Mysql Show InnoDB status not reporting buffer pool stats correctly on Linux

Question

We have multiple slaves running MySQL 5.7 - some on Linux (CentOS 7) and some on Windows. We're trying to diagnose an issue where our linux boxes randomly start falling behind with no long running queries or locks or dramatic increases in writes and reads.

Our error logs on the Linux boxes are filled with "InnoDB: page_cleaner: 1000ms intended loop took x ms. The settings might not be optimal." messages. Which suggests that we're flushing a lot of dirty pages from the buffer pool.

When viewing the innodb status: SHOW ENGINE INNODB STATUS

Our Windows boxes show non-zero values:

However, the Linux machines' buffer pool status show 0.00 for youngs/s, non-youngs/s, reads/s etc. which is not true.

Any ideas how to get the Linux boxes to report correctly?

score 0 · Answer 1 · answered May 18 '21 at 21:13

0

Lowering lru_scan_depth (which is probably 1024 now) to, say, 20, may help performance and may make those messages go away. (Please report back on what results you get.)

answered May 18 '21 at 21:13

Rick James

135,179
13
127
222

Thanks you - our innodb_lru_scan_depth is 1024 on both the windows boxes and the linux boxes. But I'll try it out and report back. – Matthew Conradie May 19 '21 at 09:01

Mysql Show InnoDB status not reporting buffer pool stats correctly on Linux

1 Answers1