Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

innodb_buffer_pool_size and innodb_buffer_pool_chunk_size calculation does not match per formula in autotune.go

Description

Hello Team,

I have tried to deploy Operator v1.11.0 and Operator v1.13.0 without any explicit configuration of innodb_buffer_pool_size and innodb_buffer_pool_chunk_size.

Resource limits for PXC used in cr.yaml file were

With Operator v1.11.0

With Operator v1.13.0

Operator when trying to setup buffer pool is not setting it according to the calculations in autotune.go

Here is the code that does the calculation,

As per the calculation it should be as below,

We get proper calculations in Operator v1.11.0 but it does not matches in Operator v1.13.0.

The variable innodb_dedicated_server is also set to off by default.

Is it calculating buffer_pool_size from something else?

Let me know if there is any question.

Regards,

Yunus Shaikh

 

 

Environment

None

Activity

ege.gunes December 22, 2023 at 11:02 AM

It’d be nice to get input from and

inel.pandzic December 21, 2023 at 11:59 AM
Edited

Hello

The reason why we have different values right now is because we execute this code:

We want to make poolSize a multiple of chunkSizeDefault (or innodb_buffer_pool_chunk_size which is 128). So now poolSize is not 3000000000 but 3087007744

Because of this we now have:

As you can see the result is still a good calculation in terms of poolSize is innodb_buffer_pool_chunk_size * innodb_buffer_pool_instances (386924544 * 8 = 3095396352)

Now my question is is this still wrong?

 

To have values like in 1.11.0, we just need to stop making poolSize a multiple of chunkSizeDefault, meaning we can just remove the chunk of code I pasted above and it should be good.

 

fyi

Details

Assignee

Reporter

Reviewer

Needs Review

No

Affects versions

Priority

Smart Checklist

Created August 30, 2023 at 1:25 PM
Updated March 5, 2024 at 5:23 PM