Age | Commit message (Collapse) | Author | Files | Lines |
|
|
|
|
|
When using mapping file with mac address, mapping logic will try
to fetch the mac address of the interface, which will fail. Storing
the mac address also in the DPDK mapping file so that we can satisfy
mapping logic.
Closes-Bug: #1619330
Change-Id: I92ba7f589c8d848feb083f07c3f937b50aca388e
|
|
* os-net-config is called multiple times during the deploy.
Once the interface is bound to a driver, it will not be
listed for ethtool to get the pci address, which will
through exception. Handled this exception.
* Stored the DPDK bound nic configs at '/var/lib/os-net-config/
dpdk_mappings.yaml' file to emulate the same nic numbering
after the nic has been bound to the DPDK driver.
Partial-Bug: #1619330
Change-Id: I6b1e45003f851f1fcf5b8730890c75331e8d0f8f
|
|
- refactor test_object unit-test for IVS to follow other object class tests
- add new CLI unit test for IVS
- cleanup comments from https://review.openstack.org/#/c/345599
Change-Id: I0c337b019ae90ee7b2f207ff9b5060ed67bf182f
|
|
Modified the sample code to use numbered nics instead of physical
device name and added test code for testing the numbered nic usecases.
Implements: blueprint tripleo-ovs-dpdk
Depends-On: If1c91402d2d393140dc1b4a678e68a1bcdbe81e4
Change-Id: Ifadb495be57fcef56a97250de0c52fd03f2dd817
|
|
Add functionality to os-net-config to allow DPDK bonding of
interfaces, and implement support for parameters to be passed
by TripleO Heat Templates.
Implements: blueprint tripleo-ovs-dpdk
Depends-On: Id4a23ced28b92a642c180a35c55080e5f4e2e05d
Change-Id: If1c91402d2d393140dc1b4a678e68a1bcdbe81e4
|
|
Add support in os-net-config for DPDK ports and OVS user bridges,
and implement parameters which will be set by the TripleO Heat
Templates when using TripleO.
Implements: blueprint tripleo-ovs-dpdk
Change-Id: Id4a23ced28b92a642c180a35c55080e5f4e2e05d
|
|
The original implementation supported a strictly numeric mapping scheme
(e.g. nic1, nic2) that could misbehave if an active NIC was not listed in the
user's mapping file. This change fixes the misbehavior, and enhances the
feature by not requiring NIC aliases follow the numeric mapping scheme. This
allows the user to choose meaningful names for the NIC aliases.
NIC mapping now happens in two steps:
1) Process any user supplied mappings
- NIC alias does not need to follow the numeric "nicN" scheme
- Existing validation rules apply: mappings for inactive NICs are ignored
(but logged), and duplicate mappings are rejected
2) Generate default mappings as needed
- Only applies to active NICs that aren't already mapped
- Follows the numeric scheme (nic1, nic2) using the NIC number based on
the list of active NICs
- No default mapping is assigned if another NIC is already using that
numeric alias
Change-Id: I6943623a51702349f6a7dcf2de8a8429078a3ab0
Closes-Bug: 1612723
|
|
|
|
The latest version of os-net-config is not reliably enabling bonding
when there are no existing bonds on the system. This is due to an
order-of-operations problem where it tries to bring up the bond
before the slave interfaces. This can result in an error because
the bonding module is not present.
This change activates the bond only after the interfaces have been
activated. This should ensure that the bonding module is loaded.
Change-Id: I62e4b10a88168948ed59170285223fb4f8d8de88
Fixes-bug: 1611471
|
|
These changes are to generate /etc/sysconf/network-scripts/ifcfg-* and
/etc/sysconfig/nfvswitch configuration files for nfvswitch and its interfaces.
NFVSwitch is a virtual switch implementation based on DPDK for datacenter
workloads with very high throughput needs.
Change-Id: If02edb9c4c54c014f67290fe0c34e2fc73cb95bd
|
|
This change adds support for Linux adapter teams using teamd to
manage the bonds instead of the kernel bonding module. Adapter
teams using teamd can act like bonds, but also support additional
features and possibly more robust fault tolerance.
This implementation is fairly straightforward, in order to maintain
backward compatibility with templates made for Linux bonds. The only
difference in the syntax between the two is type: team instead of
type: linux_bond, and the bonding_options format is different.
The configuration files for teams should contain the team options
as a JSON string. The options that can be used are documented in
the teamd.conf(5) man page.
If an interface is marked as primary, the priority will be changed
from default 0 to 100, making this interface the preferred one. In
addition, the MAC address of the Team and all member interfaces will
be set to that of the primary interface. At this time, there is no
way to set the priority of link members individually, only the
interface marked primary will have a non-default priority.
This change has been tested on bare metal and worked for a team
with two bonded interfaces using LACP. The team was part of an
OVS bridge, and there was a VLAN interface on the team. Everything
worked as expected. Unit tests are included and passing.
Change-Id: If1d516ce8f9ada76375c3a52c5557d3f7348981a
Implements: blueprint os-net-config-teaming
|
|
|
|
|
|
This patch adds support for Infiniband interfaces. The only difference
between Inifiniband and regular interfaces at this time is that an
interface with type "ib_interface" will have "TYPE=Infiniband" added
to the ifcfg file.
However, the Infiniband interface is implemented as a full new class,
so in the future we can add script functions or additional config
options to the Infiniband interface config if needed.
Unit tests for both the object and the ifcfg implementation are
included. This patch does not include an implementation for systems
that use /etc/network/interfaces (Debian-based systems).
Note that this change has not yet been tested on bare metal with
Infiniband hardware.
Fixes bug: https://bugzilla.redhat.com/show_bug.cgi?id=1326616
Change-Id: Iaeaca9cd71e2cea6147951d49aecc7458be4ca0b
|
|
|
|
This small change fixes a serious bug in os-net-config which made
mapping NICs by MAC address not idempotent if Linux bonds were used.
The first time os-net-config runs, it sees the MAC address of each
interface, then it sets up the bond. On subsequent runs, any
non-primary interfaces have taken the MAC address of the primary
interface in the bond. At this time, they all show the same MAC
address. This causes the mapping to fail, and it appears that
the system has fewer matching interfaces than it had before.
The impact of this bug is that bonding doesn't work for
anyone doing mapping by MAC address. This is known to be affecting
installations with a particular 3rd-party Neutron plugin vendor.
This change simply looks for the permanent hardware address under
the bonding_slave directory, if it exists.
Change-Id: I5b0087370f74ecc319d2285b0f9f5f3dd951dbc2
|
|
OVS patch ports are used to connect two OVS bridges so traffic can flow
between them. This is generally useful for various cases.
Specifically it could be used to connect the bridges created by TripleO
networking configurations (br-ex, etc) to the bridge created by the
multinode networking setup in infra's nodepool (br_pub). This allows the
nodes in a multinode deployment to have connectivity across private
subnets where such traffic is typically firewalled off in public clouds.
Change-Id: I11404106cb3f53734f6fc9a35c22f905a0770245
|
|
|
|
|
|
Occassionally I've found myself debugging os-net-config's behavior
related to detecting active nics. This adds a little debugging output to
help see what's going on in the ordered_active_nics function.
Change-Id: If7ea010071d2253b29aaaabb242690ea5fbfb165
|
|
|
|
Adds support for configuring OVS Tunnels via os-net-config. Tunnels are
configured as members of ovs_bridge's where the type is set to
ovs_tunnel. The object also supports setting OVS extra and options so
that additional tunnel data can be defined, such as remote_ip.
Change-Id: I31ac1cbe8a13247a1529c0f99a0aea5807888844
|
|
When there are no active nics, strange errors can happen later
in the process of applying the configuration, and it's often not
obvious what caused them. Logging a warning should make it
easier to track down such problems.
Note that this should never happen legitimately in a TripleO
environment since we always need to have at least one active nic
to even get configuration to the system. However, it is a valid
case for someone who might be applying a configuration with local
access to a system, so it should be handled sanely. It's also
helpful in case of future bugs in the active nic checking.
Change-Id: Iaf6d4b1b215b70d61e0857e093a834702829e1b9
Related-Bug: 1569403
|
|
The previous check for an interface operstate only looked for the
literal "UP" in caps, but in my environments I'm seeing operstate
returned as "up" in lower-case, which causes the _is_active_nic
check to fail incorrectly. Example:
[root@overcloud-controller-0 heat-admin]# cat /sys/class/net/eth0/operstate
up
In this environment os-net-config is failing with the exception
in the linked bug.
Change-Id: I85c2d074ce43673c691523ca146ff7cdfdf1c7ca
Closes-Bug: 1569403
|
|
|
|
Fixes a couple of small types in a docstring and exception message.
Change-Id: Id6977ca50418f9fbdca2d1a26174637968b305a6
|
|
|
|
|
|
PEERDNS=no ensures resolv.conf is not modified by ifup/ifdown when
restarting an interface.
Closes-bug: #1567004
Change-Id: Idec15d4c1630638ce8362f9d95ae3c09bcae86a3
|
|
Bumps the version of hacking to match that in global-requirements.txt.
Change-Id: I7bb98dc773ce821b025d534f537f1a1704a50f9e
Closes-bug: #1566463
|
|
When configuring a linux bond with interfaces we have a bug
(due to a missing unit test) where the interfaces weren't
being configured with MASTER=<bond_name> SLAVE=yes. This
patch adds a unit test to check for the required configuration
and cleans up the logic in impl_ifcfg.py to handle it correctly
by simply checking for the linux_bond_name property which
is already set for us via objects.py.
See also: https://bugzilla.redhat.com/show_bug.cgi?id=1323717
Change-Id: Ic632c52265dcbbe6c0ace782e633a2030ad25e6f
Closes-bug: 1566428
|
|
Fixes lp bug 1555669
Currently _is_active_nic performs multiple checks to see if a link is live.
However some hardware such as Ethernet over USB devices fulfil this criteria.
This causes tripleo deployments to fail if automatic interface naming is used.
We can therefore use operstate to determine if the link is actually up.
https://www.kernel.org/doc/Documentation/networking/operstates.txt
This patch implements this.
Implements: better link state detection
Closes-Bug: #1555669
Change-Id: I0bdd0c987f4f177935df4f7cdcc70f4d99c988a5
Signed-off-by: Christopher Brown <snecklifter@gmail.com>
|
|
|
|
The current order of operations of bringing up interfaces starts
with the bridges, and restarts the members of the bridges. This
works well when an OVS bond is part of an OVS bridge, but does not
work well when Linux bonding is used with no bridges (such as in
the case of the Nuage plugin, which creates the bridges outside of
the ifcfg files).
This change will ensure that Linux bonds are brought up first, then
any bridges are brought up, interfaces are brought up next, and the
VLAN interfacs are brought up last. This corrects a race condition
where any VLAN which was brought up before the bond would not be
active, and the VLAN would have to be brought up by hand.
This change also fixes a logging issue where os-net-config would
report that no changes were necessary when changes were made. Now,
the logging messages indicating that "No changes were necessary"
show up when no changes are required, instead of when changes are
actually made.
Change-Id: I1efee3dfd8e8cef01b054bb57a3085cc7eb60372
|
|
|
|
|
|
This change adds the ability to write IPv6 routes to the system
NIC configuration. In the ifcfg provider, IPv6 routes must be
written to network-scripts/route6-<interface> rather than
route-interface. This patch checks the next hop, and if it is an
IPv6 address (assumed because of the presence of ":"), then the
route will be written to route6-<interface>. This patch includes
tests to confirm proper operation.
Change-Id: I3563800c774a6082a896c7508f5a92e25418fe63
|
|
This change cleans up some of the Linux Bond and Linux Bridge logic
that processes member interfaces and VLANs.
One bug is fixed, where a VLAN might be assigned as a slave
interface on a bond.
Changes were made so that if VLANs are placed under a Linux Bond,
the PHYSDEV is set to the bond, so the device: does not need to
be specified in the config.
This change ensures that if a bridge has a bond and VLANs as members,
that the VLAN will be a member of the bridge and not the bond.
Change-Id: Id329737f0032e781b111741086ded83f378297a7
|
|
This change generates /etc/sysconf/network-scripts/ifcfg-* for ivs.
It also generates /etc/sysconf/ivs configuration file for ivs.
It supports only RedHat at this point.
Indigo Virtual Switch (IVS, https://github.com/floodlight/ivs)
is a virtual switch for Linux. It is compatible with the KVM
hypervisor and leveraging the Open vSwitch kernel module for
packet forwarding. There are three major differences between
IVS and OVS:
1. Each node can have at most one ivs, name is not required.
2. Bond is not allowed to attach to an ivs. It is the SDN
controller's job to dynamically form bonds on ivs.
3. IP address can only be statically assigned.
Change-Id: I276d736794d123405de793c2a4eb2c1ee55a0fad
|
|
|
|
There are times when it is desirable to set the MTU of an
"interface" to 1500 -- when the external network and the storage
network are different VLANs on the same physical interface, for
example.
Change-Id: Ic5ea6ad05118fd57c26c898a460c19969de4109d
Closes-Bug: #1539821
|
|
The detailed bug description is at
https://bugs.launchpad.net/os-net-config/+bug/1537330
Since kilo, os-net-config starts to support linux bond.
However, the addr_assign_type value of an active member
interface of a linux bond is 3, but not 0. The consequence
of this problem is that the linux bond member links are
never considered as active, this further messes up the
nicX to interface name mapping.
This change checks if device directory exists, instead
of checking the addr_assign_type value. This directory
exists only when the interface is an actual physical
interface.
Change-Id: I8e4c95a2efa809fd236b07cbce1b81a2e774f858
|
|
The usage of assertEqual(True/False, ***) should be changed
to a meaningful format of assertTrue/False(***).
Change-Id: Ic15b6ebff7f050c1d516d9d680f362609803da4c
Closes-Bug:#1512207
|
|
Add the VLAN=yes parameter if the interface name is in the foo.123 format.
This is the behavior and Debian allow the same os-net-config configuration
file to be consistent with Red Hat systems.
See also: https://bugzilla.redhat.com/show_bug.cgi?id=1283812
Change-Id: Iab2a4861e2231b34899c379963abbef59cf73328
Closes-Bug: #1518449
|
|
|
|
|
|
This way device eth10 is after eth9, and not directly after eth1
Change-Id: I81eba97cccf6c4f314f9037d16ee1f244dbade02
Closes-Bug: #1482818
|
|
|