创建两个集群而不是一个集群的galera puppet代码

Question

我有一个木偶代码，它应该创建一个包含两个节点的galera集群，但实际上它创建了两个集群，每个集群都有一个节点。

这两个节点的名称分别为testbox1和testbox2

以下是我的./hiera/role/testbox.yaml

---
classes:
  - '::galera'

selinux::mode: 'permissive'

yum::repos::enabled:
  - percona-x86_64
yum::repos:
  contrail-3.2.1-mitaka:
    enabled: 0

packages:
  - 'Percona-XtraDB-Cluster-shared-compat-57'
  - 'Percona-Server-selinux-56'

galera::configure_repo: false
galera::package_ensure: 'present'
galera::galera_package_ensure: 'absent'
galera::galera_package_name: 'Percona-XtraDB-Cluster-galera-3'
galera::client_package_name: 'Percona-XtraDB-Cluster-client-57'
galera::mysql_package_name: 'Percona-XtraDB-Cluster-server-57'
galera::bootstrap_command: 'systemctl start mysql@bootstrap.service'


galera::mysql_service_name: 'mysql'
mysql::server_service_name: 'mysql'
galera::service_enabled: true
galera::mysql_restart: true

galera::configure_firewall: false
mysql::server::purge_conf_dir: true
galera::purge_conf_dir: true

galera::grep_binary: '/bin/grep'
galera::mysql_binary: '/usr/bin/mysql'
galera::rundir: '/var/run/mysqld'
galera::socket: '/var/lib/mysql/mysql.sock'

galera::create_root_user: true
galera::create_root_my_cnf: true
galera::create_status_user: true
galera::status_check: true

galera::galera_servers: ['testbox-1', 'testbox-2']
galera::galera_master: 'testbox-1'
galera::status_password: 'bla'
galera::bind_address: '0.0.0.0'
galera::override_options:
  mysqld:
    pxc_strict_mode: 'ENFORCING'
    wsrep_provider: '/usr/lib64/galera3/libgalera_smm.so'
    wsrep_slave_threads: 8
    wsrep_sst_method: 'rsync'
    wsrep_cluster_name: 'grafana-galera-cluster'
    wsrep_node_address: "%{ipaddress}"
    wsrep_node_name: "%{hostname}"
    wsrep_sst_auth: "sstuser:%{hiera('galera::sstuser_password')}"
    binlog_format: 'ROW'
    default_storage_engine: 'InnoDB'
    innodb_locks_unsafe_for_binlog: 1
    innodb_autoinc_lock_mode: 2
    innodb_buffer_pool_size: '40000M'
    innodb_log_file_size: '100M'
    query_cache_size: 0
    query_cache_type: 0
    datadir: '/var/lib/mysql'
    socket: '/var/lib/mysql/mysql.sock'
    log-error: '/var/log/mysqld.log'
    pid-file: '/var/run/mysql/mysql.pid'
    max_connections: '10000'
    max_connect_errors: '10000000'
  mysqld_safe:
    log-error: '/var/log/mysqld.log'

galera::status_user: 'clustercheck'
galera::status_allow: '%'
galera::status_available_when_donor: 0
galera::status_available_when_readonly: -1
galera::status_host: 'localhost'
galera::status_log_on_success: ''
galera::status_log_on_success_operator: '='
galera::status_port: 9200

galera::validate::action: 'select count(1);'
galera::validate::catch: 'ERROR'
galera::validate::delay: 3
galera::validate::inv_catch: undef
galera::validate::retries: 20

我使用的是fraenki/galera模块

这段代码的问题是，我最终在一个集群中使用testbox1，在另一个集群中使用testbox2，而不是将它们都放在同一个集群中，在排除了与jira.percona.com/browse/PXC-2258相关的问题后，我发现puppet代码将创建没有wsrep_cluster_address值的wsrep.cnf，这将覆盖具有正确值的/etc/my.cnf.d/server.cnf。我知道如何通过删除wsrep.cnf手动修复这个问题，但我希望让Puppet在没有我手动修复的情况下做到这一点，但我不知道怎么做。

puppet版本3.8.7 (开源)(我无法升级它)

Answer 1

mysql@bootstrap只需要在一个节点上执行。另一个节点进行正常启动，然后它将从第一个节点开始SST。

对于两个节点，您将很难获得仲裁，并且作为HA系统是不可行的。