如何阅读Cassandra性能跟踪表

时间:2016-05-11 17:29:00

标签: cassandra

在Cassandra 3.5中,我设置了" TRACING ON"并运行一个选择并获得下表。我不知道该怎么读。你能告诉我第二个子任务"解析语句"需要139微秒或192-139 = 53微秒?数据存储网站上的指南不是很清楚。

如果格式太杂乱,请注意,请参阅我的链接。 http://i.stack.imgur.com/hHbpm.png

activity    timestamp   source  source_elapsed
Execute CQL3 query  2016-05-11 12:33:21.124000  192.160.75.0    0
Parsing SELECT * FROM ratesbt.metrics where name = 'A' ; [SharedPool-Worker-1]  2016-05-11 12:33:21.124000  192.160.75.0    139
Preparing statement [SharedPool-Worker-1]   2016-05-11 12:33:21.124000  192.160.75.0    192
Executing single-partition query on metrics [SharedPool-Worker-2]   2016-05-11 12:33:21.124000  192.160.75.0    348
Acquiring sstable references [SharedPool-Worker-2]  2016-05-11 12:33:21.124000  192.160.75.0    366
Key cache hit for sstable 142 [SharedPool-Worker-2] 2016-05-11 12:33:21.124000  192.160.75.0    397
Skipped 0/1 non-slice-intersecting sstables, included 0 due to tombstones [SharedPool-Worker-2] 2016-05-11 12:33:21.124001  192.160.75.0    409
Merging data from memtables and 1 sstables [SharedPool-Worker-2]    2016-05-11 12:33:21.124001  192.160.75.0    421
Read 3972 live and 0 tombstone cells [SharedPool-Worker-2]  2016-05-11 12:33:21.135000  192.160.75.0    10962
Request complete    2016-05-11 12:33:21.141442  192.160.75.0    17442

1 个答案:

答案 0 :(得分:0)

从请求开始的时间戳开始经过的时间。因此,根据消息(大多数事件是在一个动作发生后,但有些事件可以在它们开始之前标记),你可以弄清楚什么是花费大部分时间。

在你的情况下,从磁盘上读取3k个单元并将其与记忆合并,耗时11毫秒。如果sstable [s]没有被OS缓存(取决于磁盘),这并不是坏事。