postgresql - 加入 Postgres BDR 节点使用过时的 DSN

Question

我有一个 Postgres BDR 集群，它有 3 个节点“Ready”和 3 个节点“Parted”。

如果我执行 SELECT * FROM bdr.bdr_nodes，将显示以下信息:

-[ RECORD 1 ]------+-------------------------
node_sysid         | 6153716379158074503
node_timeline      | 1
node_dboid         | 16385
node_status        | r
node_name          | node3
node_local_dsn     | host=x.x.x.241 [...]
node_init_from_dsn | host=x.x.x.47 [...]
-[ RECORD 2 ]------+-------------------------
node_sysid         | 6153716914784688297
node_timeline      | 1
node_dboid         | 16385
node_status        | r
node_name          | node2
node_local_dsn     | host=x.x.x.5 [...]
node_init_from_dsn | host=x.x.x.47 [...]
-[ RECORD 3 ]------+-------------------------
node_sysid         | 6170758438846557459
node_timeline      | 1
node_dboid         | 16384
node_status        | r
node_name          | node4
node_local_dsn     | host=x.x.x.128 [...]
node_init_from_dsn | host=x.x.x.47 [...]
-[ RECORD 4 ]------+-------------------------
node_sysid         | 6153716402564903569
node_timeline      | 1
node_dboid         | 16385
node_status        | k
node_name          | node1
node_local_dsn     | host=x.x.x.47 [...]
node_init_from_dsn | 
-[ RECORD 5 ]------+-------------------------
node_sysid         | 6170830020100809103
node_timeline      | 1
node_dboid         | 16385
node_status        | k
node_name          | node6
node_local_dsn     | host=x.x.x.48 [...]
node_init_from_dsn | host=x.x.x.241 [...]
-[ RECORD 6 ]------+-------------------------
node_sysid         | 6170839982079996801
node_timeline      | 1
node_dboid         | 16385
node_status        | c
node_name          | node8
node_local_dsn     | host=x.x.x.142 [...]
node_init_from_dsn | host=x.x.x.241 [...]
-[ RECORD 7 ]------+-------------------------
node_sysid         | 6170833985333433816
node_timeline      | 1
node_dboid         | 16385
node_status        | k
node_name          | node7
node_local_dsn     | host=x.x.x.48 [...]
node_init_from_dsn | host=x.x.x.241 [...]

我正在尝试加入 node8。但这不会发生。错误如下:

d= p=5521 a=ERROR:  08006: could not connect to the primary server: could not connect to server: Connection timed out
        Is the server running on host "x.x.x.48" and accepting
        TCP/IP connections on port 5432?
d= p=5521 a=DETAIL:  Connection string is 'host=x.x.x.48 [...]'

该错误意味着它正在尝试连接到已被杀死或删除的节点。为什么要尝试连接到被杀死或删除的节点？我该如何解决这种情况？

以下命令用于加入node8

SELECT bdr.bdr_group_join(
      local_node_name := 'node8',
      node_external_dsn := 'host=x.x.x.142 [...]',
      join_using_dsn := 'host=x.x.x.241 [...]'
);

BDR 已根据 this instructions 安装(Debian Wheezy):

curl -sSL https://manageacloud.com/api/cm/configuration/postgresql-bdr/debian/manageacloud-production-script.sh | bash

表bdr.bdr_connections:

-[ RECORD 1 ]----------+---------------------
conn_sysid             | 6170839982079996801
conn_timeline          | 1
conn_dboid             | 16385
conn_origin_sysid      | 0
conn_origin_timeline   | 0
conn_origin_dboid      | 0
conn_is_unidirectional | f
conn_dsn               | host=x.x.x.142 [...]
conn_apply_delay       | 
conn_replication_sets  | {default}
-[ RECORD 2 ]----------+----------------------
conn_sysid             | 6153716402564903569
conn_timeline          | 1
conn_dboid             | 16385
conn_origin_sysid      | 0
conn_origin_timeline   | 0
conn_origin_dboid      | 0
conn_is_unidirectional | f
conn_dsn               | host=x.x.x.47 [...]
conn_apply_delay       | 
conn_replication_sets  | {default}
-[ RECORD 3 ]----------+-----------------------
conn_sysid             | 6153716379158074503
conn_timeline          | 1
conn_dboid             | 16385
conn_origin_sysid      | 0
conn_origin_timeline   | 0
conn_origin_dboid      | 0
conn_is_unidirectional | f
conn_dsn               | host=x.x.x.241 [...]
conn_apply_delay       | 
conn_replication_sets  | {default}
-[ RECORD 4 ]----------+-----------------------
conn_sysid             | 6153716914784688297
conn_timeline          | 1
conn_dboid             | 16385
conn_origin_sysid      | 0
conn_origin_timeline   | 0
conn_origin_dboid      | 0
conn_is_unidirectional | f
conn_dsn               | host=x.x.x.5 [...]
conn_apply_delay       | 
conn_replication_sets  | {default}
-[ RECORD 5 ]----------+-----------------------
conn_sysid             | 6170758438846557459
conn_timeline          | 1
conn_dboid             | 16384
conn_origin_sysid      | 0
conn_origin_timeline   | 0
conn_origin_dboid      | 0
conn_is_unidirectional | f
conn_dsn               | host=x.x.x.128 [...]
conn_apply_delay       | 
conn_replication_sets  | {default}

版本:

# SELECT bdr.bdr_version();
    bdr_version    
-------------------
 0.9.1-2015-05-26-
(1 row)

Answer 1

这是 BDR 中的错​​误。我刚刚在我的 bdr-plugin/next 树的本地副本中修复了它，并将更改推送到 bdr-plugin/REL0_9_STABLE 以包含在 0.9.3 中一次我已经在本地测试过了。

问题是在作为节点加入的一部分在对等节点上创建插槽期间，我们没有根据 bdr.bdr_nodes.state 过滤掉 bdr.bdr_connections 行.

删除没有相应bdr.bdr_nodes 条目的任何bdr.bdr_connections 条目，或者bdr.bdr_nodes 条目有state = 'k' 在 0.9.2 及更早版本中解决此问题。