clickhouse（十二、踩坑之路）

最新推荐文章于 2025-11-04 10:50:57 发布

原创最新推荐文章于 2025-11-04 10:50:57 发布 · 2.3w 阅读

40 ·

CC 4.0 BY-SA版权

文章标签：

#clickhouse #异常

大数据同时被 2 个专栏收录

73 篇文章

订阅专栏

clickhouse

16 篇文章

订阅专栏

本文汇总了ClickHouse数据库在运行过程中遇到的多种常见错误及其详细的解决方案，包括元数据与数据不一致、分布式DDL语句超时、字符串转数字异常等问题，提供了实用的修复策略和技术建议。

Q1
DB::Exception: Cannot create table from metadata file /data/clickhouse/metadata/default/dwd_test.sql, error: DB::Exception: The local set of parts of table default.dwd_test doesn’t look like the set of parts in ZooKeeper: 65.88 million rows of 85.04 million total rows in filesystem are suspicious. There are 545 unexpected parts with 65883643 rows (191 of them is not just-written with 65883643 rows), 0 missing parts (with 0 blocks).
A1
这是由于truncate、alter等ddl语句用了on cluster，偶尔导致zookeeper同步异常。解决方法1：删除有问题节点的本地表数据 rm -r /data/clickhouse/data/default/dwd_test，再重新启动CK，副本会自动重新同步该表数据。(如果没有副本请不要使用此方法。)
解决方法2：命令行执行sudo -u clickhouse touch /data/clickhouse/flags/force_restore_data 然后手动恢复有问题的partition
Q2
Connected to ClickHouse server version 20.3.8 revision 54433.
Poco::Exception. Code: 1000, e.code() = 13, e.displayText() = Access to file denied: /home/qspace/.clickhouse-client-history (version 20.3.8.53 (official build))
A2

创建该文件、并设置开放权限。
chown clickhouse:clickhouse /home/qspace/.clickhouse-client-history (没有则创建)
Q3
Application: DB::Exception: Listen [::]:8124 failed: Poco::Exception. Code: 1000, e.code() = 0, e.displayText() = DNS error: EAI: Address family for hostname not supported (version 20.5.2.7 (official build))
A3
本机没有开放ipv6，只能对ipv4生效。在/etc/click-house/config.xml中，把<listen_host> 改成0.0.0.0 或者 ::
Q4
Code: 32, e.displayText() = DB::Exception: Received from hadoop8:9900. DB::Exception: Attempt to read after eof: Cannot parse Int32 from String, because value is too short. (version 20.3.8.53 (official build))
字符串转数字异常，有些为空或者非数字的字符导致转换失败
A4
使用toUInt64OrZero函数、转换失败则为0。
Q5
Application: DB::Exception: Suspiciously many (125) broken parts to remove.: Cannot attach table default.test
Code: 231. DB::Exception: Received from ck10:9000. DB::Exception: Suspiciously many (125) broken parts to remove…
A5
因为写入数据造成的元数据和数据不一致问题。先删除磁盘数据、再重启节点删除本地表，如果是复制表再上zookeeper删除副本，然后重新建表。复制表的话数据会同步其他副本的。
Q6
Cannot execute replicated DDL query on leader.
A6
由于分布式ddl语句比较耗时会超时响应、改为本地执行或者减少作用的数据范围。如ALTER、OPTIMIZE全表改为具体的partition.
Q7
Code: 76. DB::Exception: Received from 0.0.0.0:9900. DB::Exception: Cannot open file /data/clickhouse/data/default/test/tmp_insert_20200523_55575_55575_0/f0560_deep_conversion_optimization_goal.mrk2, errno: 24, strerror: Too many open files.
A7
修改/etc/security/limits.conf 添加：
clickhouse soft nofile 262144
clickhouse hard nofile 262144
使用ulimit -n 查询、看到的是所有用户可打开的总数，而ck能打开的大小只是系统的默认值，所以不要被这个命令干扰，重启ck后、获取ck的进程、再通过cat /proc/${pid}/limits |grep open ，判断配置是否生效
Q8
DB::Exception: Too many parts(100).
A8
官方回答：The following changes have been made on config.xml:

<merge_tree>
    <parts_to_delay_insert>300</parts_to_delay_insert>
    <parts_to_throw_insert>600</parts_to_throw_insert>
    <max_delay_to_insert>2</max_delay_to_insert>
    <max_suspicious_broken_parts>5</max_suspicious_broken_parts>
</merge_tree>

Q9
Code: 117. DB::Exception: Received from hadoop8:9900. DB::Exception: Data directory for table already containing data parts - probably it was unclean DROP table or manual intervention. You must either clear directory by hand or use ATTACH TABLE instead of CREATE TABLE if you need to use that parts…
A9
建表的时候发现之前表存储数据还在，如果需要之前的数据那么把create改为attach语句，如果需要则要清理磁盘上对应表空间的目录、如果zookeeper上也有相关信息可能也需要一起清理。
Q10
Received exception from server (version 20.3.8):
Code: 999. DB::Exception: Received from hadoop7:9900. DB::Exception: Session expired (Session expired).
A10
查看日志应该是zookeeper的状态问题。