openstack/charm-neutron-openvswitch
Code Issues Proposed changes
charm-neutron-openvswitch/config.yaml
Frode Nordahl a88259a768 SR-IOV: match on PCI address, don't do runtime config 
Replace in-charm SR-IOV code with the common ``SRIOVContext``

Do not do run-time configuration of SR-IOV or hardware adaption
for hardware offload. In addition to being detrimental to any
virtual machine instance consuming the VF this will break NIC
firmware in some configurations.

The task is delegated to the installed packages and their systemd
services and configuration will occur at system bootup time.

We may consider adding an action to perform the configuration at
run-time if the operator really wants to, but it is very
complicated to get right. For example if you are using bonding
and hardware offload the virtual functions and hardware specific
setup has to happen _BEFORE_ netplan applies network configuration
to the system.

Closes-Bug: #1908351
Change-Id: Id0b81848658a3bd34470440bd68928ae9f6682e4
2021-02-16 11:45:53 +01:00

440 lines
18 KiB
YAML
Raw Blame History

options:
debug:
type: boolean
default: False
description: Enable debug logging.
verbose:
type: boolean
default: False
description: Enable verbose logging.
use-syslog:
type: boolean
default: False
description: |
Setting this to True will allow supporting services to log to syslog.
rabbit-user:
type: string
default: neutron
description: Username used to access RabbitMQ queue
rabbit-vhost:
type: string
default: openstack
description: RabbitMQ vhost
data-port:
type: string
default:
description: |
Space-delimited list of bridge:port mappings. Ports will be added to
their corresponding bridge. The bridges will allow usage of flat or
VLAN network types with Neutron and should match this defined in
bridge-mappings.
.
Ports provided can be the name or MAC address of the interface to be
added to the bridge. If MAC addresses are used, you may provide multiple
bridge:mac for the same bridge so as to be able to configure multiple
units. In this case the charm will run through the provided MAC addresses
for each bridge until it finds one it can resolve to an interface name.
Port can also be a linuxbridge bridge. In this case a veth pair will be
created, the ovs bridge and the linuxbridge bridge will be connected. It
can be useful to connect the ovs bridge to juju bridge.
dpdk-bond-mappings:
type: string
default:
description: |
Space-delimited list of bond:port mappings. The DPDK assigned ports will
be added to their corresponding bond, which in turn will be put into the
bridge as specified in data-port.
.
This option is supported only when enable-dpdk is true.
dpdk-bond-config:
type: string
default: ":balance-tcp:active:fast"
description: |
Space delimited list of bond:mode:lacp:lacp-time, where the arguments meaning is:
.
* bond - the bond name. If not specified the configuration applies to all bonds
* mode - the bond mode of operation. Possible values are:
- active-backup - No load balancing is offered in this mode and only one of
the member ports is active/used at a time.
- balance-slb - Considered as a static load-balancing mode. Traffic is load
balanced between member ports based on the source MAC and VLAN.
- balance-tcp - This is the preferred bonding mode. It offers traffic load
balancing based on 5-tuple header fields. LACP must be enabled
at both endpoints to use this mode. The aggregate link will
fall back to default mode (active-passive) in the event of LACP
negotiation failure.
* lacp - active, passive or off
* lacp-time - fast or slow. LACP negotiation time interval - 30 ms or 1 second
disable-security-groups:
type: boolean
default: false
description: |
Disable neutron based security groups - setting this configuration option
will override any settings configured via the neutron-api charm.
.
BE CAREFUL - this option allows you to disable all port level security
within an OpenStack cloud.
bridge-mappings:
type: string
default: 'physnet1:br-data'
description: |
Space-delimited list of ML2 data bridge mappings with format
<provider>:<bridge>.
flat-network-providers:
type: string
default:
description: |
Space-delimited list of Neutron flat network providers.
vlan-ranges:
type: string
default: "physnet1:1000:2000"
description: |
Space-delimited list of <physical_network>:<vlan_min>:<vlan_max> or
<physical_network> specifying physical_network names usable for VLAN
provider and tenant networks, as well as ranges of VLAN tags on each
available for allocation to tenant networks.
firewall-driver:
type: string
default:
description: |
Firewall driver to use to support use of security groups with
instances; valid values include iptables_hybrid (default) and
openvswitch (>= Mitaka on Ubuntu 16.04 or later).
ext-port:
type: string
default:
description: |
Deprecated: Use bridge-mappings and data-port to create a network
which can be used for external connectivity. You can call the network
external and the bridge br-ex by convention, but neither is required
A space-separated list of external ports to use for routing of instance
traffic to the external public network. Valid values are either MAC
addresses (in which case only MAC addresses for interfaces without an IP
address already assigned will be used), or interfaces (eth0)
enable-local-dhcp-and-metadata:
type: boolean
default: false
description: |
Enable local Neutron DHCP and Metadata Agents. This is useful for
deployments which do not include a neutron-gateway (do not require l3,
lbaas or vpnaas services) and should only be used in-conjunction with
flat or VLAN provider networks configurations.
NOTE: This configuration option will be ignored when deployed in a LXD
container.
dnsmasq-flags:
type: string
default:
description: |
Comma-separated list of key=value config flags with the additional dhcp
options for neutron dnsmasq. Note, this option is only valid when
enable-local-dhcp-and-metadata option is set to True.
NOTE: This configuration option will be ignored when deployed in a LXD
container.
instance-mtu:
type: int
default:
description: |
Configure DHCP services to provide MTU configuration to instances
within the cloud. This is useful in deployments where its not
possible to increase MTU on switches and physical servers to
accommodate the packet overhead of using GRE tunnels.
NOTE: This configuration option will be ignored when deployed in a LXD
container.
dns-servers:
type: string
default:
description: |
A comma-separated list of DNS servers which will be used by dnsmasq as
forwarders. This option only applies when the enable-local-dhcp-and-metadata
options is set to True.
NOTE: This configuration option will be ignored when deployed in a LXD
container.
prevent-arp-spoofing:
type: boolean
default: true
description: |
Enable suppression of ARP responses that don't match an IP address that
belongs to the port from which they originate.
.
Only supported in OpenStack Liberty or newer, which has the required
minimum version of Open vSwitch.
.
NOTE: this feature is deprecated and removed in Openstack >= Ocata. As of
that point the only way to disable protection will be via the port
security extension (see LP 1691080 for info).
enable-dpdk:
type: boolean
default: false
description: |
Enable DPDK fast userspace networking; this requires use of DPDK
supported network interface drivers and must be used in conjunction with
the data-port configuration option to configure each bridge with an
appropriate DPDK enabled network device.
dpdk-socket-memory:
type: int
default: 1024
description: |
Amount of hugepage memory in MB to allocate per NUMA socket in deployed
systems.
.
Only used when DPDK is enabled.
dpdk-socket-cores:
type: int
default: 1
description: |
Number of cores to allocate to DPDK per NUMA socket in deployed systems.
.
Only used when DPDK is enabled.
dpdk-driver:
type: string
default:
description: |
Kernel userspace device driver to use for DPDK devices, valid values
include:
.
vfio-pci
uio_pci_generic
.
Only used when DPDK is enabled.
enable-sriov:
type: boolean
default: false
description: |
Enable SR-IOV NIC agent on deployed units; use with sriov-device-mappings
to map SR-IOV devices to underlying provider networks. Enabling this
option allows instances to be plugged into directly into SR-IOV VF
devices connected to underlying provider networks alongside the default
Open vSwitch networking options.
.
NOTE: This configuration option will be ignored on Trusty and older.
.
NOTE: Changing this value will have no effect on runtime configuration. A
manual restart of the `sriov-netplan-shim` service or reboot of the
system is required to apply configuration.
sriov-device-mappings:
type: string
default:
description: |
Space-delimited list of SR-IOV device mappings with format
.
<provider>:<interface>
.
Multiple mappings can be provided, delimited by spaces.
.
NOTE: This configuration option will be ignored if enable-sriov is false.
sriov-numvfs:
type: string
default: auto
description: |
Number of VF's to configure each PF with; by default, each SR-IOV PF will
be configured with the maximum number of VF's it can support. In the case
sriov-device-mappings is set, only the devices in the mapping are configured.
Either use a single integer to apply the same VF configuration to all
detected SR-IOV devices or use a per-device configuration in the following
format
.
<device>:<numvfs>
.
Multiple devices can be configured by providing multi values delimited by
spaces.
.
NOTE: Changing this value will disrupt networking on all SR-IOV capable
interfaces for blanket configuration or listed interfaces when per-device
configuration is used.
enable-hardware-offload:
type: boolean
default: false
description: |
Enable support for hardware offload of flows from Open vSwitch to supported
network adapters. This requires use of OpenStack Stein or later and has
only been tested on Mellanox ConnectX 5 adapters.
.
Enabling this option will make use of the sriov-numvfs option to configure
the VF functions of the physical network adapters detected in each unit.
.
Enabling this option also requires that the firewall-driver option be
set to 'openvswitch'; this will allow security groups to be applied to
hardware offloaded ports (note that this feature is currently not
supported by OVS or the Linux kernel).
.
This option must not be enabled with either enable-sriov or enable-dpdk.
.
NOTE: Changing this value will not perform hardware specific adaption. A
manual restart of the hardware specific adaption service or reboot of the
system is required to apply configuration.
networking-tools-source:
type: string
default: ppa:openstack-charmers/networking-tools
description: |
Package archive source to use for utilities associated with configuring
SR-IOV VF's and switchdev mode in Mellanox network adapters.
.
This PPA can be mirrored for offline deployments.
.
NOTE: This configuration option will be ignored if enable-sriov and
enable-hardware-offload are both false.
worker-multiplier:
type: float
default:
description: |
The CPU core multiplier to use when configuring worker processes for
this services e.g. metadata-agent. By default, the number of workers for
each daemon is set to twice the number of CPU cores a service unit has.
When deployed in a LXD container, this default value will be capped to 4
workers unless this configuration option is set.
# Network config (by default all access is over 'private-address')
os-data-network:
type: string
default: ""
description: |
The IP address and netmask of the OpenStack Data network (e.g.,
192.168.0.0/24)
.
This network will be used for tenant network traffic in overlay
networks.
.
In order to support service zones spanning multiple network
segments, a space-delimited list of a.b.c.d/x can be provided
The address of the first network found to have an address
configured will be used.
ipfix-target:
type: string
default:
description: |
IPFIX target wit the format "IP_Address:Port". This will enable IPFIX
exporting on all OVS bridges to the target, including br-int and br-ext.
security-group-log-output-base:
type: string
default:
description: |
This option allows setting a path for Network Security Group logs.
A valid file system path must be provided. If this option is not
provided Neutron will use syslog as a destination.
(Available from Queens)
security-group-log-rate-limit:
type: int
default:
description: |
Log entries are queued for writing to a log file when a packet rate
exceeds the limit set by this option.
Possible values: null (no rate limitation), integer values greater than 100.
WARNING: Should be NOT LESS than 100, if set
(or, if null, logging will log unlimited.)
security-group-log-burst-limit:
type: int
default: 25
description: |
This option sets the maximum queue size for log entries.
Can be used to avoid excessive memory consumption.
WARNING: Should be NOT LESS than 25.
use-dvr-snat:
type: boolean
default: False
description: |
This option controls a mode of the l3 agent when DVR is used. There are
2 modes 'dvr' (default) and dvr_snat (gateway mode). Neutron server
(deployed by neutron-api charm) will schedule a network:router_centralized_snat port
and a (centralized) snat namespace to dvr_snat mode agents only. If this option
is enabled, all neutron-openvswitch nodes become candidates for being centralized
snat nodes. If l3ha is enabled on neutron-api, relevant packages are also installed
on every unit making them capable of hosting parts of an L3HA router. The min and max
numbers of L3 agents per router need to be taken into account in this case
(see max_l3_agents_per_router and min_l3_agents_per_router Neutron options).
Practically, this option can be used to allow DVR routers (L3HA or not) to
be scheduled without a requirement for a dedicated network node to host
centralized SNAT. This is especially important if only floating IPs are
used in the network design and SNAT traffic is minimal or non-existent.
NOTE: This configuration option will be ignored when deployed in a LXD
container.
sysctl:
type: string
default: |
{ net.ipv4.neigh.default.gc_thresh1 : 128,
net.ipv4.neigh.default.gc_thresh2 : 28672,
net.ipv4.neigh.default.gc_thresh3 : 32768,
net.ipv6.neigh.default.gc_thresh1 : 128,
net.ipv6.neigh.default.gc_thresh2 : 28672,
net.ipv6.neigh.default.gc_thresh3 : 32768,
net.nf_conntrack_max : 1000000,
net.netfilter.nf_conntrack_buckets : 204800,
net.netfilter.nf_conntrack_max : 1000000 }
description: |
YAML-formatted associative array of sysctl key/value pairs to be set
persistently e.g. '{ kernel.pid_max : 4194303 }'.
firewall-group-log-output-base:
type: string
default:
description: |
This option allows setting a path for Firewall Group logs.
A valid file system path must be provided. If this option is not
provided Neutron will use syslog as a destination.
(Available from Stein)
firewall-group-log-rate-limit:
type: int
default:
description: |
Log entries are queued for writing to a log file when a packet rate
exceeds the limit set by this option.
Possible values: null (no rate limitation), integer values greater than 100.
WARNING: Should be NOT LESS than 100, if set (if null logging will not be
rate limited).
(Available from Stein)
firewall-group-log-burst-limit:
type: int
default: 25
description: |
This option sets the maximum queue size for log entries.
Can be used to avoid excessive memory consumption.
WARNING: Should be NOT LESS than 25.
(Available from Stein)
ovs-use-veth:
type: string
default:
description: |
"True" or "False" string value. It is safe to leave this option unset.
This option allows the DHCP agent to use a veth interface for OVS in
order to support kernels with limited namespace support. i.e. Trusty.
Changing the value after neutron DHCP agents are created will break
access. The charm will go into a blocked state if this is attempted.
keepalived-healthcheck-interval:
type: int
default: 30
description: |
By default all HA routers will check their external network gateway
by sending a ping and if that fails they trigger a vrrp transition. This
option defines how frequently this check is performed. Setting this value
to 0 will disable the healthchecks. Note that this only applies when
using l3ha and dvr_snat.
of-inactivity-probe:
type: int
default: 10
description: |
The inactivity_probe interval in seconds for the local switch connection
to the controller. A value of 0 disables inactivity probes. Used only
for 'native' driver. The default value is 10 seconds.
disable-mlockall:
type: boolean
default:
description: |
Disable Open vSwitch use of mlockall().
.
When mlockall() is enabled, all of ovs-vswitchd's process memory is
locked into physical RAM and prevented from paging. This avoids network
interruptions but can lead to memory exhaustion in memory-constrained
environments.
.
By default, the charm will disable mlockall() if it is running in a
container. Otherwise, the charm will default to mlockall() enabled if
it is not running in a container.
.
Changing this config option will restart openvswitch-switch, resulting
in an expected data plane outage while the service restarts.
.
(Available from Mitaka)
View Git Blame Copy Permalink