修复开启备机逻辑复制，执行倒换时如果同步备先退，主机的逻辑复制walsender不会退出的问题 · Pull Request !7510 · openGauss/openGauss-server

1. 修复开启备机逻辑复制，执行倒换时如果同步备先退，主机的逻辑复制walsender不会退出的问题
---

【标题】(请简要描述下实现的内容)
修复开启备机逻辑复制，执行倒换时如果同步备先退，主机的逻辑复制walsender不会退出的问题
【实现内容】:
修复开启备机逻辑复制，执行倒换时如果同步备先退，主机的逻辑复制walsender不会退出的问题
【根因分析】:

1. 一主4备，主机有2个同步备`synchronous_standby_names = 'ANY 1(dn_6002, dn_6003),ANY 1(dn_6004, dn_6005)'`，备机上有逻辑复制，且开启了enable_slot_log，所以主机一共有10个逻辑复制的walsender(备机逻辑复制的)，5个物理复制的walsender（4个备机和1个灾备集群的）
2. 启动倒换时，12个walsender退出，剩余一个发起倒换的备机和2个逻辑复制的没退，且2个逻辑复制的刚好也属于发起倒换的备机的
3. 通过分析，发现退掉的备机最后一次的位点是这个145B/9E477EE8， 但是发起倒换的6001最后的位点是这个145B/9E477FC0， 说明主机最后的xlog flush到了145B/9E477FC0
4. 而主机的配置中要求至少有2个同步备，说明此刻有一个同步备没了，导致145B/9E477FC0这个lsn无法同步，导致逻辑复制的walsender卡在log_slot_advance->SyncRepWaitForLSN中

【实现方案】:

在syncrepwaitforlsn的循环中，判断如果当前线程是逻辑walsender且发生switchover且开启enable_slot_log，就终止等待

【关联需求或issue】:
主备切换过程中主机降备时候逻辑复制相关的walsender线程没退出导致切换失败【YC】
https://e.gitee.com/opengaussorg/issues/table?issue=IBY8ZR
【开发自验报告】:
1. 请附上自验结果(内容或者截图)

复现：通过修改代码，让物理walsender先退，再等逻辑复制发一次消息，可以复现
```
(gdb) t 11
[Switching to thread 11 (Thread 0xfffc889ce0e0 (LWP 867367))]
#0  0x0000fffe197fb50c in __GI___poll (fds=0xfffc889ccaf8, nfds=187650768285816, timeout=<optimized out>)
    at ../sysdeps/unix/sysv/linux/poll.c:41
41        return SYSCALL_CANCEL (ppoll, fds, nfds, timeout_ts_p, NULL, 0);
(gdb) bt
#0  0x0000fffe197fb50c in __GI___poll (fds=0xfffc889ccaf8, nfds=187650768285816, timeout=<optimized out>)
    at ../sysdeps/unix/sysv/linux/poll.c:41
#1  0x0000aaaad962b404 in WaitLatchOrSocket (latch=0xfffd2c068f90, wakeEvents=25, sock=-1, timeout=3000)
    at /usr1/cctest/openGauss-server_gitee/src/common/backend/port/pg_latch.cpp:309
#2  0x0000aaaad962b078 in WaitLatch (latch=0xfffd2c068f90, wakeEvents=25, timeout=3000)
    at /usr1/cctest/openGauss-server_gitee/src/common/backend/port/pg_latch.cpp:192
#3  0x0000aaaada851d20 in SyncRepWaitForLSN (XactCommitLSN=4277313296, enableHandleCancel=true)
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/syncrep.cpp:449
#4  0x0000aaaada848024 in log_slot_advance (slotInfo=0xfffe01a00e38, extra_content=0x0)
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/slotfuncs.cpp:115
#5  0x0000aaaada87850c in AdvanceLogicalReplication (cmd=0xfffc86404b48)
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/walsender.cpp:1612
#6  0x0000aaaada87a304 in IdentifyCommand (cmd_node=0xfffc86404b48, repCxt=0xfffc889cce58,
    cmd_string=0xfffc86abe060 "ADVANCE_REPLICATION SLOT \"slot1\" LOGICAL 0/F4000058 0/FE000218")
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/walsender.cpp:2215
#7  0x0000aaaada87a664 in HandleWalReplicationCommand (
    cmd_string=0xfffc86abe060 "ADVANCE_REPLICATION SLOT \"slot1\" LOGICAL 0/F4000058 0/FE000218", repCxt=0xfffc889cce58)
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/walsender.cpp:2284
#8  0x0000aaaada874de4 in WalSndHandshake ()
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/walsender.cpp:564
#9  0x0000aaaada874398 in WalSenderMain ()
    at /usr1/cctest/openGauss-server_gitee/src/gausskernel/storage/replication/walsender.cpp:408

gs_ctl switchover -D /xx/xxx/data/slave/
[2025-04-03 10:15:05.004][1028232][][gs_ctl]: gs_ctl switchover ,datadir is /xx/xxx/data/slave
[2025-04-03 10:15:05.004][1028232][][gs_ctl]: switchover term (1)
[2025-04-03 10:15:05.024][1028232][][gs_ctl]: waiting for server to switchover...............................................................
[2025-04-03 10:16:05.975][1028232][][gs_ctl]:
 switchover timeout after 60 seconds. please manually check the cluster status.

```

修复后，在相同的条件下，可以正常倒换：

```
gs_ctl switchover -D /xx/xxx/data/slave/
[2025-04-07 16:09:15.182][1030345][][gs_ctl]: gs_ctl switchover ,datadir is /xx/xxx/data/slave
[2025-04-07 16:09:15.182][1030345][][gs_ctl]: switchover term (1)
[2025-04-07 16:09:15.201][1030345][][gs_ctl]: waiting for server to switchover.............
[2025-04-07 16:09:25.390][1030345][][gs_ctl]: done
[2025-04-07 16:09:25.390][1030345][][gs_ctl]: switchover completed (/xx/xxx/data/slave)

2025-04-07 16:09:19.216 cctest [unknown] 20.20.20.107 281466813341920 0[0:0#0]  0 [BACKEND] WARNING:  canceling wait for synchronous replication due to switchover request for logical replication.

```
2. 是否可以添加fastcheck测试用例，如是，请补充fastcheck用例
3. 是否涉及资料修改，如是，在docs仓库补充资料
4. 是否考虑升级场景(系统表修改、日志持久化以及修改执行态数据格式)
5. 是否考虑在线扩容等扩展场景
5. 是否考虑异常场景/并发场景/前向兼容/性能场景
6. 是否对其他模块产生影响

【其他说明】:

openGauss/openGauss-server .gitee-modal { width: 500px !important; }

搜索帮助

openGauss/openGauss-server