PostgreSQL基于Patroni方案的高可用启动流程分析

什么是Patroni

在很多生产环境中，分布式数据库以高可用性、数据分布性、负载均衡等特性，被用户广泛应用。而作为高可用数据库的解决方案——Patroni，是专门为PostgreSQL数据库设计的，一款以Python语言实现的高可用架构模板。该架构模板，旨在通过外部共享存储软件（kubernetes、etcd、etcd3、zookeeper、aws等），实现 PostgreSQL 集群的自动故障恢复、自动故障转移、自动备份等能力。
主要特点：
1.自动故障检测和恢复：Patroni 监视 PostgreSQL 集群的健康状态，一旦检测到主节点故障，它将自动执行故障恢复操作，将其中一个从节点晋升为主节点。
2.自动故障转移：一旦 Patroni 定义了新的主节点，它将协调所有从节点和客户端，以确保它们正确地切换到新的主节点，从而实现快速、无缝的故障转移。
3.一致性和数据完整性：Patroni 高度关注数据一致性和完整性。在故障切换过程中，它会确保在新主节点接管之前，数据不会丢失或受损。
4.外部共享配置存储：Patroni 使用外部键值存储（如 ZooKeeper、etcd 或 Consul）来存储配置和集群状态信息。这确保了配置的一致性和可访问性，并支持多个 Patroni 实例之间的协作。
5.支持多种云环境和物理硬件：Patroni 不仅可以在云环境中运行，还可以部署在物理硬件上，提供了广泛的部署选项。
Patroni架构解析

●DCS（Distributed Configuration Store ）：是指分布式配置信息的存储位置，可支持kubernetes、etcd、etcd3、zookeeper、aws等存储媒介，由Patroni进行分布式配置信息的读写。
●核心Patroni：负责将分布式配置信息写入DCS中，并设置PostgreSQL节点的角色以及PostgreSQL配置信息，管理PostgreSQL的生命周期。
●PostgreSQL节点：各PostgreSQL节点，根据Patroni设置的PostgreSQL配置信息，生成主从关系链，以流复制的方式进行数据同步，最终生成一个PostgreSQL集群。
Patroni高可用源码分析
Patroni高可用启动流程

流程说明：
●加载集群信息，通过DCS支持的API接口，获取集群信息，主要内容如下：
○config：记录pg集群ID以及配置信息（包括pg参数信息、一些超时时间配置等），用于集群校验、节点重建等；
○leader：记录主节点选举时间、心跳时间、选举周期、最新的lsn等，用于主节点完成竞争后的信息记录；
○sync: 记录主节点和同步节点信息，由主节点记录，用于主从切换、故障转移的同步节点校验；
○failover: 记录最后一次故障转移的时间。
●集群状态检测，主要检测集群配置信息的内容校验，当前集群的整体状态及节点状态，判断通过什么方式来启动PostgreSQL；
●启动PostgreSQL，用于初始化PostgreSQL目录，根据集群信息设置相应的PostgreSQL配置信息，并启动；
●生成PostgreSQL集群，指将完成启动的PostgreSQL节点，通过设置主从角色，关联不同角色的PostgreSQL节点，最终生成完整的集群。
Patroni高可用启动流程分析
加载集群信息
加载集群信息，是高可用流程启动的第一步，也是生成PostgreSQL集群的最关键信息。

第一步，记载集群信息

…

try:
    self.load_cluster_from_dcs()
    self.state_handler.reset_cluster_info_state(self.cluster, self.patroni.nofailover)
except Exception:
    self.state_handler.reset_cluster_info_state(None, self.patroni.nofailover)
    raise
......

通过DCS接口加载集群信息

def load_cluster_from_dcs(self):
cluster = self.dcs.get_cluster()

# We want to keep the state of cluster when it was healthy
if not cluster.is_unlocked() or not self.old_cluster:
    self.old_cluster = cluster
self.cluster = cluster

if not self.has_lock(False):
    self.set_is_leader(False)

self._leader_timeline = None if cluster.is_unlocked() else cluster.leader.timeline

集群接口

def get_cluster(self, force=False):
    if force:
        self._bypass_caches()
    try:
        cluster = self._load_cluster()
    except Exception:
        self.reset_cluster()
        raise

self._last_seen = int(time.time())

with self._cluster_thread_lock:
    self._cluster = cluster
    self._cluster_valid_till = time.time() + self.ttl
    return cluster

@abc.abstractmethod
def _load_cluster(self):
    """Internally this method should build  `Cluster` object which
       represents current state and topology of the cluster in DCS.
       this method supposed to be called only by `get_cluster` method.

   raise `~DCSError` in case of communication or other problems with DCS.
   If the current node was running as a master and exception raised,
   instance would be demoted."""

以Kubernetes作为DCS为例

def _load_cluster(self):
    stop_time = time.time() + self._retry.deadline
    self._api.refresh_api_servers_cache()
    try:
        with self._condition:
            self._wait_caches(stop_time)

        members = [self.member(pod) for pod in self._pods.copy().values()]
        nodes = self._kinds.copy()

    config = nodes.get(self.config_path)
    metadata = config and config.metadata
    annotations = metadata and metadata.annotations or {}

    # get initialize flag
    initialize = annotations.get(self._INITIALIZE)

    # get global dynamic configuration
    config = ClusterConfig.from_node(metadata and metadata.resource_version,
                                     annotations.get(self._CONFIG) or '{}',
                                     metadata.resource_version if self._CONFIG in annotations else 0)

    # get timeline history
    history = TimelineHistory.from_node(metadata and metadata.resource_version,
                                        annotations.get(self._HISTORY) or '[]')

    leader = nodes.get(self.leader_path)
    metadata = leader and leader.metadata
    self._leader_resource_version = metadata.resource_version if metadata else None
    annotations = metadata and metadata.annotations or {}

    # get last known leader lsn
    last_lsn = annotations.get(self._OPTIME)
    try:
        last_lsn = 0 if last_lsn is None else int(last_lsn)
    except Exception:
        last_lsn = 0

    # get permanent slots state (confirmed_flush_lsn)
    slots = annotations.get('slots')
    try:
        slots = slots and json.loads(slots)
    except Exception:
        slots = None

    # get leader
    leader_record = {n: annotations.get(n) for n in (self._LEADER, 'acquireTime',
                     'ttl', 'renewTime', 'transitions') if n in annotations}
    if (leader_record or self._leader_observed_record) and leader_record != self._leader_observed_record:
        self._leader_observed_record = leader_record
        self._leader_observed_time = time.time()

    leader = leader_record.get(self._LEADER)
    try:
        ttl = int(leader_record.get('ttl')) or self._ttl
    except (TypeError, ValueError):
        ttl = self._ttl

    if not metadata or not self._leader_observed_time or self._leader_observed_time + ttl < time.time():
        leader = None

    if metadata:
        member = Member(-1, leader, None, {})
        member = ([m for m in members if m.name == leader] or [member])[0]
        leader = Leader(metadata.resource_version, None, member)

    # failover key
    failover = nodes.get(self.failover_path)
    metadata = failover and failover.metadata
    failover = Failover.from_node(metadata and metadata.resource_version,
                                  metadata and (metadata.annotations or {}).copy())

    # get synchronization state
    sync = nodes.get(self.sync_path)
    metadata = sync and sync.metadata
    sync = SyncState.from_node(metadata and metadata.resource_version,  metadata and metadata.annotations)

    return Cluster(initialize, config, leader, last_lsn, members, failover, sync, history, slots)
except Exception:
    logger.exception('get_cluster')
    raise KubernetesError('Kubernetes API is not responding properly')

上述集群信息中，主要以xxx-config、xxx-leader、xxx-failover、xxx-sync作为配置信息，具体内容如下：
●xxx-config

% kubectl get cm pg142-1013-postgresql-config -oyaml
apiVersion: v1
kind: ConfigMap
metadata:
  annotations:
    config: '{"loop_wait":10,"maximum_lag_on_failover":33554432,"postgresql":{"parameters":{"archive_command":"/bin/true","archive_mode":"on","archive_timeout":"1800s","autovacuum":"on","autovacuum_analyze_scale_factor":0.02,"autovacuum_max_workers":"3","autovacuum_naptime":"5min","autovacuum_vacuum_cost_delay":"2ms","autovacuum_vacuum_cost_limit":"-1","autovacuum_vacuum_scale_factor":0.05,"autovacuum_work_mem":"128MB","backend_flush_after":"0","bgwriter_delay":"200ms","bgwriter_flush_after":"256","bgwriter_lru_maxpages":"100","bgwriter_lru_multiplier":"2","checkpoint_completion_target":"0.9","checkpoint_flush_after":"256kB","checkpoint_timeout":"5min","commit_delay":"0","constraint_exclusion":"partition","datestyle":"iso,
      mdy","deadlock_timeout":"1s","default_text_search_config":"pg_catalog.english","dynamic_shared_memory_type":"posix","effective_cache_size":"32768","fsync":"on","full_page_writes":"on","hot_standby":"on","hot_standby_feedback":"off","huge_pages":"off","idle_in_transaction_session_timeout":"600000","lc_messages":"en_US.UTF-8","lc_monetary":"en_US.UTF-8","lc_numeric":"en_US.UTF-8","lc_time":"en_US.UTF-8","listen_addresses":"*","log_autovacuum_min_duration":"0","log_checkpoints":"on","log_connections":"off","log_disconnections":"off","log_error_verbosity":"default","log_line_prefix":"%t
      [%p]: [%l-1] %c %x %d %u %a %h","log_lock_waits":"on","log_min_duration_statement":"500","log_rotation_size":"0","log_statement":"none","log_temp_files":0,"log_timezone":"Asia/Shanghai","maintenance_work_mem":"32768","max_connections":"170","max_parallel_maintenance_workers":"2","max_parallel_workers":"2","max_parallel_workers_per_gather":"2","max_replication_slots":"10","max_standby_archive_delay":"30s","max_standby_streaming_delay":"30s","max_wal_senders":"10","max_wal_size":"2048","max_worker_processes":"8","old_snapshot_threshold":"-1","pg_stat_statements.max":"10000","pg_stat_statements.save":"on","pg_stat_statements.track":"all","pgaudit.log":"NONE","pgaudit.log_catalog":"on","pgaudit.log_client":"off","pgaudit.log_level":"log","pgaudit.log_parameter":"off","pgaudit.log_relation":"off","pgaudit.log_rows":"off","pgaudit.log_statement":"on","pgaudit.log_statement_once":"off","pgaudit.role":"","random_page_cost":"4","restart_after_crash":"on","synchronous_commit":"on","tcp_keepalives_count":"0","tcp_keepalives_idle":"900","tcp_keepalives_interval":"100","temp_buffers":"8MB","timezone":"Asia/Shanghai","track_activity_query_size":"1kB","track_functions":"all","track_io_timing":"off","unix_socket_directories":"/var/run/postgresql","vacuum_cost_delay":"0ms","vacuum_cost_limit":"200","wal_buffers":"2048","wal_compression":"on","wal_keep_segments":"128","wal_keep_size":"2048MB","wal_level":"replica","wal_log_hints":"on","wal_receiver_status_interval":"10s","wal_sender_timeout":"1min","wal_writer_delay":"200ms","wal_writer_flush_after":"1MB","work_mem":"4MB"},"use_pg_rewind":true,"use_slots":true},"retry_timeout":10,"synchronous_mode":true,"ttl":30}'
    initialize: "7289263672843878470"
  creationTimestamp: "2023-10-13T02:25:51Z"
  labels:
    application: spilo
    cluster-name: pg142-1013-postgresql
  name: pg142-1013-postgresql-config
  namespace: default
  resourceVersion: "22858249"
  uid: dfa64d28-e939-4bdd-8db1-a3485fa09637
上述例子中，下有和2个参数，
1.定义集群的整体配置信息，这里包含了PostgreSQL配置参数以及集群参数（选举等待时间、允许的最大WAL延迟量、是否开启同步模式等）等；
2.定义了集群的ID，该值对应pg_controldata命令内的值，因此，所有集群内的PostgreSQL节点有相同的sys_id。
root@pg142-1013-postgresql-1:/home/postgres# pg_controldata | grep "Database system identifier"
Database system identifier:           7289263672843878470
●xxx-leader
% kubectl get cm pg142-1013-postgresql-leader -oyaml
apiVersion: v1
kind: ConfigMap
metadata:
  annotations:
    acquireTime: "2023-10-13T02:26:06.973552+00:00"
    leader: pg142-1013-postgresql-0
    optime: "67109192"
    renewTime: "2023-10-16T07:02:57.418940+00:00"
    transitions: "0"
    ttl: "30"
  creationTimestamp: "2023-10-13T02:26:07Z"
  labels:
    application: spilo
    cluster-name: pg142-1013-postgresql
  name: pg142-1013-postgresql-leader
  namespace: default
  resourceVersion: "23286847"
  uid: cb235c85-6a21-454d-8320-222205eaa77f

上述下，各参数含义：
1.acquireTime：获取集群leader锁时间；
2.leader：集群leader锁的拥有者，这里表示某个PostgreSQL节点名称；
3.optime：集群leader的最新LSN的十进制数，这里;
4.renewTime：集群leader锁的拥有者心跳时间，心跳周期与xxx-config中的对应；
5.transitions：集群leader锁占用次数，一般发生在主从切换或故障转移场景，依次累加；
6.ttl：故障转移前的选举时间，即超过TTL时间下，没有获取到renewTime值更新，便触发选举，由新的节点占用leader锁。
●xxx-sync

% kubectl get cm pg142-1013-postgresql-sync -oyaml
apiVersion: v1
kind: ConfigMap
metadata:
  annotations:
    leader: pg142-1013-postgresql-1
    sync_standby: pg142-1013-postgresql-0
  creationTimestamp: "2023-10-16T06:54:39Z"
  labels:
    application: spilo
    cluster-name: pg142-1013-postgresql
  name: pg142-1013-postgresql-sync
  namespace: default
  resourceVersion: "23288352"
  uid: 1c46e63b-8b90-4fc6-9596-8e2f71fba2ab

上述内容记录了2个信息：
1.leader：显示leader节点的名称；
2.sync_standby：显示同步节点的名称，多个同步节点以逗号分隔。
●xxx-failover

% kubectl get cm pg142-1013-postgresql-failover -oyaml
apiVersion: v1
kind: ConfigMap
metadata:
  creationTimestamp: "2023-10-16T07:16:03Z"
  labels:
    application: spilo
    cluster-name: pg142-1013-postgresql
  managedFields:
  - apiVersion: v1
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:labels:
          .: {}
          f:application: {}
          f:cluster-name: {}
    manager: Patroni
    operation: Update
    time: "2023-10-16T07:36:56Z"
  name: pg142-1013-postgresql-failover
  namespace: default
  resourceVersion: "23290596"
  uid: 72d50c58-bc65-4b77-8870-93d0b8f8b7a2

上述内容，主要记录最后一次故障转移发生的时间。
集群状态检测

 if self.is_paused():
      self.watchdog.disable()
      self._was_paused = True
  else:
      if self._was_paused:
          self.state_handler.schedule_sanity_checks_after_pause()
      self._was_paused = False
  
  if not self.cluster.has_member(self.state_handler.name):
      self.touch_member()
  
  # cluster has leader key but not initialize key
  if not (self.cluster.is_unlocked() or self.sysid_valid(self.cluster.initialize)) and self.has_lock():
      self.dcs.initialize(create_new=(self.cluster.initialize is None), sysid=self.state_handler.sysid)
  
  if not (self.cluster.is_unlocked() or self.cluster.config and self.cluster.config.data) and self.has_lock():
      self.dcs.set_config_value(json.dumps(self.patroni.config.dynamic_configuration, separators=(',', ':')))
      self.cluster = self.dcs.get_cluster()
  
  if self._async_executor.busy:
      return self.handle_long_action_in_progress()
  
  msg = self.handle_starting_instance()
  if msg is not None:
      return msg
  
  # we've got here, so any async action has finished.
  if self.state_handler.bootstrapping:
      return self.post_bootstrap()
  
  if self.recovering:
      self.recovering = False
  
      if not self._rewind.is_needed:
          # Check if we tried to recover from postgres crash and failed
          msg = self.post_recover()
          if msg is not None:
              return msg
  
      # Reset some states after postgres successfully started up
      self._crash_recovery_executed = False
      if self._rewind.executed and not self._rewind.failed:
          self._rewind.reset_state()
  
      # The Raft cluster without a quorum takes a bit of time to stabilize.
      # Therefore we want to postpone the leader race if we just started up.
      if self.cluster.is_unlocked() and self.dcs.__class__.__name__ == 'Raft':
          return 'started as a secondary'

检测集群是否暂停

集群暂停，是指集群中的PostgreSQL节点不由Patroni管理，当集群异常时，不再出发故障转移等措施。
集群暂停一般由用户主动出发，可以用在单个PostgreSQL节点的维护上，触发方式：

root@pg142-1013-postgresql-0:/home/postgres# patronictl list
+ Cluster: pg142-1013-postgresql (7289263672843878470) ---+---------+----+-----------+
| Member                  | Host           | Role         | State   | TL | Lag in MB |
+-------------------------+----------------+--------------+---------+----+-----------+
| pg142-1013-postgresql-0 | 10.244.117.143 | Leader       | running |  3 |           |
| pg142-1013-postgresql-1 | 10.244.165.220 | Sync Standby | running |  3 |         0 |
+-------------------------+----------------+--------------+---------+----+-----------+
root@pg142-1013-postgresql-0:/home/postgres# patronictl pause
Success: cluster management is paused
root@pg142-1013-postgresql-0:/home/postgres# patronictl list
+ Cluster: pg142-1013-postgresql (7289263672843878470) ---+---------+----+-----------+
| Member                  | Host           | Role         | State   | TL | Lag in MB |
+-------------------------+----------------+--------------+---------+----+-----------+
| pg142-1013-postgresql-0 | 10.244.117.143 | Leader       | running |  3 |           |
| pg142-1013-postgresql-1 | 10.244.165.220 | Sync Standby | running |  3 |         0 |
+-------------------------+----------------+--------------+---------+----+-----------+
 Maintenance mode: on
上述，即表示当前集群已停止。此时，PostgreSQL进程仍然存活，如果故障，将需要用户自行启动。
集群暂停恢复方式：
root@pg142-1013-postgresql-0:/home/postgres# patronictl list
+ Cluster: pg142-1013-postgresql (7289263672843878470) ---+---------+----+-----------+
| Member                  | Host           | Role         | State   | TL | Lag in MB |
+-------------------------+----------------+--------------+---------+----+-----------+
| pg142-1013-postgresql-0 | 10.244.117.143 | Leader       | running |  3 |           |
| pg142-1013-postgresql-1 | 10.244.165.220 | Sync Standby | running |  3 |         0 |
+-------------------------+----------------+--------------+---------+----+-----------+
 Maintenance mode: on
root@pg142-1013-postgresql-0:/home/postgres# patronictl resume
Success: cluster management is resumed
root@pg142-1013-postgresql-0:/home/postgres# patronictl list
+ Cluster: pg142-1013-postgresql (7289263672843878470) ---+---------+----+-----------+
| Member                  | Host           | Role         | State   | TL | Lag in MB |
+-------------------------+----------------+--------------+---------+----+-----------+
| pg142-1013-postgresql-0 | 10.244.117.143 | Leader       | running |  3 |           |
| pg142-1013-postgresql-1 | 10.244.165.220 | Sync Standby | running |  3 |         0 |
+-------------------------+----------------+--------------+---------+----+-----------+

集群初始化检测

cluster has leader key but not initialize key通过命令，即可恢复集群。

在恢复集群后，需要对集群中PostgreSQL节点进行处理：
1.重新配置PostgreSQL的参数；
2.根据xxx-sync中最后一次记录的主、同步节点名称信息，在主节点上设置同步复制槽信息；
3.检测恢复后的PostgreSQL节点的是否变更，与最后一次xxx-config中的值，是否一致，否则将无法恢复集群。

if not (self.cluster.is_unlocked() or self.sysid_valid(self.cluster.initialize)) and self.has_lock():
    self.dcs.initialize(create_new=(self.cluster.initialize is None), sysid=self.state_handler.sysid)

if not (self.cluster.is_unlocked() or self.cluster.config and self.cluster.config.data) and self.has_lock():
    self.dcs.set_config_value(json.dumps(self.patroni.config.dynamic_configuration, separators=(',', ':')))
    self.cluster = self.dcs.get_cluster()

集群初始化检测，主要检测2个方面的信息：
●集群当前存在leader节点，但xxx-config中的不存在，此时，需要将leader节点上PostgreSQL的sysid设置到xxx-config中；
●集群当前存在leader节点，但未获取到xxx-config信息，需要将leader节点上的配置信息和sysid都设置到xxx-config中，并重新获取集群信息。
该步骤的用途是，防止xxx-config文件被删除，导致从节点加载集群信息失败。
节点状态检测

检测当前PostgreSQL的进程启动到了什么阶段

if self._async_executor.busy:
    return self.handle_long_action_in_progress()

检测启动中的PostgreSQL是否出现异常

msg = self.handle_starting_instance()
if msg is not None:
return msg
节点状态检测，是通过检测PostgreSQL节点的当前运行状态，来确定是否需要进行具体的操作，节点状态检测的方式可分为2种：
1.通过PostgreSQL的运行状态确定；
2.通过异步进程（_async_executor）监听，当前节点处于什么阶段。

节点检测通过后基础操作

# we've got here, so any async action has finished.
if self.state_handler.bootstrapping:
    return self.post_bootstrap()

if self.recovering:
    self.recovering = False

    if not self._rewind.is_needed:
        # Check if we tried to recover from postgres crash and failed
        msg = self.post_recover()
        if msg is not None:
            return msg

    # Reset some states after postgres successfully started up
    self._crash_recovery_executed = False
    if self._rewind.executed and not self._rewind.failed:
        self._rewind.reset_state()

    # The Raft cluster without a quorum takes a bit of time to stabilize.
    # Therefore we want to postpone the leader race if we just started up.
    if self.cluster.is_unlocked() and self.dcs.__class__.__name__ == 'Raft':
        return 'started as a secondary'

节点状态检测通过后，需要对PostgreSQL进行操作：

1.PostgreSQL启动后操作
def post_bootstrap(self):
    with self._async_response:
        result = self._async_response.result
    # bootstrap has failed if postgres is not running
    if not self.state_handler.is_running() or result is False:
        self.cancel_initialization()

if result is None:
    if not self.state_handler.is_leader():
        return 'waiting for end of recovery after bootstrap'

    self.state_handler.set_role('master')
    ret = self._async_executor.try_run_async('post_bootstrap', self.state_handler.bootstrap.post_bootstrap,
                                             args=(self.patroni.config['bootstrap'], self._async_response))
    return ret or 'running post_bootstrap'

self.state_handler.bootstrapping = False
if not self.watchdog.activate():
    logger.error('Cancelling bootstrap because watchdog activation failed')
    self.cancel_initialization()
self._rewind.ensure_checkpoint_after_promote(self.wakeup)
self.dcs.initialize(create_new=(self.cluster.initialize is None), sysid=self.state_handler.sysid)
self.dcs.set_config_value(json.dumps(self.patroni.config.dynamic_configuration, separators=(',', ':')))
self.dcs.take_leader()
self.set_is_leader(True)
self.state_handler.call_nowait(ACTION_ON_START)
self.load_cluster_from_dcs()

return 'initialized a new cluster'

上述操作，包括pg_rewind后的checkpoint检测、初始化DCS的xxx-config资源、生成xxx-leader资源、加载集群信息等。
2.恢复中的PostgreSQL检测是否需要执行pg_rewind
if self.recovering:
self.recovering = False

if not self._rewind.is_needed:
    # Check if we tried to recover from postgres crash and failed
    msg = self.post_recover()
    if msg is not None:
        return msg

# Reset some states after postgres successfully started up
self._crash_recovery_executed = False
if self._rewind.executed and not self._rewind.failed:
    self._rewind.reset_state()

pg_rewind命令用于将从节点的WAL与主节点的WAL拉齐，一般用于从节点WAL因异常后滞后于主节点WAL。
启动PostgreSQL

# is data directory empty?
if self.state_handler.data_directory_empty():
    self.state_handler.set_role('uninitialized')
    self.state_handler.stop('immediate', stop_timeout=self.patroni.config['retry_timeout'])
    # In case datadir went away while we were master.
    self.watchdog.disable()

    # is this instance the leader?
    if self.has_lock():
        self.release_leader_key_voluntarily()
        return 'released leader key voluntarily as data dir empty and currently leader'

    if self.is_paused():
        return 'running with empty data directory'
    return self.bootstrap()  # new node
else:
    # check if we are allowed to join
    data_sysid = self.state_handler.sysid
    if not self.sysid_valid(data_sysid):
        # data directory is not empty, but no valid sysid, cluster must be broken, suggest reinit
        return ("data dir for the cluster is not empty, "
                "but system ID is invalid; consider doing reinitialize")

    if self.sysid_valid(self.cluster.initialize):
        if self.cluster.initialize != data_sysid:
            if self.is_paused():
                logger.warning('system ID has changed while in paused mode. Patroni will exit when resuming'
                               ' unless system ID is reset: %s != %s', self.cluster.initialize, data_sysid)
                if self.has_lock():
                    self.release_leader_key_voluntarily()
                    return 'released leader key voluntarily due to the system ID mismatch'
            else:
                logger.fatal('system ID mismatch, node %s belongs to a different cluster: %s != %s',
                             self.state_handler.name, self.cluster.initialize, data_sysid)
                sys.exit(1)
    elif self.cluster.is_unlocked() and not self.is_paused():
        # "bootstrap", but data directory is not empty
        if not self.state_handler.cb_called and self.state_handler.is_running() \
                and not self.state_handler.is_leader():
            self._join_aborted = True
            logger.error('No initialize key in DCS and PostgreSQL is running as replica, aborting start')
            logger.error('Please first start Patroni on the node running as master')
            sys.exit(1)
        self.dcs.initialize(create_new=(self.cluster.initialize is None), sysid=data_sysid)

无数据目录启动

无数据目录启动，是指在执行初始化目录异常、恢复节点异常、WAL拉齐异常等场景下，会触发的流程：
1.设置角色，用于后续重新初始化集群；
2.立即停止当前PostgreSQL进程；
3.判断当前节点是否为主节点，主动释放主节点锁；
4.执行启动操作。

def bootstrap(self):
  if not self.cluster.is_unlocked():  # cluster already has leader
      clone_member = self.cluster.get_clone_member(self.state_handler.name)
      member_role = 'leader' if clone_member == self.cluster.leader else 'replica'
      msg = "from {0} '{1}'".format(member_role, clone_member.name)
      ret = self._async_executor.try_run_async('bootstrap {0}'.format(msg), self.clone, args=(clone_member, msg))
      return ret or 'trying to bootstrap {0}'.format(msg)
  
  # no initialize key and node is allowed to be master and has 'bootstrap' section in a configuration file
  elif self.cluster.initialize is None and not self.patroni.nofailover and 'bootstrap' in self.patroni.config:
      if self.dcs.initialize(create_new=True):  # race for initialization
          self.state_handler.bootstrapping = True
          with self._async_response:
              self._async_response.reset()
  
          if self.is_standby_cluster():
              ret = self._async_executor.try_run_async('bootstrap_standby_leader', self.bootstrap_standby_leader)
              return ret or 'trying to bootstrap a new standby leader'
          else:
              ret = self._async_executor.try_run_async('bootstrap', self.state_handler.bootstrap.bootstrap,
                                                       args=(self.patroni.config['bootstrap'],))
              return ret or 'trying to bootstrap a new cluster'
      else:
          return 'failed to acquire initialize lock'
  else:
      create_replica_methods = self.get_standby_cluster_config().get('create_replica_methods', []) \
                               if self.is_standby_cluster() else None
      if self.state_handler.can_create_replica_without_replication_connection(create_replica_methods):
          msg = 'bootstrap (without leader)'
          return self._async_executor.try_run_async(msg, self.clone) or 'trying to ' + msg
      return 'waiting for {0}leader to bootstrap'.format('standby_' if self.is_standby_cluster() else '')

上述代码，表示启动的几种方式：
1.当前集群已有leader节点，当前PostgreSQL将以从节点从主节点上同步数据启动；
2.当前集群没有leader节点，当前PostgreSQL将以主节点启动，如果是备用集群，将以备用集群主节点启动；
3.当前集群为备用集群且没有主节点，从节点通过方式，一般通过协议流方式从主集群上进行数据同步。
有数据目录启动
有数据目录启动，主要校验集群ID与PostgreSQL节点sysid的一致性，触发的主要流程：
1.校验PostgreSQL节点sysid是否有效，如果无效，表示PostgreSQL出现了异常需要重启；
2.校验校验集群ID与PostgreSQL节点sysid是否一致，不一致将无法加入集群，如果集群已暂停，将会释放leader锁占用；
3.检验集群没有leader节点，当前节点将重新初始化集群，将sysid作为新的集群ID启动。
生成PostgreSQL集群

try:
    if self.cluster.is_unlocked():
        ret = self.process_unhealthy_cluster()
    else:
        msg = self.process_healthy_cluster()
        ret = self.evaluate_scheduled_restart() or msg
finally:
    # we might not have a valid PostgreSQL connection here if another thread
    # stops PostgreSQL, therefore, we only reload replication slots if no
    # asynchronous processes are running (should be always the case for the master)
    if not self._async_executor.busy and not self.state_handler.is_starting():
        create_slots = self.state_handler.slots_handler.sync_replication_slots(self.cluster,
                                                                               self.patroni.nofailover)
        if not self.state_handler.cb_called:
            if not self.state_handler.is_leader():
                self._rewind.trigger_check_diverged_lsn()
            self.state_handler.call_nowait(ACTION_ON_START)
        if create_slots and self.cluster.leader:
            err = self._async_executor.try_run_async('copy_logical_slots',
                                                     self.state_handler.slots_handler.copy_logical_slots,
                                                     args=(self.cluster.leader, create_slots))
            if not err:
                ret = 'Copying logical slots {0} from the primary'.format(create_slots)

生成PostgreSQL集群，主要根据当前集群是否存在主节点，判断走健康的集群流程还是非健康的集群流程。
非健康的集群流程

def process_unhealthy_cluster(self):
  """Cluster has no leader key"""

  if self.is_healthiest_node():
      if self.acquire_lock():
          failover = self.cluster.failover
          if failover:
              if self.is_paused() and failover.leader and failover.candidate:
                  logger.info('Updating failover key after acquiring leader lock...')
                  self.dcs.manual_failover('', failover.candidate, failover.scheduled_at, failover.index)
              else:
                  logger.info('Cleaning up failover key after acquiring leader lock...')
                  self.dcs.manual_failover('', '')
          self.load_cluster_from_dcs()

          if self.is_standby_cluster():
              # standby leader disappeared, and this is the healthiest
              # replica, so it should become a new standby leader.
              # This implies we need to start following a remote master
              msg = 'promoted self to a standby leader by acquiring session lock'
              return self.enforce_follow_remote_master(msg)
          else:
              return self.enforce_master_role(
                  'acquired session lock as a leader',
                  'promoted self to leader by acquiring session lock'
              )
      else:
          return self.follow('demoted self after trying and failing to obtain lock',
                             'following new leader after trying and failing to obtain lock')
  else:
      # when we are doing manual failover there is no guaranty that new leader is ahead of any other node
      # node tagged as nofailover can be ahead of the new leader either, but it is always excluded from elections
      if bool(self.cluster.failover) or self.patroni.nofailover:
          self._rewind.trigger_check_diverged_lsn()
          time.sleep(2)  # Give a time to somebody to take the leader lock

      if self.patroni.nofailover:
          return self.follow('demoting self because I am not allowed to become master',
                             'following a different leader because I am not allowed to promote')
      return self.follow('demoting self because i am not the healthiest node',
                         'following a different leader because i am not the healthiest node')

非健康的集群流程，是确定leader节点的候选，首要条件必须找到一个健康的节点，如何判断健康的节点，主要有以下几个条件：
1.PostgreSQL集群状态非暂停；
2.PostgreSQL节点状态非启动中；
3.PostgreSQL节点允许故障转移；
4.PostgreSQL节点WAL与集群缓存中的（最后一次主节点同步的lsn值）的滞后量在允许的范围内。

def is_healthiest_node(self):
if time.time() - self._released_leader_key_timestamp < self.dcs.ttl:
logger.info(‘backoff: skip leader race after pre_promote script failure and releasing the lock voluntarily’)
return False
if self.is_paused() and not self.patroni.nofailover and
self.cluster.failover and not self.cluster.failover.scheduled_at:
ret = self.manual_failover_process_no_leader()
if ret is not None: # continue if we just deleted the stale failover key as a master
return ret

    if self.state_handler.is_starting():  # postgresql still starting up is unhealthy
        return False

    if self.state_handler.is_leader():
        # in pause leader is the healthiest only when no initialize or sysid matches with initialize!
        return not self.is_paused() or not self.cluster.initialize\
                or self.state_handler.sysid == self.cluster.initialize

    if self.is_paused():
        return False

    if self.patroni.nofailover:  # nofailover tag makes node always unhealthy
        return False

    if self.cluster.failover:
        # When doing a switchover in synchronous mode only synchronous nodes and former leader are allowed to race
        if self.is_synchronous_mode() and self.cluster.failover.leader and \
                self.cluster.failover.candidate and not self.cluster.sync.matches(self.state_handler.name):
            return False
        return self.manual_failover_process_no_leader()

    if not self.watchdog.is_healthy:
        logger.warning('Watchdog device is not usable')
        return False

    # When in sync mode, only last known master and sync standby are allowed to promote automatically.
    all_known_members = self.cluster.members + self.old_cluster.members
    if self.is_synchronous_mode() and self.cluster.sync and self.cluster.sync.leader:
        if not self.cluster.sync.matches(self.state_handler.name):
            return False
        # pick between synchronous candidates so we minimize unnecessary failovers/demotions
        members = {m.name: m for m in all_known_members if self.cluster.sync.matches(m.name)}
    else:
        # run usual health check
        members = {m.name: m for m in all_known_members}

    return self._is_healthiest_node(members.values())

…

 def _is_healthiest_node(self, members, check_replication_lag=True):
        """This method tries to determine whether I am healthy enough to became a new leader candidate or not."""

    my_wal_position = self.state_handler.last_operation()
    if check_replication_lag and self.is_lagging(my_wal_position):
        logger.info('My wal position exceeds maximum replication lag')
        return False  # Too far behind last reported wal position on master

    if not self.is_standby_cluster() and self.check_timeline():
        cluster_timeline = self.cluster.timeline
        my_timeline = self.state_handler.replica_cached_timeline(cluster_timeline)
        if my_timeline < cluster_timeline:
            logger.info('My timeline %s is behind last known cluster timeline %s', my_timeline, cluster_timeline)
            return False

    # Prepare list of nodes to run check against
    members = [m for m in members if m.name != self.state_handler.name and not m.nofailover and m.api_url]

    if members:
        for st in self.fetch_nodes_statuses(members):
            if st.failover_limitation() is None:
                if not st.in_recovery:
                    logger.warning('Master (%s) is still alive', st.member.name)
                    return False
                if my_wal_position < st.wal_position:
                    logger.info('Wal position of %s is ahead of my wal position', st.member.name)
                    # In synchronous mode the former leader might be still accessible and even be ahead of us.
                    # We should not disqualify himself from the leader race in such a situation.
                    if not self.is_synchronous_mode() or st.member.name != self.cluster.sync.leader:
                        return False
                    logger.info('Ignoring the former leader being ahead of us')
    return True

当前节点为健康节点，因当前集群没有主节点，需要执行leader锁抢占。如果当前节点抢占leader锁失败，将作为从节点加入到集群中。
当前节点为异常节点，则会一直等待PostgreSQL节点正常后，参与集群的选举行为。
健康的集群流程

def process_healthy_cluster(self):
  if self.has_lock():
      if self.is_paused() and not self.state_handler.is_leader():
          if self.cluster.failover and self.cluster.failover.candidate == self.state_handler.name:
              return 'waiting to become master after promote...'

      if not self.is_standby_cluster():
          self._delete_leader()
          return 'removed leader lock because postgres is not running as master'

  if self.update_lock(True):
      msg = self.process_manual_failover_from_leader()
      if msg is not None:
          return msg

      # check if the node is ready to be used by pg_rewind
      self._rewind.ensure_checkpoint_after_promote(self.wakeup)

      if self.is_standby_cluster():
          # in case of standby cluster we don't really need to
          # enforce anything, since the leader is not a master.
          # So just remind the role.
          msg = 'no action. I am ({0}), the standby leader with the lock'.format(self.state_handler.name) \
                if self.state_handler.role == 'standby_leader' else \
                'promoted self to a standby leader because i had the session lock'
          return self.enforce_follow_remote_master(msg)
      else:
          return self.enforce_master_role(
              'no action. I am ({0}), the leader with the lock'.format(self.state_handler.name),
              'promoted self to leader because I had the session lock'
          )
  else:
      # Either there is no connection to DCS or someone else acquired the lock
      logger.error('failed to update leader lock')
      if self.state_handler.is_leader():
          if self.is_paused():
              return 'continue to run as master after failing to update leader lock in DCS'
          self.demote('immediate-nolock')
          return 'demoted self because failed to update leader lock in DCS'
      else:
          return 'not promoting because failed to update leader lock in DCS'

 else:
      logger.debug('does not have lock')
  lock_owner = self.cluster.leader and self.cluster.leader.name
  if self.is_standby_cluster():
      return self.follow('cannot be a real primary in a standby cluster',
                         'no action. I am ({0}), a secondary, and following a standby leader ({1})'.format(
                              self.state_handler.name, lock_owner), refresh=False)
  return self.follow('demoting self because I do not have the lock and I was a leader',
                     'no action. I am ({0}), a secondary, and following a leader ({1})'.format(
                          self.state_handler.name, lock_owner), refresh=False)

健康的集群流程，是指当前的集群存在leader节点，对该流程的处理，主要有2个方向：
1.检测当前节点为主节点，进行更新leader锁操作，保持主节点心跳，避免从节点竞争锁，如果更新锁失败，将立即释放锁，让其他从节点抢占；
2.检测当前节点非主节点，作为从节点加入集群。

总结
综上所述，Patroni 是一个用于管理 PostgreSQL 数据库集群的高可用性（HA）管理工具，旨在确保数据库系统的连续可用性，以应对节点故障和维护操作等挑战。Patroni 提供了一系列关键功能和特点，使得它成为强大的高可用性解决方案。
总之，在很多场景中，Patroni能够保持PostgreSQL集群友好的运行，保证在集群异常的情况下，通过自动故障转移、数据同步和备份策略等功能，确保数据库集群的稳定性和可用性，使得应用程序能够持续访问数据，即使在节点故障或维护时也不会中断服务。

参考资源
Patroni配置参数https://patroni.readthedocs.io/en/latest/patroni_configuration.html
Patroni基于2.1.5分支源码https://github.com/zalando/patroni/tree/v2.1.5