Copyset is not automatically migrated after stopping metaserver

[YunhuiChen]

Describe the bug (描述bug)

Copyset is not automatically migrated after stopping metaserver
To Reproduce (复现方法)
1、stop metaserver
2、wait 4 hours
3、

Expected behavior (期望行为)

Versions (各种版本)
OS:
Compiler:
branch:
commit id:

Additional context/screenshots (更多上下文/截图)

[cw123]

I 2022-04-29T12:02:54.543547+0800 80 recoverScheduler.cpp:34] recoverScheduler begin.
I 2022-04-29T12:02:54.543730+0800 80 operatorControllerTemplate.h:141] add operator [startEpoch: 3, copysetID: (1,5), priority: 2, step: change peer from 5 to 6] fail because of oncurrency exceed
W 2022-04-29T12:02:54.543761+0800 80 recoverScheduler.cpp:95] recover scheduler add operator [startEpoch: 3, copysetID: (1,5), priority: 2, step: change peer from 5 to 6] on [copysetId:(1,5), epoch:3, leader:2, peers:(1,2,5,), canidate:0, has configChangeInfo:0] fail
I 2022-04-29T12:02:54.543777+0800 80 operatorControllerTemplate.h:141] add operator [startEpoch: 3, copysetID: (1,6), priority: 2, step: change peer from 5 to 6] fail because of oncurrency exceed
W 2022-04-29T12:02:54.543783+0800 80 recoverScheduler.cpp:95] recover scheduler add operator [startEpoch: 3, copysetID: (1,6), priority: 2, step: change peer from 5 to 6] on [copysetId:(1,6), epoch:3, leader:4, peers:(2,4,5,), canidate:0, has configChangeInfo:0] fail

[cw123]

the recover generate operator, but the operator exceed the concurrency
the concurrent is set 1 at mds.conf.
mds.schduler.operator.concurrent=1

[cw123]

[cw123]

过了6天，copyset配置变更还是卡在这里

[cw123]

mds日志里面，心跳有报这个错误
W 2022-05-05T17:43:33.544570+0800 106 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 1
W 2022-05-05T17:43:37.469429+0800 100 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 4
W 2022-05-05T17:43:42.525686+0800 90 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 2
W 2022-05-05T17:43:43.544281+0800 98 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 1
W 2022-05-05T17:43:47.469271+0800 94 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 4
W 2022-05-05T17:43:52.526186+0800 25 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 2
W 2022-05-05T17:43:53.545147+0800 107 heartbeat_manager.cpp:233] leader not found, poolid: 1, copysetid: 1

[cw123]

metaserver有很多这个日志

[wu-hanqing]

this problem is introduced by #1211, and the reason is when dispatching configuration change to metaserver, MDS will ensure that the target copyset is not creating state.

but #1211 doesn't reset creating state, so the configuration change doesn't dispatch.

[cw123]

this problem is introduced by #1211, and the reason is when dispatching configuration change to metaserver, MDS will ensure that the target copyset is not creating state.

but #1211 doesn't reset creating state, so the configuration change doesn't dispatch.

will fix it in #1417

[cw123]

fix it, merge to master #1417