Age | Commit message (Collapse) | Author | Files | Lines |
|
We're seeing extra IGMP traffic which interferes with tests which must
be disabled. This is done by disabling the setting by default on all
interfaces.
The default setting is also adopted for other traffic-related
configuration:
Default sets the configuration when the interface is created. For
hardware interfaces, that is at boot time. For VF interfaces, when we
create them with the csit-initialize-vfs.sh script and for virtual
interfaces when they're created.
On the other hand, the all configuration option doesn't set the
configuration. It uses the current configuration value along with the
passed configuration value to produce a new value, which could be
expressed as:
f(old_value, value_from_all) -> new_value
Where f is a function specific to each configuration option. Using all
thus requires us to know the function and the old value which is less
reliable than just using default everywhere.
Also fix the 1n-alt VF init script.
Change-Id: I3da15abfb5a1e76769076edd95d20e2168b6f71f
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: Iea69bcd730f641504428f1400e5c6eb6d6b755dd
|
|
Signed-off-by: Peter Mikus <peter.mikus@protonmail.ch>
Change-Id: I3d2bb0408f15fd14989d1179d7f08710a3d86e67
|
|
Signed-off-by: pmikus <peter.mikus@protonmail.ch>
Change-Id: Id90ca8323e6f49b51f19526e1089bf711e4e8182
|
|
Arm VPP Device requires newer drivers, otherwise we're frequently seeing
a deadlock when running csit-initialize-vfs.
We've observed server slowdown in VPP Device jobs in the past and
testing revealed that disabling the cppc_cpufreq seemingly resolves the
issue (there was no observed slowdown with a sample of 500 job runs).
Also update Nomad config (server IPs).
Change-Id: I484417f2cdb701239ca7517406147d0f2f7adde4
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
|
|
Signed-off-by: Peter Mikus <pmikus@cisco.com>
Change-Id: I298625dc75cdc6ffc168484e19ca30f02c2f46a0
|
|
We're not using unsafe noiommu on perf testbeds and we should mirror
that config in vpp device.
As a side note, on arm vpp device, enabling unsafe noiommu leads to
corrupted packets, so it must be disabled at least on arm vpp device.
Also update arm vpp device kernel parameters to match perf testbeds.
Change-Id: Id3d3e8fcf71583cc04887b4180cd20561856d902
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
|
|
Signed-off-by: pmikus <pmikus@cisco.com>
Change-Id: If00b376bd78b7b603d7887c1e149e7dccdb6d217
|
|
VFs need /sys/module/vfio/parameters/enable_unsafe_noiommu_mode to be
enabled on the host. Add this to Ansible.
Change-Id: Iee1d0ccfdb4ea9b9d6981d39490fe7feb33bf70c
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
|
|
* move iavf driver install from vpp_device to intel role
* differentiate the names of tasks of various drivers to get better
ansible logs
* add .5 sleep when creating VFs with iavf to avoid improper init
* fix install commands
* add XL710QDA2 NIC to supported i40e NICs
Change-Id: Id50e1552a322b9473e58222bfed3405d2d73c824
Signed-off-by: Juraj Linkeš <juraj.linkes@pantheon.tech>
|
|
Signed-off-by: pmikus <pmikus@cisco.com>
Change-Id: I902439bd078886e30bf0990647796038c36ed598
|
|
Signed-off-by: pmikus <pmikus@cisco.com>
Change-Id: If24c385b93e1ba310ea258984c97494b4cab467c
|
|
+ Better accessibility
+ Compliant with fdio.infra._function_
- function [pxe|terraform|ansible|vagrant]
+ dill==0.3.3 also applied on TBs
- ci-man to follow today
- Docs to be updated in separate patch
Signed-off-by: pmikus <pmikus@cisco.com>
Change-Id: Iff9eaa29d63044188cc8160db2d9b44b7635782a
|