEndpoint": "/api/status",
  "StatusSimpleHealth": true,
  "StatusOUVerify": false,
  "AgentPollMinutes": 60,
  "UnseenAgentForgetHours": 6,
  "StaleSeedFailMinutes": 60,
  "SeedAcceptableBytesDiff": 8192,
  "PseudoGTIDPattern": "",
  "PseudoGTIDPatternIsFixedSubstring": false,
  "PseudoGTIDMonotonicHint": "asc:",
  "DetectPseudoGTIDQuery": "",
  "BinlogEventsChunkSize": 10000,
  "SkipBinlogEventsContaining": [],
  "ReduceReplicationAnalysisCount": true,
  "FailureDetectionPeriodBlockMinutes": 60,
  "RecoveryPeriodBlockSeconds": 3600,
  "RecoveryIgnoreHostnameFilters": [],
  "RecoverMasterClusterFilters": [
    "*"
  ],
  "RecoverIntermediateMasterClusterFilters": [
    "*"
  ],
  "OnFailureDetectionProcesses": [],
  "PreFailoverProcesses": [],
  "PostFailoverProcesses": [
    "echo -e 'Failed server is {failedHost}:{failedPort}.\nSuccessed server is {successorHost}:{successorPort}.' | mail -s 'Complete Orchestrator Failover' yourname@sample.com"
  ],
  "PostUnsuccessfulFailoverProcesses": [
    "echo -e 'Failed server is {failedHost}:{failedPort}.' | mail -s 'Failed Orchestrator Failover' yourname@sample.com"
  ],
  "PostMasterFailoverProcesses": [
    "/usr/local/bin/detect_orchestrator_failover.py -u adminuser -p adminpassword -H {successorHost} -P {successorPort}"
  ],
  "PostIntermediateMasterFailoverProcesses": [],
  "CoMasterRecoveryMustPromoteOtherCoMaster": true,
  "DetachLostSlavesAfterMasterFailover": true,
  "ApplyMySQLPromotionAfterMasterFailover": true,
  "MasterFailoverDetachSlaveMasterHost": false,
  "MasterFailoverLostInstancesDowntimeMinutes": 0,
  "PostponeSlaveRecoveryOnLagMinutes": 0
}

"MySQLHostnameResolveMethod": "@@report_host"
"DetectClusterAliasQuery": "SELECT @@report_host"

"PostMasterFailoverProcesses"

/usr/local/bin/detect_orchestrator_failover.py

#!/usr/bin/python
# -*- coding: utf-8 -*-
import os
import sys
import traceback
import datetime
import re
import argparse
import MySQLdb

parser = argparse.ArgumentParser(description = "detect orchestrator failover.")
parser.add_argument("-u", type=str, help = "(required) mysql user", required=True)
parser.add_argument("-p", type=str, help = "(required) mysql pass", required=True)
parser.add_argument("-H", type=str, help = "(required) master host", required=True)
parser.add_argument("-P", type=str, help = "(required) master port", required=True)
args = parser.parse_args()

mysql_user = args.u
mysql_pass = args.p
master_host = args.H
master_port = args.P

fp = None
master_conn = None
master_c = None
try:
  # log
  log_path = '/path_to_your_dir/detect_orchestrator_failover.log'
  fp = open(log_path, 'a')

  # connect master
  master_conn = MySQLdb.connect(
    user=mysql_user,
    passwd=mysql_pass,
    host=master_host,
    port=int(master_port),
    db='mysql'
  )
  master_c = master_conn.cursor(MySQLdb.cursors.DictCursor)
  fp.write('[{}]new master is {}:{}.\n'.format(datetime.datetime.now(), master_host, master_port))

  # set master
  master_queries = []
  master_queries.append('set global read_only = 0')
  for sql in master_queries:
    master_c.execute(sql)
    fp.write('[{}]new master query is {}.\n'.format(datetime.datetime.now(), sql))

  # get slaves
  sql = "show slave hosts"
  master_c.execute(sql)
  rows = master_c.fetchall()
  if isinstance(rows, tuple):
    for row in rows:
      slave_conn = None
      slave_c = None
      try:
        slave_host = row['Host']
        slave_port = row['Port']

        # connect slave
        slave_conn = MySQLdb.connect(
          user=mysql_user,
          passwd=mysql_pass,
          host=slave_host,
          port=int(slave_port),
          db='mysql'
        )
        slave_c = slave_conn.cursor(MySQLdb.cursors.DictCursor)
        fp.write('[{}]new slave is {}:{}.\n'.format(datetime.datetime.now(), slave_host, slave_port))

        # set slave
        slave_queries = []
        slave_queries.append('set global read_only = 1')
        for sql in slave_queries:
          slave_c.execute(sql)
          fp.write('[{}]new slave query is {}.\n'.format(datetime.datetime.now(), sql))

      except:
        fp.write('[{}]{}\n'.format(datetime.datetime.now(), traceback.format_exc()))
      finally:
          if slave_c is not None:
            slave_c.close()
          if slave_conn is not None:
            slave_conn.close()

except:
  fp.write('[{}]{}\n'.format(datetime.datetime.now(), traceback.format_exc()))
finally:
  if master_c is not None:
    master_c.close()
  if master_conn is not None:
    master_conn.close()
  if fp is not None:
    fp.close()