starlingx/docs
Code Issues Proposed changes

Merge "Stale SMT information (r7,dsR7)"

Browse Source
This commit is contained in:
Zuul 2022-10-18 13:29:01 +00:00 committed by Gerrit Code Review
commit 60ce38a1ce
1 changed files with 9 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
doc/source/admintasks/kubernetes/isolating-cpu-cores-to-enhance-application-performance.rst
View File

@ -35,14 +35,16 @@ The use of application-isolated cores is only applicable when using the static
Kubernetes CPU Manager policy. For more information, Kubernetes CPU Manager policy. For more information,
see :ref:`Kubernetes CPU Manager Policies <kubernetes-cpu-manager-policies>`. see :ref:`Kubernetes CPU Manager Policies <kubernetes-cpu-manager-policies>`.
**Limitation**: If Hyperthreading is enabled in the BIOS and .. note::
application-isolated CPUs are configured, and these CPUs are allocated to more
than one container, the |SMT| siblings may be allocated to different containers
and that could adversely impact the performance of the application.
**Workaround**: The suggested workaround is to allocate all |prod| isolcpus allocation is |SMT|-aware. If a container requests multiple
application-isolated CPUs on a host to a single pod. For more information, see isolcpus it will provide |SMT| siblings to the extent possible. If an odd
Node Management: :ref:`Changing the Hyper-threading Status <changing-the-hyper-threading-status>`. number of isolcpus are requested it will provide as many |SMT| siblings as
are available, then allocate singletons whose sibling has already been
allocated, then allocate one sibling from a free |SMT| sibling pair. If
hyperthreading is enabled in the BIOS then containers should request isolcpus
in pairs. If all containers on a system do this then they will never have
different containers being allocated |SMT| siblings from the same core.
When using the static CPU manager policy before increasing the number of When using the static CPU manager policy before increasing the number of
platform CPUs or changing isolated CPUs to application CPUs on a host, ensure platform CPUs or changing isolated CPUs to application CPUs on a host, ensure