1G hugepages for memory allocation is not enable in Virtual external storage config

Bug #1799539 reported by Fernando Hernandez Gonzalez on 2018-10-23
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
StarlingX
Low
Tao Liu

Bug Description

Brief Description
-----------------
Once you lock a compute and go to the Memory tab and hit update memory button the 1G hugepages for memory allocation is not enable in Virtual external storage config

Severity
--------
Minor

Steps to Reproduce
------------------
Go to Admin--Platform--Host inventory
Lock a compute node
Once locked hit the compute link and go to Memory host detail tab.
Go to "Update Memory"
Update Memory Allocation pop up windos should be displayed.
Go to "# of VM 1G Hugepages node" field and you will see the field is not editable.

Even y you reduce the VM 2M Hugepages node size the VM 1G remains not editable.
Not sure if this is a feature enable for Virtual External storage.

Expected Behavior
------------------
"# of VM 1G Hugepages node" field should be editable

Actual Behavior
----------------
"# of VM 1G Hugepages node" field is not editable.

Reproducibility
---------------
100% reproducible.

System Configuration
--------------------

Virtual Multinode External Ceph Storage + 2 controllers + 2 computes + 2 storages.

Tao Liu (tliu88) wrote :

Can you verify that the compute node supports 1G huge page?

Use CLI
system host-memory-show <compute-node> <processor>
for example system host-memory-show compute-0 0
if the VM Huge Pages (1G): Available shows None, it means 1G huge pages is not supported on this node.

From Linux, if /mnt/huge-1048576kB does not exist, it means 1G huge page is not supported.
sudo mount | grep /mnt/huge
none on /mnt/huge-1048576kB type hugetlbfs (rw,relatime,pagesize=1048576kB)
none on /mnt/huge-2048kB type hugetlbfs (rw,relatime,pagesize=2048kB)

Ghada Khalil (gkhalil) wrote :

This is not stx.2018.10 gating - it's a minor issue and maybe expected based on the used system. Removing the label and waiting for more info from the reporter.

Changed in starlingx:
status: New → Incomplete
tags: removed: stx.2018.10

Tao Liu you are right, the compute node is not supporting 1G
[wrsroot@compute-0 ~(keystone_admin)]$ system host-memory-show compute-0 0
+--------------------------------+--------------------------------------+
| Property | Value |
+--------------------------------+--------------------------------------+
| Memory: Usable Total (MiB) | 14195 |
| Platform (MiB) | 1600 |
| Available (MiB) | 13171 |
| Huge Pages Configured | True |
| vSwitch Huge Pages: Size (MiB) | 2 |
| Total | 512 |
| Available | 0 |
| VM Pages (4K): Total | 1638656 |
| VM Huge Pages (2M): Total | 3385 |
| Available | 3385 |
| VM Huge Pages (1G): Total | 0 |
| Available | *None* |
| uuid | 2e8f1f83-d436-46ac-8ba9-7f49d0717c49 |
| ihost_uuid | d2976cf4-d0bd-49ef-ad14-ac35892459f0 |
| inode_uuid | f6cdbe90-2837-47e4-89f3-46cd5a6a0fe1 |
| created_at | 2018-10-23T09:04:02.066457+00:00 |
| updated_at | 2018-10-23T10:28:41.805357+00:00 |
+--------------------------------+--------------------------------------+

I went through cat /proc/cmdline from my compute and I can not see hugeapgesz=1G neither hugepages=1

BOOT_IMAGE=/vmlinuz-3.10.0-862.11.6.el7.36.tis.x86_64 root=UUID=7e095964-d952-4265-9bd9-5999a8a71058 ro security_profile=standard module_blacklist=integrity,ima audit=0 tboot=false crashkernel=auto biosdevname=0 console=ttyS0,115200 iommu=pt usbcore.autosuspend=-1 selinux=0 enforcing=0 nmi_watchdog=panic,1 softlockup_panic=1 intel_iommu=on user_namespace.enable=1 nopti nospectre_v2 hugepagesz=2M hugepages=0 default_hugepagesz=2M isolcpus=1 rcu_nocbs=1-3 kthread_cpus=0 irqaffinity=0

I asked to Juan Carlos to check his /proc/cmdline in his configuration (MULTINODE LOCAL STORAGE VIRT) and his computes has the 1G declared.

Tao Liu my question should be if Virtual Mode with External Ceph Storage should have this 1G hugepages declared in the system?

I also went through the /proc/cmdline on my controller-0 and it does not have 1G neither.

Changed in starlingx:
status: Incomplete → New
Ghada Khalil (gkhalil) wrote :

Marking as Invalid. The reporter confirmed that the compute node used does not support 1G huge pages.

Changed in starlingx:
status: New → Incomplete
assignee: nobody → Tao Liu (tliu88)
status: Incomplete → Invalid
importance: Undecided → Low
tags: added: stx.metal
Tao Liu (tliu88) wrote :

If the cpuinfo shows 1G hugepage is supported, it will be declared in the system.

Tao Liu (tliu88) wrote :

The 'system' in my last comment refers to the node that has compute function.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers