Age | Commit message (Collapse) | Author | Files | Lines |
|
Interface options is needed for nosdn-fdio scenario to disable
vlan strip offloading for enic.
Change-Id: I7d7e3d7e6e0c822a13113911a41a3d076ecbd509
Signed-off-by: Feng Pan <fpan@redhat.com>
|
|
This fixes nosdn OVS DPDK scenarios to work with OVS 2.6. It also adds
support for the ODL DPDK scenario. OVS DPDK configuration is done
during preconfig on compute nodes. All nodes make use of first boot
script to get kernel params like hugepages. The network settings are
modified to use real nic names for the virtual network settings file.
This is required due to a timing bug with os-net-config and using
first-boot script. The network settings file also contains the UIO
driver to use with DPDK.
JIRA: APEX-314
opnfv-tht-pr: 117
Change-Id: I46b6fca71ecec38981968133ba2411f64d7c6445
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
This patch adds support for ODL L3 FDIO scenario.
Introduces a new deploy settings parameter, odl_routing_node, to
indicate the routing node's hostname. This parameter will only
be used for this scenario.
Change-Id: I8df3eb701299007761c6cec42c4ca318b124a0b8
opnfv-tht-pr: 108
os-net-config-pr: 4
Signed-off-by: Feng Pan <fpan@redhat.com>
|
|
Changes:
- Kernel parameters are now set through first-boot.yaml during deployment. No custom
images will be built. Note that all nodes will be configured the same way, we only
use compute's kernel parameter settings currently (from deploy settings file)
- Add support for VPP interface type in network settings file, it is now possible
to specify vpp_interface type and uio_driver for tenant nic in network settings
file. A new example config, network_settings_vpp.yaml is added.
- Add support for odl_l2-fdio scenario.
Limitations:
- Physical NIC names must be specified in network settings file, numbered nic names
such as nic1 are not supported for fdio scenarios.
- The same kernel parameters will be configured for all nodes. The paramters will be
taken from compute kernel parameter section in deploy settings file.
opnfv-tht-pr: 104
opnfv-puppet-tripleo-pr: 12
os-net-config-pr: 3
Change-Id: Ie9d6151e6e58d11da3c66fbcabe4a0886c3fa152
Signed-off-by: Feng Pan <fpan@redhat.com>
|
|
Also contains various network parsing fixes.
JIRA: APEX-361
Change-Id: I5e40ed67267d5ded85adc2982a8726ebbceaeae5
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
- proposed in genesis
https://wiki.opnfv.org/display/genesis/Common+Network+Settings
- This patch updates the network_settings files in the config/network dir
and updates the associated bash, python and python tests relative to
the new yaml structure and nomenclature used in the proposed settings
file
JIRA: APEX-262
Change-Id: Ib0dc9d0f21465b38f2a53cff20f71fb9230bfc48
Signed-off-by: Dan Radez <dradez@redhat.com>
|
|
|
|
Currently there is no way to specify logically or physically the nic
order to be used on overcloud nodes. We always hardcode to use nic1 for
admin network, nic2 for private, etc. This patch allows a user to not
only decide which logical nics to use for which network, but also
specify physical interface names if they need to.
This is done on a per role basis, due to tripleO limitation. So a user
is able to specify nic order/names for compute and controller roles
separately.
If a user specifies nic order, they must specify it for all networks
other than admin network. We assume if admin network is unspecified it
uses "nic1", so that name is reserved in this case. A user is also
allowed to specify a mixture of logical and physical names, for example
"nic2" and "eth3" on another network.
JIRA: APEX-151
Change-Id: Ie9d4abb463cf8f8788913cb4bcf9486830adc449
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
JIRA: APEX-201
Change-Id: I3ba8d5e38e203bb4f23679a3b91de97c591de146
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Change-Id: Ibed808350163379fcfa5178c78fabe2f47e491fc
Signed-off-by: Dan Radez <dradez@redhat.com>
|
|
Issue where in default network_settings, api_network is not specified.
This means that the settings do not have api_network in its output, but
the nic template checks to see if api_network vlan is native in its
logic. This patch assumes a vlan will be numeric. A blank value or a
value of 'native' will not be numeric therefore default is expanded to
allow blank values as non-vlan networks.
Also fixes controller to always have br-ex.
Change-Id: Ibebb231901e653127175aef194eec708e8507da2
Signed-off-by: Tim Rozet <trozet@redhat.com>
Signed-off-by: Dan Radez <dradez@redhat.com>
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Changes include:
- Fixes nova schedulers to be an array instead of string
- Fixes os-net-config to only run one time with dpdk nodes
- Fixes br-phy creation to only happen on compute nodes
opnfv-tht-pr: 32
Change-Id: Ic3ef9ba247535ed00e40fe3fc4f0e78ac7d01c1a
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
Moves the (to be) dpdk interface to br-phy so that the ping check passes
after os-net-config. The puppet then moves the interface off of br-phy
before binding to dpdk.
opnfv-tht-pr: 31
Change-Id: I803671831ea5fd2665a7fa1368a7175d3e0a845d
Signed-off-by: Tim Rozet <trozet@redhat.com>
|
|
JIRA: APEX-136
JIRA: APEX-30
Change-Id: I7cbbe11f4c1f1455dba253733c78fe8dc215de97
Signed-off-by: Dan Radez <dradez@redhat.com>
|