Skip to content

Latest commit

 

History

History
142 lines (106 loc) · 5.13 KB

sap-post-provisioning-intel-bare-metal-optane.md

File metadata and controls

142 lines (106 loc) · 5.13 KB
copyright lastupdated keywords subcollection
years
2021
2021-05-10
SAP, {{site.data.keyword.cloud_notm}} SAP-Certified Infrastructure, {{site.data.keyword.ibm_cloud_sap}}, SAP Workloads
sap

{:shortdesc: .shortdesc} {:codeblock: .codeblock} {:screen: .screen} {:external: target="_blank" .external} {:pre: .pre} {:table: .aria-labeledby="caption"} {:note: .note} {:tip: .tip} {:important: .important}

Post-Provisioning Intel Optane PMem

After you order your Bare Metal server and the server is provisioned, you need to do complete the provisioning by creating namespaces for your system.

PMem on Bare Metal supports the App Dir mode. In App Dir mode, PMem and DRAM DIMMs act as independent memory resources that are directly accessed by the applications.

The App Dir mode uses regions, and namespaces to represent persistent memory devices in an interleaved set. A region represents the physical persistent memory devices. A region is made up of one or more namespaces. A namespace represents a unit of storage that can be used for Input/Output (I/O).

Regions are created as part of the provisioning process. You have to create the namespaces, create mount points, and mount the PMem devices.

For more information on regions and namespaces, see <link to Intel Optane Persistent Memory and SAP HANA Platform Configuration (https://www.sap.com/documents/2019/05/7a278c36-507d-0010-87a3-c30de2ffd8ff.html)>.

Procedure

After patching the machine, do these steps:

KT comment 8 – what patch is needed? Does IBM do this or the customer?

KT comment 9 – is the ndctl software already on the server and just needs to be installed? Or does the customer have to download it?

  1. Install the ndctl software for your system:

    • For SLES:

      $ zypper in ndctl

      Refreshing service 
'SMT-https_susesapsmtamr_service_networklayer_com'.
Loading repository data...
Reading installed packages...
Resolving package dependencies...
The following 2 NEW packages are going to be installed:
libndctl6 ndctl

    • For RHEL:

      $ yum install ndctl

      Refreshing service 
'SMT-https_susesapsmtamr_service_networklayer_com'.
Loading repository data...
Reading installed packages...
Resolving package dependencies...

  2. List the available regions for your system: $ ndctl list -R -v

    [
{
    "dev":"region1",
    "size":1623497637888,
    "available_size":1623497637888,
    "max_available_extent":1623497637888,
    "type":"pmem",
    "numa_node":1,
    "target_node":3,
    "iset_id":-7847382400914477876,
    "persistence_domain":"memory_controller"
},
{
    "dev":"region0",
    "size":1623497637888,
    "available_size":1623497637888,
    "max_available_extent":1623497637888,
    "type":"pmem",
    "numa_node":0,
    "target_node":2,
    "iset_id":9195364489516887244,
    "persistence_domain":"memory_controller"
}

  3. Create a namespace for each of your regions.

    $ ndctl create-namespace -r region0

     {
 "dev":"namespace0.0",
 "mode":"fsdax",
 "map":"dev",
 "size":"1488.37 GiB (1598.13 GB)",
 "uuid":"75af4e28-1b99-4d34-a39a-a342a370232a",
 "sector_size":512,
 "align":2097152,
 "blockdev":"pmem0"
 }

    Repeat this command for region1

KT comment 10 – are the customers always going to want to use fsdax mode for namespaces? Or is it the only one we support? There are 4. How much information do I need to provide about the namespace modes to help the users make this decision? Or is it a non-decision?

  1. Create an xfs on-top of the pmem devices: $ mkfs -t xfs -f /dev/pmem0

     meta-data=/dev/pmem0             isize=512    agcount=4, agsize=97542016 blks
         =                       sectsz=4096  attr=2, projid32bit=1
         =                       crc=1        finobt=1, sparse=0, rmapbt=0
         =                       reflink=0
 data    =                       bsize=4096   blocks=390168064, imaxpct=5
         =                       sunit=0      swidth=0 blks
 naming   =version 2              bsize=4096   ascii-ci=0, ftype=1
 log      =internal log           bsize=4096   blocks=190511, version=2
         =                       sectsz=4096  sunit=1 blks, lazy-count=1
 realtime =none                   extsz=4096   blocks=0, rtextents=0

    Repeat this command for pmem1

  2. Mount the file system you just created by adding the following lines to /etc/fstab:

    /dev/pmem0 /hana/pmem/nvmem0 xfs dax 0 0 /dev/pmem1 /hana/pmem/nvmem1 xfs dax 0 0

  3. Create the path to the device.

    $ mkdir -p /hana/pmem/nvmem0 $ mkdir -p /hana/pmem/nvmem1 $ mount -a

  4. Verify the devices.

    $ df -h | grep pmem

     /dev/pmem0      1.5T  1.6G  1.5T   1% /hana/pmem/nvmem0
 /dev/pmem1      1.5T  1.6G  1.5T   1% /hana/pmem/nvmem1