Re: [PATCH v6 0/2] Methodology change for hugepage configuration

["Juraj Linkeš" <juraj.linkes@pantheon.tech>]

On Wed, May 15, 2024 at 4:50 PM Nicholas Pratte <npratte@iol.unh.edu> wrote:
>
> >What's the difference between this version and v4?
>
> Version 5 was a response to your suggestions regarding the semantics
> of the hugepage variable names as it relates to countable or
> uncountable nouns. This patch, which was originally just a single
> patch, was expanded into a patch series since the variable changes
> suggested are sort of logically independent of the original patch;
> there were talks about potentially making this a separate patch
> entirely, but we agreed that a patch series for this change is
> appropriate.
>

Ah, ok. There were a lot of patches so that confused me.

> But, to answer a question that I think you are indirectly asking, I
> ran into some issues when sending out the latest version of this
> patch. While the latest version of this patch, version 6, was sent
> last week and does show up on patchwork, it seems like the emails
> never reached the mailing list for some (I just asked Patrick in the
> office and he cannot seem to find them).
>
> >      "type": "object",
> >       "description": "Optional hugepage configuration. If not specified, hugepages won't be configured and DTS will use system configuration.",
> >       "properties": {
> >-        "amount": {
> >+        "quantity": {
> >           "type": "integer",
> >-          "description": "The amount of hugepages to configure. Hugepage size will be the system default."
> >+          "description": "The number of hugepages to configure. Hugepage size will be the system default."
> >         },
> >         "force_first_numa": {
> >           "type": "boolean",
>
> The only change that version 6 made is in reference to the suggestion
> that Bruce made in a separate thread; I changed the description text
> from "amount of" to "number of."
>

As you mentioned in the other thread, I also like changing it in all
instances (in code and conf.yaml).

> Hopefully this answers your question. Again, I ran into some issues
> sending the latest version out for this patch
>
> On Mon, May 13, 2024 at 5:53 AM Juraj Linkeš <juraj.linkes@pantheon.tech> wrote:
> >
> > What's the difference between this version and v4?
> >
> > On Tue, May 7, 2024 at 7:44 PM Nicholas Pratte <npratte@iol.unh.edu> wrote:
> > >
> > > In order to prevent accidental misconfiguration of hugepages at runtime,
> > > the following changes are made to only allow for configuration of 2MB
> > > hugepages within the DTS config.yaml. In the previous implementation, a
> > > default hugepage size was selected via the size listed in /proc/meminfo.
> > > The problem with this implementation is that, assuming the end-user has
> > > made prior modifications to the system, /proc/meminfo may default to
> > > hugepage sizes that are not recommended to be configured at runtime
> > > (i.e. 1GB hugepages). This can lead to two problems: overallocation of
> > > hugepages (which may crash the remote host) configuration of hugepages
> > > sizes that are not recommended during runtime. In this new implementation,
> > > we stipulate that any runtime hugepage configuration size that is not 2MB
> > > is considered an outlier. If the end-user would like to configure either
> > > 1GB hugepages or any unique hugepage size outside of 2MB, then they should
> > > make these configurations either at startup (in the case of 1GB hugepages)
> > > or runtime outside of DTS configuration (if a user would like hugepages
> > > that are not 2MB). In either case, the expectation is that, if wish to
> > > use hugepage sizes that are not 2MB, you will make these changes outside
> > > and prior to the initialization of DTS.
> > >
> > > The end-user has two options: remove the option for hugepage
> > > configuration in the conf.yaml, or keep the option and specify the
> > > amount of 2MB hugepages desired. In the case of the former, then we assume
> > > that hugepages are already configured prior to DTS initialization. In
> > > the latter case, the user must define the amount of 2MB hugepages to be
> > > configured at runtime. If the amount of 2MB hugepages requested exceeds
> > > the amount of 2MB hugepages already configured on the system, then the
> > > system will remount hugepages to cover the difference. If the amount of
> > > hugepages requested is either greater than or equal to the amount
> > > already configured on the system, then nothing is done.
> > >
> > > Nicholas Pratte (2):
> > >   dts: Change hugepage runtime config to 2MB Exclusively
> > >   dts: Change hugepage 'amount' to a different term
> > >
> > >  doc/guides/tools/dts.rst                     |  6 ++++-
> > >  dts/conf.yaml                                |  8 +++---
> > >  dts/framework/config/__init__.py             |  8 +++---
> > >  dts/framework/config/conf_yaml_schema.json   | 12 ++++-----
> > >  dts/framework/config/types.py                |  4 +--
> > >  dts/framework/testbed_model/linux_session.py | 28 +++++++++++---------
> > >  dts/framework/testbed_model/node.py          |  4 ++-
> > >  dts/framework/testbed_model/os_session.py    |  7 ++++-
> > >  8 files changed, 45 insertions(+), 32 deletions(-)
> > >
> > > --
> > > 2.44.0
> > >