由于docker目录迁移，导致容器启动的时候，然后不断restarting，服务不能正常运行。

root@db1:~# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
9f795ad40afc redis:6.0.9 “docker-entrypoint.s…” 3 days ago Up 41 minutes redis-6378
995d1c983134 redis:6.0.9 “docker-entrypoint.s…” 3 days ago Up 41 minutes redis-6379
60ff8ad4efe8 clickhouse/clickhouse-server:23.6.1.1524 “/entrypoint.sh” 4 weeks ago Restarting (232) 14 seconds ago clickhouse-01
f352dacd9f1d clickhouse/clickhouse-server:23.6.1.1524 “/entrypoint.sh” 4 weeks ago Restarting (232) 12 seconds ago clickhouse-02
b4a02ed016f1 clickhouse/clickhouse-keeper:23.6.1.1524 “/entrypoint.sh” 4 weeks ago Restarting (1) 6 seconds ago clickhouse-keeper-01
06f77e12e6bd mysql:5.7 “docker-entrypoint.s…” 5 weeks ago Restarting (1) 19 seconds ago mysql-master1
root@db1:~#

查看docker logs报错日志，发现是目录权限问题，导致命令报错直接退出。

root@db1:~# docker logs -f –tail 100 clickhouse-keeper-01
Logging information to /var/log/clickhouse-keeper/clickhouse-keeper.log
Logging errors to /var/log/clickhouse-keeper/clickhouse-keeper.err.log
Processing configuration file ‘/etc/clickhouse-keeper/keeper_config.xml’.
Logging information to /var/log/clickhouse-keeper/clickhouse-keeper.log
Logging errors to /var/log/clickhouse-keeper/clickhouse-keeper.err.log
Necessary directory ‘/var/lib/clickhouse’ isn’t accessible by user with id ‘101’
Necessary directory ‘/var/lib/clickhouse’ isn’t accessible by user with id ‘101’
…
正常情况下，通过宿主机执行docker exec -it 容器ID sh 是可以进去容器，然后就可以进行尝试修复。

但是容器现在处于不可运行状态，根本无法进去：

root@db1:~# docker exec -it -uroot clickhouse-keeper-01 bash
Error response from daemon: Container b4a02ed016f178eea7339f5c986c774dfbe766368b3d42e90338bb338f8a6869 is restarting, wait until the container is running
root@db1:~#

删除容器，再来run一个？但那是新的了，数据会丢失。

那容器里的数据复制docker cp复制出来，然后再run一个？但不知道容器内哪些数据是需要cp的，总不能都复制出来吧，再说组合一个新的“原装货”也很麻烦。

灵机一动，修改一下入口文件entrypoint.sh，再最前面加上sleep 3600，这样好歹可以先”卡住”容器，至少这样它是不会错误退出了。

遂docker cp 容器ID:/entrypoint.sh ./

vim entrypoint.sh，

#!/bin/bash
sleep 3600
set +x
set -eo pipefail
shopt -s nullglob

…

docker cp entrypoint.sh 容器ID:/

然后docker restart 容器ID

root@db1:~# docker exec -it -uroot clickhouse-keeper-01 bash
clickhouse-keeper-01:/#

于是可以正常进去了，为了方便，指定用root用户登录，否则权限还是无法修改。

实际上，用root执行一下entrypoint.sh，就可以把权限按照脚本期望的“刷”回来。（当然，也可以进行其他操作）

这里需要注意一下，因为我们已经修改了入口文件，所以需要来个副本，并去掉sleep时间，才能立即执行。

cp entrypoint.sh repair-entrypoint.sh

vi repair-entrypoint.sh

#!/bin/bash
# sleep 3600
set +x
set -eo pipefail
shopt -s nullglob

…

然后执行./repair-entrypoint.sh：

my id: 1, leader: 3, term: 1020
2024.12.10 06:44:37.452752 [ 58 ] {} <Information> RaftInstance: config at index 2122055 is committed, prev config log idx 2121879
2024.12.10 06:44:37.452804 [ 58 ] {} <Information> RaftInstance: new config log idx 2122055, prev log idx 2121879, cur config log idx 2121879, prev log idx 2121643
2024.12.10 06:44:37.452845 [ 58 ] {} <Information> RaftInstance: new configuration: log idx 2122055, prev log idx 2121879
peer 1, DC ID 0, clickhouse-keeper-01:9234, voting member, 1
peer 2, DC ID 0, clickhouse-keeper-02:9234, voting member, 1
peer 3, DC ID 0, clickhouse-keeper-03:9234, voting member, 1
my id: 1, leader: 3, term: 1020

没问题了。

切换为默认身份 su clickhouse

重新执行./repair-entrypoint.sh

发现正常了。

最后用root用户repair-entrypoint.sh 覆盖entrypoint.sh，退出容器，重启容器，OK。

观察日志，没问题，集群正常了~