写关注 w:1、majority、j:true 的耗时对比

  成员状态:9901=PRIMARY, 9902=SECONDARY, 9903=SECONDARY
  # 集群的默认写关注(没有显式设置时)
  test> db.adminCommand({getDefaultRWConcern: 1}).defaultWriteConcern
  { w: 'majority', wtimeout: 0 }
  test> var h = db.hello(); ({setName: h.setName, isWritablePrimary: h.isWritablePrimary, members: h.hosts.length, majority: Math.floor(h.hosts.length / 2) + 1})
  { setName: 'rsm46', isWritablePrimary: true, members: 3, majority: 2 }

  写关注                              中位数       p99
  w:0                           0.0 ms    0.2 ms
  w:1                           0.2 ms    0.7 ms
  w:1, j:true                   4.0 ms    9.0 ms
  w:2                           0.6 ms    1.5 ms
  w:majority                    4.0 ms    8.1 ms
  w:3                           0.6 ms    1.3 ms
  w:majority, j:true            4.0 ms    8.4 ms

  w:majority(只剩 2 个节点在线):成功,耗时 23.5 ms
  w:3 wtimeout=1000:1016 ms 后返回 WriteConcernError code=64: waiting for replication timed out
  但 w:3 那条文档已经写进了 primary:count({k: "w3"}) = 1

  只剩 1 个节点在线:isWritablePrimary = False,secondary = True
  向它写入:NotPrimaryError: not primary, full error: {'topologyVersion': {'processId': ObjectId('6acc4c1e275f45487358a

  w:0 重复插入 _id:1:没有抛异常(服务端没把错误告诉客户端)
  w:1 重复插入 _id:1:DuplicateKeyError
  t5 里实际文档数 1

  连接串 w=2:client.write_concern = {'w': 2}
  集合上 with_options(write_concern=WriteConcern(w=1)):{'w': 1}
  # mongosh:单次操作指定
  test> db.t.insertOne({x: 1}, {writeConcern: {w: "majority", wtimeout: 5000}}).acknowledged
  true