Understanding the PMM Server requirements(CPU and RAM or if any)

Percona Monitoring and Management (PMM) PMM 2.x
1

Hi Team,

Can you please help me the system requirements for the below? We need table stats and Query analytics stats as well.

pmm-admin add mysql --server-insecure-tls --username=‘mon’ --password=‘XXXXX’ --host=‘prod401’ --port=3111 --service-name=Prod --disable-tablestats-limit=5000 --query-source=slowlog --size-slow-logs=1GiB

I am using all the above options for all the 30 servers. so please suggest me accrodingly.
Total production DB servers: 30

Please give me the below details?
RAM:
CPU:

If any other requirements, please suggest me.

2

Hi Naresh,

Such requirements are provided on the next page:

1 Like
3

Hi Adivinho,

Thanks for the update.

As per the above document, I got clarified for memory requirement(20 nodes). But need clarification on CPU, if we have a 20 nodes then how many CPU’s required, this information is not understandable from the above page.

1 Like
4

Hi, PMM doesn’t normally use a lot of CPU. A 2 core instance should be enough for a setup like yours.

1 Like
5

Hi Igroene,

Thanks for the update. I just want to know what are the requirements and how it will be, whether I am doing anything wrong or not.
Actually what I did and what issue I have faced in the below.

After adding the --disable-tablestats-limit=5000 option to the mysql service then we are seeing the huge CPU utilization.

We are seeing the CPU utilization only for Victoria metrics.

We have added total 20 servers and we are seeing High CPU utilization for only 20 servers. We have another 80 servers need to add the PMM server.

RAM: 16 GB

Current CPU’s: 8 ( But initially we tried with 4 CPU and then 6 CPU even though same issue, so we increased it to 8CPU)

PMM Version: 2.25.0

Attaching the screenshots for reference. After removing the --disable-tablestats-limit=5000 option then the CPU utilization is normal.

image
image1809×379 69.5 KB

image
image1138×649 322 KB

Even memory utilization also high sometimes but after removing the --disable-tablestats-limit=5000 option then memory is coming down.

image
image1828×539 65.7 KB

1 Like
6

The option changes the number of tables beyond which per-table statistics collection is disabled.
The default value is 1000.

1 Like
7

Hi Adivinho,

Thanks for the update. I knew the default value is 1000. but why CPU is high when we enabled the --disable-tablestats-limit and when we completely disabled the tables stats then the CPU is normal. If we see only victoriametrics taking more CPU’s and remaining services are normal.

1 Like
8

It’s required more CPU time for collecting statistics for 5000 tables than for 1000 or none.

1 Like
9

Hi Adivinho,

Thanks for the update.

Seems we have few bugs in the Victoria Metrics and some are related to CPU and Memory, currently there is a version upgradation planned for the next release as per Jira ticket.

https://jira.percona.com/browse/PMM-9358

1 Like