general pool is less than 95% of total physical RAM

Izayda

We observe strange situation with memory available for general pool.

There is 252,3 GB total physical memory on each node.
The maxmemorysize parameter for general pool is 95%.
According to documentation, we are expecting that 239,685 GB of memory will be available for general pool.

But we see only 199,9 GB memory available for general in v_monitor.resource_pool_status (memory_size_kb column). That is only 79% of total physical RAM.
Moreover, memory_size_kb data slightly differs for each node. That is also strange.

Thank you.

PS: ulimit -a and /etc/security/limits.conf don't have any restrictions for dbadmin user

bhanu_kishore

Hi ,

Can u please share output of below queries.

Select * from host_resources:

Select * from resource_pool_status:

Izayda

Hi,
thanks for answer.

host_resources

https://www.dropbox.com/s/eknit1fgqu5uslm/host_resources.xlsx?dl=0

resource_pool_status

https://www.dropbox.com/s/2nazhp9w769bia5/resource_pool_status.xlsx?dl=0

[Deleted User]

If you are using 8.x version of Vertica, Vertica has introduced new system pool "metadata" that is used to track memory used by catalog . Can you please check memorysize of each resource pool in your cluster and add them .

Izayda

Yes, we are using 8.0.1-3.
Average size of metadata pool is about 5-6GB.
Total size of built-in pools (including metadata pool) is about 14,8GB.
Detailed info can be found in previous message.

[Deleted User]

You will have to total memorysize of all pools built-in pool and user pools .

Izayda

Now i realize, that the memorysize parameter sum of other pools, beside general is about 39GB. So, missed gigabytes are found We didn't have proper understanding of general pool.

Am I right that maxmemorysize parameter for general pool (95%) is applied before extracting reserved memory for other pools?

[Deleted User]

yes , you got it

Izayda

Ok, thank you very much!
We will reconfigure our pools.