HANA Data File Partitioning - Installation
Published Jan 26 2021 07:53 AM 4,241 Views
Microsoft

HANA Data Path Striping (partitioning)

Since SAP HANA 2.0 SPS03 (Rev 35) and above versions, data volumes on the Indexserver can be partitioned so that read and write operations can run in parallel with increased data throughput. Also, the HANA startup time will benefit from this feature because the data throughput will increase massively, depending in the number and performance of the additional disks or volumes.

HANA data partitioning is only available for data volumes not for log volumes.

This, of course, can be configured on our HLI’s and on VM’s with Premium Managed Disks or Azure NetApp Files (ANF).

There is no additional certification required.

Interesting Links:

SAP HANA – Partitioning Data Volumes | SAP Blogs

2400005 - FAQ: SAP HANA Persistence - SAP ONE Support Launchpad

2700123 - Indexserver Crash at PageAccess::PageFlushCallback::pageChunkWritten With "Unexpected offs...

Partitioning Data Volumes - SAP Help Portal

It can be one option to create a HANA partitioning with 2 or 4 volumes…

This documentation is covering the configuration of 2 data volumes.

HANA Disk PartitioningHANA Disk Partitioning

Create a second data path directory.

 

 

mkdir -p /hana/data2/ML4/mnt00001

 

 

Create the /etc/fstab entry.

 

 

vi /etc/fstab

10.20.255.31:/hana_data2_a91_t250_vol            /hana/data2/ML4/mnt00001 nfs      rw,hard,nosuid,timeo=600,vers=4.1,rsize=1048576,wsize=1048576,intr,noatime,lock 0 0

 

 

Set the user permissions.

 

 

chown -R ml4adm:sapsys /hana/data2

 

 

Enable HANA to work with multi data partitioning.

 

 

su – ml4adm
cdcoc
vi global.ini
[customizable_functionalities]
persistence_datavolume_partition_multipath = true

 

 

restart HANA

 

 

sapcontrol -nr 02 -function StopSystem
sapcontrol -nr 02 -function StartSystem

 

 

See the actual size of the datavolume 1:

 

 

ls -l /hana/data/ML4/mnt00001/hdb00003.00003/
ml4adm sapsys 127708168192 Jan 18 08:01 datavolume_0000.dat

 

 

 Execute the following SQL statement against the tenant database:

 

 

hdbsql -u system -p <PASSWD> -n localhost:30215
ALTER SYSTEM ALTER DATAVOLUME ADD PARTITION PATH '/hana/data2/ML4';

 

 

HANA will do the striping over time.

 

 

ls -l /hana/data2/ML4/mnt00001/hdb00003.00003/
ml4adm sapsys 0 Jan 18 08:04 datavolume_0001.dat

 

 

After some time the situation will change….

 

 

ls -l /hana/data2/ML4/mnt00001/hdb00003.00003/
ml4adm sapsys 23219666944 Jan 18  2021 datavolume_0001.dat


ls -l /hana/data/ML4/mnt00001/hdb00003.00003/
ml4adm sapsys 127708168192 Jan 18 08:43 datavolume_0000.dat

 

 

Benefits of Data Partitioning

With HANA data partitioning the HANA Savepoint will benefit massively from the striping over many disks/volumes. This Graph shown the savepoint behavior with a single data volume:

HCMT with multiple volumes

To measure the additional data files with HCMT use the appropriate json config file.

This config file is called  multivolume.json

Here is a place where we can add the additional datafiles…

 

 

 

 

  "Variables": [
      {
         "Comment": "The <Value> should be adapted to the HANA log multi volume paths",
         "Name": "LogVolumes",
         "Value": "/hana/log/ML4/mnt00001",
         "Request": "true"
      },
      {
         "Comment": "The <Value> should be adapted to the HANA data multi volume paths",
         "Name": "DataVolumes",
         "Value": "/hana/data/ML4/mnt00001,/hana/data2/ML4/mnt00001",
         "Request": "true",
         "Profile": "LNX"
      },

 

 

 

This is just a result on my system. Performance surly is dependent on many different parameter and configuration options.

 

As a result, we can compare the data write results.

Block size         Data Write values with one data-file              Data write values with two data-files

                                   Initialwrite               Overwrite                         Initialwrite              Overwrite   

Screenshot 2021-01-26 164200.jpg

Backup

For snapshot-based backups, please play close attention that you snapshot all data volumes belonging to the same database.

Version history
Last update:
‎Jan 26 2021 07:56 AM
Updated by: