诊断工具
更新时间：2025-08-21
概述
高效好用的性能诊断工具对于数据库系统的调优至关重要，因为这取决于是否能快速定位到存在性能问题的业务 SQL，继而快速定位和解决性能瓶颈，保证数据库系统服务的 SLA。
当前，PALO 系统默认将执行时间超过 5 秒的 SQL 认定为慢 SQL，此阈值可通过 config.qe_slow_log_ms 进行配置。目前 PALO 提供了以下三种诊断渠道，能够帮助快速定位存在性能问题的慢 SQL，分别如下：
PALO Manager 日志
PALO Manager 的日志模块提供了慢 SQL 筛选功能。用户可以通过选择特定 FE 节点上的 fe.audit.log 来查看慢 SQL。只需在搜索框中输入“slow_query”，即可在页面上展示当前系统的历史慢 SQL 信息。
Audit Log
当前 PALO FE 提供了四种类型的 Audit Log，包括 slow_query、query、load 和 stream_load。Audit Log 除了在安装部署 Manager 服务的集群上通过日志页面访问获取之外，也可以直接访问 FE 所在节点的 fe/log/fe.audit.log 文件获取信息。
通过直查 fe.audit.log 中的 slow_query 标签，可以快速筛选出执行缓慢的查询 SQL，如下所示：
SQL
12024-07-18 11:23:13,042 [slow_query] |Client=127.0.0.1:63510|User=root|Ctl=internal|Db=tpch_sf1000|State=EOF|ErrorCode=0|ErrorMessage=|Time(ms)=11603|ScanBytes=236667379712|ScanRows=13649979418|ReturnRows=100|StmtId=1689|QueryId=91ff336304f14182-9ca537eee75b3856|IsQuery=true|isNereids=true|feIp=172.21.0.10|Stmt=select c_name, c_custkey, o_orderkey, o_orderdate, o_totalprice, sum(l_quantity) from customer, orders, lineitem where o_orderkey in ( select l_orderkey from lineitem group by l_orderkey having sum(l_quantity) > 300 ) and c_custkey = o_custkey and o_orderkey = l_orderkey group by c_name, c_custkey, o_orderkey, o_orderdate, o_totalprice order by o_totalprice desc, o_orderdate limit 100|CpuTimeMS=918556|ShuffleSendBytes=3267419|ShuffleSendRows=89668|SqlHash=b4e1de9f251214a30188180f37907f7d|peakMemoryBytes=38720935552|SqlDigest=d41d8cd98f00b204e9800998ecf8427e|cloudClusterName=UNKNOWN|TraceId=|WorkloadGroup=normal|FuzzyVariables=|scanBytesFromLocalStorage=0|scanBytesFromRemoteStorage=0
22024-07-18 11:23:33,043 [slow_query] |Client=127.0.0.1:26672|User=root|Ctl=internal|Db=tpch_sf1000|State=EOF|ErrorCode=0|ErrorMessage=|Time(ms)=8978|ScanBytes=334985555968|ScanRows=10717654374|ReturnRows=100|StmtId=1815|QueryId=6e1fae453cb04d9a-b1e5f94d9cea1885|IsQuery=true|isNereids=true|feIp=172.21.0.10|Stmt=select s_name, count(*) as numwait from supplier, lineitem l1, orders, nation where s_suppkey = l1.l_suppkey and o_orderkey = l1.l_orderkey and o_orderstatus = 'F' and l1.l_receiptdate > l1.l_commitdate and exists ( select * from lineitem l2 where l2.l_orderkey = l1.l_orderkey and l2.l_suppkey <> l1.l_suppkey ) and not exists ( select * from lineitem l3 where l3.l_orderkey = l1.l_orderkey and l3.l_suppkey <> l1.l_suppkey and l3.l_receiptdate > l3.l_commitdate ) and s_nationkey = n_nationkey and n_name = 'SAUDI ARABIA' group by s_name order by numwait desc, s_name limit 100|CpuTimeMS=990127|ShuffleSendBytes=59208164|ShuffleSendRows=3651504|SqlHash=f8a30e4182d72cce3eff6cb385005b1f|peakMemoryBytes=10495660672|SqlDigest=d41d8cd98f00b204e9800998ecf8427e|cloudClusterName=UNKNOWN|TraceId=|WorkloadGroup=normal|FuzzyVariables=|scanBytesFromLocalStorage=0|scanBytesFromRemoteStorage=0
32024-07-18 11:23:41,044 [slow_query] |Client=127.0.0.1:26684|User=root|Ctl=internal|Db=tpch_sf1000|State=EOF|ErrorCode=0|ErrorMessage=|Time(ms)=8514|ScanBytes=334986551296|ScanRows=10717654374|ReturnRows=100|StmtId=1833|QueryId=4f91483464ce4aa8-beeed7dcb8675bc8|IsQuery=true|isNereids=true|feIp=172.21.0.10|Stmt=select s_name, count(*) as numwait from supplier, lineitem l1, orders, nation where s_suppkey = l1.l_suppkey and o_orderkey = l1.l_orderkey and o_orderstatus = 'F' and l1.l_receiptdate > l1.l_commitdate and exists ( select * from lineitem l2 where l2.l_orderkey = l1.l_orderkey and l2.l_suppkey <> l1.l_suppkey ) and not exists ( select * from lineitem l3 where l3.l_orderkey = l1.l_orderkey and l3.l_suppkey <> l1.l_suppkey and l3.l_receiptdate > l3.l_commitdate ) and s_nationkey = n_nationkey and n_name = 'SAUDI ARABIA' group by s_name order by numwait desc, s_name limit 100|CpuTimeMS=925841|ShuffleSendBytes=59223190|ShuffleSendRows=3651602|SqlHash=f8a30e4182d72cce3eff6cb385005b1f|peakMemoryBytes=10505123104|SqlDigest=d41d8cd98f00b204e9800998ecf8427e|cloudClusterName=UNKNOWN|TraceId=|WorkloadGroup=normal|FuzzyVariables=|scanBytesFromLocalStorage=0|scanBytesFromRemoteStorage=0
42024-07-18 11:23:49,044 [slow_query] |Client=127.0.0.1:10748|User=root|Ctl=internal|Db=tpch_sf1000|State=EOF|ErrorCode=0|ErrorMessage=|Time(ms)=8660|ScanBytes=334987673600|ScanRows=10717654374|ReturnRows=100|StmtId=1851|QueryId=4599cb1bab204f80-ac430dd78b45e3da|IsQuery=true|isNereids=true|feIp=172.21.0.10|Stmt=select s_name, count(*) as numwait from supplier, lineitem l1, orders, nation where s_suppkey = l1.l_suppkey and o_orderkey = l1.l_orderkey and o_orderstatus = 'F' and l1.l_receiptdate > l1.l_commitdate and exists ( select * from lineitem l2 where l2.l_orderkey = l1.l_orderkey and l2.l_suppkey <> l1.l_suppkey ) and not exists ( select * from lineitem l3 where l3.l_orderkey = l1.l_orderkey and l3.l_suppkey <> l1.l_suppkey and l3.l_receiptdate > l3.l_commitdate ) and s_nationkey = n_nationkey and n_name = 'SAUDI ARABIA' group by s_name order by numwait desc, s_name limit 100|CpuTimeMS=932664|ShuffleSendBytes=59223178|ShuffleSendRows=3651991|SqlHash=f8a30e4182d72cce3eff6cb385005b1f|peakMemoryBytes=10532849344|SqlDigest=d41d8cd98f00b204e9800998ecf8427e|cloudClusterName=UNKNOWN|TraceId=|WorkloadGroup=normal|FuzzyVariables=|scanBytesFromLocalStorage=0|scanBytesFromRemoteStorage=0通过 fe.audit.log 获取的慢 SQL，使用者可以方便地获取执行时间、扫描行数、返回行数、SQL 语句等详细信息，为进一步重现和定位性能问题奠定了基础。
audit_log 系统表
PALO 2.1 以后的版本在__internal_schema 数据库下提供了 audit_log 系统表，供用户查看 SQL 运行的情况。使用前需要打开全局配置 set global enable_audit_plugin=true;（此开关默认关闭）
SQL
1mysql> use __internal_schema;
2Reading table information for completion of table and column names
3You can turn off this feature to get a quicker startup with -A
4
5Database changed
6mysql> show tables;
7+-----------------------------+
8| Tables_in___internal_schema |
9+-----------------------------+
10| audit_log |
11| column_statistics |
12| histogram_statistics |
13| partition_statistics |
14+-----------------------------+
154 rows in set (0.00 sec)
16
17mysql> desc audit_log;
18+-------------------+--------------+------+-------+---------+-------+
19| Field | Type | Null | Key | Default | Extra |
20+-------------------+--------------+------+-------+---------+-------+
21| query_id | varchar(48) | Yes | true | NULL | |
22| time | datetime | Yes | true | NULL | |
23| client_ip | varchar(128) | Yes | true | NULL | |
24| user | varchar(128) | Yes | false | NULL | NONE |
25| catalog | varchar(128) | Yes | false | NULL | NONE |
26| db | varchar(128) | Yes | false | NULL | NONE |
27| state | varchar(128) | Yes | false | NULL | NONE |
28| error_code | int | Yes | false | NULL | NONE |
29| error_message | text | Yes | false | NULL | NONE |
30| query_time | bigint | Yes | false | NULL | NONE |
31| scan_bytes | bigint | Yes | false | NULL | NONE |
32| scan_rows | bigint | Yes | false | NULL | NONE |
33| return_rows | bigint | Yes | false | NULL | NONE |
34| stmt_id | bigint | Yes | false | NULL | NONE |
35| is_query | tinyint | Yes | false | NULL | NONE |
36| frontend_ip | varchar(128) | Yes | false | NULL | NONE |
37| cpu_time_ms | bigint | Yes | false | NULL | NONE |
38| sql_hash | varchar(128) | Yes | false | NULL | NONE |
39| sql_digest | varchar(128) | Yes | false | NULL | NONE |
40| peak_memory_bytes | bigint | Yes | false | NULL | NONE |
41| stmt | text | Yes | false | NULL | NONE |
42+-------------------+--------------+------+-------+---------+-------+通过 audit_log 内部表，用户可以查询详细的 SQL 执行信息，进行如慢查询筛选等详细统计分析。
总结
PALO Manager 日志，audit log 以及 audit_log 系统表等工具，可以提供慢 SQL 自动或手动筛选过滤，以及细粒度 SQL 执行信息统计分析等能力。这些工具为系统性的性能诊断和调优提供了强大支撑。