HI team In today Test Team Call was presented this commit: https://opendev.org/starlingx/config/commit/e51976083e6042546b1922e8a00bbb2e... To disable pre-allocation of 2M huge pages on VBox . Cristopher and I had the AR to test if the $ facter is_virtual return true in qemu , and it does. Which means that we should that the amount of memory for the baremetal system will not be the same than virtual environments We truly want to thank Tao Liu for this change, this reduces the footprint on VMs but not in bare metal. Is still open the CLI to fix the constraint of a VM/ bare metal system. For now, we can work with this patch and make some poc for this wished command that could be useful for developers and users of starlign x Thanks Victor R On Wed, May 8, 2019 at 12:01 PM Curtis <serverascode@gmail.com> wrote:

On Wed, May 8, 2019 at 11:33 AM Lemus Contreras, Cristopher J <cristopher.j.lemus.contreras@intel.com> wrote:

...

Hello,

During today’s Community meeting, memory usage on virtual environments was discussed. In summary, the memory requirements for each VM has increased significantly (see bug: https://bugs.launchpad.net/bugs/1826445 ).

I think it's important that we as a project try to keep memory resources as low as reasonably possible. I've seen this happen in other OpenStack projects where they find that their memory usage is increasing and they go and look and find some low hanging fruit where they can reduce requirements. I think we definitely need to take a look and see where we can reduce some memory usage.

IHMO, we also need need to make sure it's possible to do a PoC with low requirements. I find that during workshops etc that memory requirements are quite high. We need to document a virtual PoC All-in-one which is quite a bit smaller than prod environments. The place for this to happen may be starlingx-in-a-box but I haven't been able to look at that code yet.

Finally, I would expect to find people wanting to use stx in small ucpe style systems where memory could be hardware with less than 16gb. Certainly we could decide not to support that, but we shouldn't be surprised to see some people suggest that use case. :)

Thanks, Curtis

...

An alternative was mentioned, during the deployment it’s possible to limit the amount of memory that starlingx is going to use. Does somebody has the instructions to implement such limits? It’ll help us a lot to continue working with Virtual Environments.

Thanks in advance.

Cristopher Lemus

_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss

-- Blog: serverascode.com _______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss

Thanks a lot Brent This is really helpful, we will test and came back with feedback. Thanks a lot Victor R On Tue, May 14, 2019 at 3:01 PM Rowsell, Brent <Brent.Rowsell@windriver.com> wrote:

You can modify the memory allocations per function on the host using the command below. The node must be locked. After updating, unlock the node.

system help host-memory-modify

usage: system host-memory-modify [-m <Platform Reserved MiB>]

[-2M <2M hugepages number>]

[-1G <1G hugepages number>] [-f <function>]

<hostname or id> <processor>

Modify platform reserved and/or application huge page memory attributes for

worker nodes.

Positional arguments:

<hostname or id> Name or ID of host

<processor> processor

Optional arguments:

-m <Platform Reserved MiB>, --platform_reserved_mib <Platform Reserved MiB>

The amount of platform memory (MiB) for the numa node

-2M <2M hugepages number>, --hugepages_nr_2M_pending <2M hugepages number>

The number of 2M vm huge pages for the numa node

-1G <1G hugepages number>, --hugepages_nr_1G_pending <1G hugepages number>

The number of 1G vm huge pages for the numa node

-f <function>, --function <function>

The Memory Function.

Brent

From: Lemus Contreras, Cristopher J [mailto:cristopher.j.lemus.contreras@intel.com] Sent: Wednesday, May 8, 2019 11:30 AM To: starlingx-discuss@lists.starlingx.io Cc: Rowsell, Brent <Brent.Rowsell@windriver.com> Subject: Limit memory usage on Virtual Environment

Hello,

During today’s Community meeting, memory usage on virtual environments was discussed. In summary, the memory requirements for each VM has increased significantly (see bug: https://bugs.launchpad.net/bugs/1826445 ).

An alternative was mentioned, during the deployment it’s possible to limit the amount of memory that starlingx is going to use. Does somebody has the instructions to implement such limits? It’ll help us a lot to continue working with Virtual Environments.

Thanks in advance.

Cristopher Lemus

_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss