Uploaded image for project: 'OpenShift Request For Enhancement'
  1. OpenShift Request For Enhancement
  2. RFE-8380

LVM Storage operator - Ability to stripe local NVMe disks

XMLWordPrintable

    • Icon: Feature Request Feature Request
    • Resolution: Unresolved
    • Icon: Normal Normal
    • None
    • None
    • LVMS
    • None
    • Product / Portfolio Work
    • None
    • False
    • Hide

      None

      Show
      None
    • None
    • None
    • None
    • None
    • None
    • None
    • None
    • None

      1. Proposed title of this feature request
      -->> LVM Storage operator - Ability to stripe local NVMe disks
      2. What is the nature and description of the request?
      -->> When we use LVM Storage operator, we are facing an issue where the operator uses NVMe disks in a linear way.
      The problem is that this fills the disks sequentially, using one disk at a time instead of all disks in parallel.
      For example: 
      sh-5.1# dmsetup table       
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 0 3750739968 linear 259:1 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 3750739968 3750739968 linear 259:0 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 7501479936 3750739968 linear 259:2 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 11252219904 3750739968 linear 259:3 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 15002959872 3750739968 linear 259:4 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 18753699840 3750739968 linear 259:5 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 22504439808 3750739968 linear 259:6 2048
      splunk-localnvme-ac51c8c5bb29439e840e-0bd73c3cfcd0: 26255179776 1007796224 linear 259:7 2048
      3. Why does the customer need this? (List the business requirements here)
      -->> 

      We would like to know if there is a way for the LVM Storage operator to create a stripe array instead of a linear.

      We have already consulted the documentation, and the manual command "mdadm --create /dev/md0 --level=1 --raid-devices=2 /dev/sda1 /dev/sdc1" does not allow us to automate this because each time a new node is created, the command needs to be entered again, so this does not work. We would need this to implemented at operator level

      4. List any affected packages or components.
      -->> LVMS
      We discussed the above over slack, and based on that  converstaion we shared the below update with the customer:
      ~~~

      (1) - Current Allocation: The LVM Storage (LVMS) Operator currently provisions volumes using a linear allocation strategy by default. This is why you are seeing volumes fill one disk before moving to the next.

      (2) - Striping/RAID Support: Unfortunately, native operator-level support for automated striping or LVM RAID configurations is currently not available in the LVM Storage Operator.

       

      NOTE: This limitation is noted in the public documentation, here: https://github.com/openshift/lvm-operator?tab=readme-ov-file#missing-native-lvm-raid-configuration-support.

      We recognize that the manual mdadm RAID setup you tested is not a viable, scalable solution since it requires manual reapplication on every new node.
      ~~~
      Slack thread: 
      https://redhat-internal.slack.com/archives/C060V75JF4J/p1759764404230579

      To which customer asked us to create an RFE

              dfroehli42rh Daniel Fröhlich
              rhn-support-gakendre Gaurav Kendre
              None
              Votes:
              0 Vote for this issue
              Watchers:
              5 Start watching this issue

                Created:
                Updated:
                None
                None