| Message ID | 1493056114-9816-1-git-send-email-ktraynor@redhat.com |
|---|---|
| State | Accepted |
| Headers | show |
On Mon, Apr 24, 2017 at 06:48:34PM +0100, Kevin Traynor wrote: > Using dpdk-socket-mem to allocate memory for some NUMA nodes > but leaving blank for subsequent ones is equivalent of assigning > 0 MB memory to those subsequent nodes. Document this behavior. > > Signed-off-by: Kevin Traynor <ktraynor@redhat.com> Applied to master. Thank you for working on the documentation! (Let me know if this should be applied to any release branches.)
On 04/24/2017 07:50 PM, Ben Pfaff wrote: > On Mon, Apr 24, 2017 at 06:48:34PM +0100, Kevin Traynor wrote: >> Using dpdk-socket-mem to allocate memory for some NUMA nodes >> but leaving blank for subsequent ones is equivalent of assigning >> 0 MB memory to those subsequent nodes. Document this behavior. >> >> Signed-off-by: Kevin Traynor <ktraynor@redhat.com> > > Applied to master. Thank you for working on the documentation! > > (Let me know if this should be applied to any release branches.) > Well, it's probably not obvious that a trailing blank means 0 MB and those NUMA nodes will not be useable by the DPDK datapath, so could be useful info to have in 2.6/2.7 too. It will apply cleanly on branch-2.7. I've sent separate patch for branch-2.6. thanks, Kevin.
On Tue, Apr 25, 2017 at 02:06:25PM +0100, Kevin Traynor wrote: > On 04/24/2017 07:50 PM, Ben Pfaff wrote: > > On Mon, Apr 24, 2017 at 06:48:34PM +0100, Kevin Traynor wrote: > >> Using dpdk-socket-mem to allocate memory for some NUMA nodes > >> but leaving blank for subsequent ones is equivalent of assigning > >> 0 MB memory to those subsequent nodes. Document this behavior. > >> > >> Signed-off-by: Kevin Traynor <ktraynor@redhat.com> > > > > Applied to master. Thank you for working on the documentation! > > > > (Let me know if this should be applied to any release branches.) > > > > Well, it's probably not obvious that a trailing blank means 0 MB and > those NUMA nodes will not be useable by the DPDK datapath, so could be > useful info to have in 2.6/2.7 too. > > It will apply cleanly on branch-2.7. I've sent separate patch for > branch-2.6. Thanks for thinking it through. I crossported to branch-2.7 and applied the separate patch to branch-2.6.
diff --git a/Documentation/intro/install/dpdk.rst b/Documentation/intro/install/dpdk.rst index f29ac05..d1c0e65 100644 --- a/Documentation/intro/install/dpdk.rst +++ b/Documentation/intro/install/dpdk.rst @@ -227,9 +227,14 @@ listed below. Defaults will be provided for all values not explicitly set. If allocating more than one GB hugepage, you can configure the amount of memory used from any given NUMA nodes. For example, to use 1GB from -NUMA node 0, run:: +NUMA node 0 and 0GB for all other NUMA nodes, run:: $ ovs-vsctl --no-wait set Open_vSwitch . \ other_config:dpdk-socket-mem="1024,0" +or:: + + $ ovs-vsctl --no-wait set Open_vSwitch . \ + other_config:dpdk-socket-mem="1024" + Similarly, if you wish to better scale the workloads across cores, then multiple pmd threads can be created and pinned to CPU cores by explicity diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml index 870c813..02980b1 100644 --- a/vswitchd/vswitch.xml +++ b/vswitchd/vswitch.xml @@ -260,6 +260,7 @@ <p> The specifier is a comma-separated string, in ascending order of CPU - socket (ex: 1024,2048,4096,8192 would set socket 0 to preallocate - 1024MB, socket 1 to preallocate 2048MB, etc.) + socket. E.g. On a four socket system 1024,0,2048 would set socket 0 + to preallocate 1024MB, socket 1 to preallocate 0MB, socket 2 to + preallocate 2048MB and socket 3 (no value given) to preallocate 0MB. </p> <p>
Using dpdk-socket-mem to allocate memory for some NUMA nodes but leaving blank for subsequent ones is equivalent of assigning 0 MB memory to those subsequent nodes. Document this behavior. Signed-off-by: Kevin Traynor <ktraynor@redhat.com> --- Documentation/intro/install/dpdk.rst | 7 ++++++- vswitchd/vswitch.xml | 5 +++-- 2 files changed, 9 insertions(+), 3 deletions(-)