The ovs-ofctl program is a command line tool for monitoring and
administering OpenFlow switches. It can also show the current
state of an OpenFlow switch, including features, configuration,
and table entries. It should work with any OpenFlow switch, not
just Open vSwitch.
OpenFlow Switch Management Commands
These commands allow ovs-ofctl to monitor and administer an
OpenFlow switch. It is able to show the current state of a
switch, including features, configuration, and table entries.
Most of these commands take an argument that specifies the method
for connecting to an OpenFlow switch. The following connection
methods are supported:
ssl:host[:port]
tcp:host[:port]
The specified port on the given host, which can be
expressed either as a DNS name (if built with
unbound library) or an IP address in IPv4 or IPv6
address format. Wrap IPv6 addresses in square
brackets, e.g. tcp:[::1]:6653. On Linux, use
%device to designate a scope for IPv6 link-level
addresses, e.g. tcp:[fe80::1234%eth0]:6653. For
ssl, the --private-key, --certificate, and
--ca-cert options are mandatory.
If port is not specified, it defaults to 6653.
unix:file
On POSIX, a Unix domain server socket named file.
On Windows, connect to a local named pipe that is
represented by a file created in the path file to
mimic the behavior of a Unix domain socket.
file This is short for unix:file, as long as file does
not contain a colon.
bridge This is short for
unix:/usr/local/var/run/openvswitch/bridge.mgmt, as
long as bridge does not contain a colon.
[type@]dp
Attempts to look up the bridge associated with dp
and open as above. If type is given, it specifies
the datapath provider of dp, otherwise the default
provider system is assumed.
show switch
Prints to the console information on switch, including
information on its flow tables and ports.
dump-tables switch
Prints to the console statistics for each of the flow
tables used by switch.
dump-table-features switch
Prints to the console features for each of the flow tables
used by switch.
dump-table-desc switch
Prints to the console configuration for each of the flow
tables used by switch for OpenFlow 1.4+.
mod-table switch table setting
This command configures flow table settings in switch for
OpenFlow table table, which may be expressed as a number
or (unless --no-names is specified) a name.
The available settings depend on the OpenFlow version in
use. In OpenFlow 1.1 and 1.2 (which must be enabled with
the -O option) only, mod-table configures behavior when no
flow is found when a packet is looked up in a flow table.
The following setting values are available:
drop Drop the packet.
continue
Continue to the next table in the pipeline. (This
is how an OpenFlow 1.0 switch always handles
packets that do not match any flow, in tables other
than the last one.)
controller
Send to controller. (This is how an OpenFlow 1.0
switch always handles packets that do not match any
flow in the last table.)
In OpenFlow 1.3 and later (which must be enabled with the
-O option) and Open vSwitch 2.11 and later only, mod-table
can change the name of a table:
name:new-name
Changes the name of the table to new-name. Use an
empty new-name to clear the name. (This will be
ineffective if the name is set via the name column
in the Flow_Table table in the Open_vSwitch
database as described in ovs-vswitchd.conf.db(5).)
In OpenFlow 1.4 and later (which must be enabled with the
-O option) only, mod-table configures the behavior when a
controller attempts to add a flow to a flow table that is
full. The following setting values are available:
evict Delete some existing flow from the flow table,
according to the algorithm described for the
Flow_Table table in ovs-vswitchd.conf.db(5).
noevict
Refuse to add the new flow. (Eviction might still
be enabled through the overflow_policy column in
the Flow_Table table documented in
ovs-vswitchd.conf.db(5).)
vacancy:low,high
Enables sending vacancy events to controllers using
TABLE_STATUS messages, based on percentage
thresholds low and high.
novacancy
Disables vacancy events.
dump-ports switch [netdev]
Prints to the console statistics for network devices
associated with switch. If netdev is specified, only the
statistics associated with that device will be printed.
netdev can be an OpenFlow assigned port number or device
name, e.g. eth0.
dump-ports-desc switch [port]
Prints to the console detailed information about network
devices associated with switch. To dump only a specific
port, specify its number as port. Otherwise, if port is
omitted, or if it is specified as ANY, then all ports are
printed. This is a subset of the information provided by
the show command.
If the connection to switch negotiates OpenFlow 1.0, 1.2,
or 1.2, this command uses an OpenFlow extension only
implemented in Open vSwitch (version 1.7 and later).
Only OpenFlow 1.5 and later support dumping a specific
port. Earlier versions of OpenFlow always dump all ports.
mod-port switch port action
Modify characteristics of port port in switch. port may
be an OpenFlow port number or name (unless --no-names is
specified) or the keyword LOCAL (the preferred way to
refer to the OpenFlow local port). The action may be any
one of the following:
up
down Enable or disable the interface. This is
equivalent to ip link set up or ip link set down on
a Unix system.
stp
no-stp Enable or disable 802.1D spanning tree protocol
(STP) on the interface. OpenFlow implementations
that don't support STP will refuse to enable it.
receive
no-receive
receive-stp
no-receive-stp
Enable or disable OpenFlow processing of packets
received on this interface. When packet processing
is disabled, packets will be dropped instead of
being processed through the OpenFlow table. The
receive or no-receive setting applies to all
packets except 802.1D spanning tree packets, which
are separately controlled by receive-stp or
no-receive-stp.
forward
no-forward
Allow or disallow forwarding of traffic to this
interface. By default, forwarding is enabled.
flood
no-flood
Controls whether an OpenFlow flood action will send
traffic out this interface. By default, flooding
is enabled. Disabling flooding is primarily useful
to prevent loops when a spanning tree protocol is
not in use.
packet-in
no-packet-in
Controls whether packets received on this interface
that do not match a flow table entry generate a
``packet in'' message to the OpenFlow controller.
By default, ``packet in'' messages are enabled.
The show command displays (among other information) the
configuration that mod-port changes.
get-frags switch
Prints switch's fragment handling mode. See set-frags,
below, for a description of each fragment handling mode.
The show command also prints the fragment handling mode
among its other output.
set-frags switch frag_mode
Configures switch's treatment of IPv4 and IPv6 fragments.
The choices for frag_mode are:
normal Fragments pass through the flow table like non-
fragmented packets. The TCP ports, UDP ports, and
ICMP type and code fields are always set to 0, even
for fragments where that information would
otherwise be available (fragments with offset 0).
This is the default fragment handling mode for an
OpenFlow switch.
drop Fragments are dropped without passing through the
flow table.
reassemble
The switch reassembles fragments into full IP
packets before passing them through the flow table.
Open vSwitch does not implement this fragment
handling mode.
nx-match
Fragments pass through the flow table like non-
fragmented packets. The TCP ports, UDP ports, and
ICMP type and code fields are available for
matching for fragments with offset 0, and set to 0
in fragments with nonzero offset. This mode is a
Nicira extension.
See the description of ip_frag, in ovs-fields(7), for a
way to match on whether a packet is a fragment and on its
fragment offset.
dump-flows switch [flows]
Prints to the console all flow entries in switch's tables
that match flows. If flows is omitted, all flows in the
switch are retrieved. See Flow Syntax, below, for the
syntax of flows. The output format is described in Table
Entry Output.
By default, ovs-ofctl prints flow entries in the same
order that the switch sends them, which is unlikely to be
intuitive or consistent. Use --sort and --rsort to
control display order. The --names/--no-names and
--stats/--no-stats options also affect output formatting.
See the descriptions of these options, under OPTIONS
below, for more information
dump-aggregate switch [flows]
Prints to the console aggregate statistics for flows in
switch's tables that match flows. If flows is omitted,
the statistics are aggregated across all flows in the
switch's flow tables. See Flow Syntax, below, for the
syntax of flows. The output format is described in Table
Entry Output.
queue-stats switch [port [queue]]
Prints to the console statistics for the specified queue
on port within switch. port can be an OpenFlow port
number or name, the keyword LOCAL (the preferred way to
refer to the OpenFlow local port), or the keyword ALL.
Either of port or queue or both may be omitted (or
equivalently the keyword ALL). If both are omitted,
statistics are printed for all queues on all ports. If
only queue is omitted, then statistics are printed for all
queues on port; if only port is omitted, then statistics
are printed for queue on every port where it exists.
queue-get-config switch [port [queue]]
Prints to the console the configuration of queue on port
in switch. If port is omitted or ANY, reports queues for
all port. If queue is omitted or ANY, reports all queues.
For OpenFlow 1.3 and earlier, the output always includes
all queues, ignoring queue if specified.
This command has limited usefulness, because ports often
have no configured queues and because the OpenFlow
protocol provides only very limited information about the
configuration of a queue.
dump-ipfix-bridge switch
Prints to the console the statistics of bridge IPFIX for
switch. If bridge IPFIX is configured on the switch,
IPFIX statistics can be retrieved. Otherwise, error
message will be printed.
This command uses an Open vSwitch extension that is only
in Open vSwitch 2.6 and later.
dump-ipfix-flow switch
Prints to the console the statistics of flow-based IPFIX
for switch. If flow-based IPFIX is configured on the
switch, statistics of all the collector set ids on the
switch will be printed. Otherwise, print error message.
Refer to ovs-vswitchd.conf.db(5) for more details on
configuring flow based IPFIX and collector set ids.
This command uses an Open vSwitch extension that is only
in Open vSwitch 2.6 and later.
ct-flush-zone switch zone
Flushes the connection tracking entries in zone on switch.
This command uses an Open vSwitch extension that is only
in Open vSwitch 2.6 and later.
OpenFlow Switch Flow Table Commands
These commands manage the flow table in an OpenFlow switch. In
each case, flow specifies a flow entry in the format described in
Flow Syntax, below, file is a text file that contains zero or
more flows in the same syntax, one per line, and the optional
--bundle option operates the command as a single atomic
transation, see option --bundle, below.
[--bundle] add-flow switch flow
[--bundle] add-flow switch - < file
[--bundle] add-flows switch file
Add each flow entry to switch's tables. Each flow
specification (e.g., each line in file) may start with
add, modify, delete, modify_strict, or delete_strict
keyword to specify whether a flow is to be added,
modified, or deleted, and whether the modify or delete is
strict or not. For backwards compatibility a flow
specification without one of these keywords is treated as
a flow add. All flow mods are executed in the order
specified.
[--bundle] [--strict] mod-flows switch flow
[--bundle] [--strict] mod-flows switch - < file
Modify the actions in entries from switch's tables that
match the specified flows. With --strict, wildcards are
not treated as active for matching purposes.
[--bundle] del-flows switch
[--bundle] [--strict] del-flows switch [flow]
[--bundle] [--strict] del-flows switch - < file
Deletes entries from switch's flow table. With only a
switch argument, deletes all flows. Otherwise, deletes
flow entries that match the specified flows. With
--strict, wildcards are not treated as active for matching
purposes.
[--bundle] [--readd] replace-flows switch file
Reads flow entries from file (or stdin if file is -) and
queries the flow table from switch. Then it fixes up any
differences, adding flows from flow that are missing on
switch, deleting flows from switch that are not in file,
and updating flows in switch whose actions, cookie, or
timeouts differ in file.
With --readd, ovs-ofctl adds all the flows from file, even
those that exist with the same actions, cookie, and
timeout in switch. In OpenFlow 1.0 and 1.1, re-adding a
flow always resets the flow's packet and byte counters to
0, and in OpenFlow 1.2 and later, it does so only if the
reset_counts flag is set.
diff-flows source1 source2
Reads flow entries from source1 and source2 and prints the
differences. A flow that is in source1 but not in source2
is printed preceded by a -, and a flow that is in source2
but not in source1 is printed preceded by a +. If a flow
exists in both source1 and source2 with different actions,
cookie, or timeouts, then both versions are printed
preceded by - and +, respectively.
source1 and source2 may each name a file or a switch. If
a name begins with / or ., then it is considered to be a
file name. A name that contains : is considered to be a
switch. Otherwise, it is a file if a file by that name
exists, a switch if not.
For this command, an exit status of 0 means that no
differences were found, 1 means that an error occurred,
and 2 means that some differences were found.
packet-out switch packet-out
Connects to switch and instructs it to execute the packet-
out OpenFlow message, specified as defined in Packet-Out
Syntax section.
Group Table Commands
These commands manage the group table in an OpenFlow switch. In
each case, group specifies a group entry in the format described
in Group Syntax, below, and file is a text file that contains
zero or more groups in the same syntax, one per line, and the
optional --bundle option operates the command as a single atomic
transation, see option --bundle, below.
The group commands work only with switches that support OpenFlow
1.1 or later or the Open vSwitch group extensions to OpenFlow 1.0
(added in Open vSwitch 2.9.90). For OpenFlow 1.1 or later, it is
necessary to explicitly enable these protocol versions in
ovs-ofctl (using -O). For more information, see ``Q: What
versions of OpenFlow does Open vSwitch support?'' in the Open
vSwitch FAQ.
[--bundle] add-group switch group
[--bundle] add-group switch - < file
[--bundle] add-groups switch file
Add each group entry to switch's tables. Each group
specification (e.g., each line in file) may start with
add, modify, add_or_mod, delete, insert_bucket, or
remove_bucket keyword to specify whether a flow is to be
added, modified, or deleted, or whether a group bucket is
to be added or removed. For backwards compatibility a
group specification without one of these keywords is
treated as a group add. All group mods are executed in
the order specified.
[--bundle] [--may-create] mod-group switch group
[--bundle] [--may-create] mod-group switch - < file
Modify the action buckets in entries from switch's tables
for each group entry. If a specified group does not
already exist, then without --may-create, this command has
no effect; with --may-create, it creates a new group. The
--may-create option uses an Open vSwitch extension to
OpenFlow only implemented in Open vSwitch 2.6 and later.
[--bundle] del-groups switch
[--bundle] del-groups switch [group]
[--bundle] del-groups switch - < file
Deletes entries from switch's group table. With only a
switch argument, deletes all groups. Otherwise, deletes
the group for each group entry.
[--bundle] insert-buckets switch group
[--bundle] insert-buckets switch - < file
Add buckets to an existing group present in the switch's
group table. If no command_bucket_id is present in the
group specification then all buckets of the group are
removed.
[--bundle] remove-buckets switch group
[--bundle] remove-buckets switch - < file
Remove buckets to an existing group present in the
switch's group table. If no command_bucket_id is present
in the group specification then all buckets of the group
are removed.
dump-groups switch [group]
Prints group entries in switch's tables to console. To
dump only a specific group, specify its number as group.
Otherwise, if group is omitted, or if it is specified as
ALL, then all groups are printed.
Only OpenFlow 1.5 and later support dumping a specific
group. Earlier versions of OpenFlow always dump all
groups.
dump-group-features switch
Prints to the console the group features of the switch.
dump-group-stats switch [group]
Prints to the console statistics for the specified group
in switch's tables. If group is omitted then statistics
for all groups are printed.
OpenFlow 1.3+ Switch Meter Table Commands
These commands manage the meter table in an OpenFlow switch. In
each case, meter specifies a meter entry in the format described
in Meter Syntax, below.
OpenFlow 1.3 introduced support for meters, so these commands
only work with switches that support OpenFlow 1.3 or later. It
is necessary to explicitly enable these protocol versions in
ovs-ofctl (using -O) and in the switch itself (with the protocols
column in the Bridge table). For more information, see ``Q: What
versions of OpenFlow does Open vSwitch support?'' in the Open
vSwitch FAQ.
add-meter switch meter
Add a meter entry to switch's tables. The meter syntax is
described in section Meter Syntax, below.
mod-meter switch meter
Modify an existing meter.
del-meters switch [meter]
Delete entries from switch's meter table. To delete only
a specific meter, specify its number as meter. Otherwise,
if meter is omitted, or if it is specified as all, then
all meters are deleted.
dump-meters switch [meter]
Print entries from switch's meter table. To print only a
specific meter, specify its number as meter. Otherwise,
if meter is omitted, or if it is specified as all, then
all meters are printed.
meter-stats switch [meter]
Print meter statistics. meter can specify a single meter
with syntax meter=id, or all meters with syntax meter=all.
meter-features switch
Print meter features.
OpenFlow Switch Bundle Command
Transactional updates to both flow and group tables can be made
with the bundle command. file is a text file that contains zero
or more flow mods, group mods, or packet-outs in Flow Syntax,
Group Syntax, or Packet-Out Syntax, each line preceded by flow,
group, or packet-out keyword, correspondingly. The flow keyword
may be optionally followed by one of the keywords add, modify,
modify_strict, delete, or delete_strict, of which the add is
assumed if a bare flow is given. Similarly, the group keyword
may be optionally followed by one of the keywords add, modify,
add_or_mod, delete, insert_bucket, or remove_bucket, of which the
add is assumed if a bare group is given.
bundle switch file
Execute all flow and group mods in file as a single atomic
transaction against switch's tables. All bundled mods are
executed in the order specified.
OpenFlow Switch Tunnel TLV Table Commands
Open vSwitch maintains a mapping table between tunnel option TLVs
(defined by <class, type, length>) and NXM fields tun_metadatan,
where n ranges from 0 to 63, that can be operated on for the
purposes of matches, actions, etc. This TLV table can be used for
Geneve option TLVs or other protocols with options in same TLV
format as Geneve options. This mapping must be explicitly
specified by the user through the following commands.
A TLV mapping is specified with the syntax
{class=class,type=type,len=length}->tun_metadatan. When an
option mapping exists for a given tun_metadatan, matching on the
defined field becomes possible, e.g.:
ovs-ofctl add-tlv-map br0
"{class=0xffff,type=0,len=4}->tun_metadata0"
ovs-ofctl add-flow br0
tun_metadata0=1234,actions=controller
A mapping should not be changed while it is in active use by a
flow. The result of doing so is undefined.
These commands are Nicira extensions to OpenFlow and require Open
vSwitch 2.5 or later.
add-tlv-map switch option[,option]...
Add each option to switch's tables. Duplicate fields are
rejected.
del-tlv-map switch [option[,option]]...
Delete each option from switch's table, or all option TLV
mapping if no option is specified. Fields that aren't
mapped are ignored.
dump-tlv-map switch
Show the currently mapped fields in the switch's option
table as well as switch capabilities.
OpenFlow Switch Monitoring Commands
snoop switch
Connects to switch and prints to the console all OpenFlow
messages received. Unlike other ovs-ofctl commands, if
switch is the name of a bridge, then the snoop command
connects to a Unix domain socket named
/usr/local/var/run/openvswitch/switch.snoop. ovs-vswitchd
listens on such a socket for each bridge and sends to it
all of the OpenFlow messages sent to or received from its
configured OpenFlow controller. Thus, this command can be
used to view OpenFlow protocol activity between a switch
and its controller.
When a switch has more than one controller configured,
only the traffic to and from a single controller is
output. If none of the controllers is configured as a
primary or a secondary (using a Nicira extension to
OpenFlow 1.0 or 1.1, or a standard request in OpenFlow 1.2
or later), then a controller is chosen arbitrarily among
them. If there is a primary controller, it is chosen;
otherwise, if there are any controllers that are not
primaries or secondaries, one is chosen arbitrarily;
otherwise, a secondary controller is chosen arbitrarily.
This choice is made once at connection time and does not
change as controllers reconfigure their roles.
If a switch has no controller configured, or if the
configured controller is disconnected, no traffic is sent,
so monitoring will not show any traffic.
monitor switch [miss-len] [invalid_ttl] [watch:[spec...]]
Connects to switch and prints to the console all OpenFlow
messages received. Usually, switch should specify the
name of a bridge in the ovs-vswitchd database. This is
available only in OpenFlow 1.0 as Nicira extension.
If miss-len is provided, ovs-ofctl sends an OpenFlow ``set
configuration'' message at connection setup time that
requests miss-len bytes of each packet that misses the
flow table. Open vSwitch does not send these and other
asynchronous messages to an ovs-ofctl monitor client
connection unless a nonzero value is specified on this
argument. (Thus, if miss-len is not specified, very
little traffic will ordinarily be printed.)
If invalid_ttl is passed, ovs-ofctl sends an OpenFlow
``set configuration'' message at connection setup time
that requests INVALID_TTL_TO_CONTROLLER, so that ovs-ofctl
monitor can receive ``packet-in'' messages when TTL
reaches zero on dec_ttl action. Only OpenFlow 1.1 and 1.2
support invalid_ttl; Open vSwitch also implements it for
OpenFlow 1.0 as an extension.
watch:[spec...] causes ovs-ofctl to send a ``monitor
request'' Nicira extension message to the switch at
connection setup time. This message causes the switch to
send information about flow table changes as they occur.
The following comma-separated spec syntax is available:
!initial
Do not report the switch's initial flow table
contents.
!add Do not report newly added flows.
!delete
Do not report deleted flows.
!modify
Do not report modifications to existing flows.
!own Abbreviate changes made to the flow table by
ovs-ofctl's own connection to the switch. (These
could only occur using the ofctl/send command
described below under RUNTIME MANAGEMENT COMMANDS.)
!actions
Do not report actions as part of flow updates.
table=table
Limits the monitoring to the table with the given
table, which may be expressed as a number between 0
and 254 or (unless --no-names is specified) a name.
By default, all tables are monitored.
out_port=port
If set, only flows that output to port are
monitored. The port may be an OpenFlow port number
or keyword (e.g. LOCAL).
field=value
Monitors only flows that have field specified as
the given value. Any syntax valid for matching on
dump-flows may be used.
This command may be useful for debugging switch or
controller implementations. With watch:, it is
particularly useful for observing how a controller updates
flow tables.
OpenFlow Switch and Controller Commands
The following commands, like those in the previous section, may
be applied to OpenFlow switches, using any of the connection
methods described in that section. Unlike those commands, these
may also be applied to OpenFlow controllers.
probe target
Sends a single OpenFlow echo-request message to target and
waits for the response. With the -t or --timeout option,
this command can test whether an OpenFlow switch or
controller is up and running.
ping target [n]
Sends a series of 10 echo request packets to target and
times each reply. The echo request packets consist of an
OpenFlow header plus n bytes (default: 64) of randomly
generated payload. This measures the latency of
individual requests.
benchmark target n count
Sends count echo request packets that each consist of an
OpenFlow header plus n bytes of payload and waits for each
response. Reports the total time required. This is a
measure of the maximum bandwidth to target for round-trips
of n-byte messages.
Other Commands
ofp-parse file
Reads file (or stdin if file is -) as a series of OpenFlow
messages in the binary format used on an OpenFlow
connection, and prints them to the console. This can be
useful for printing OpenFlow messages captured from a TCP
stream.
ofp-parse-pcap file [port...]
Reads file, which must be in the PCAP format used by
network capture tools such as tcpdump or wireshark,
extracts all the TCP streams for OpenFlow connections, and
prints the OpenFlow messages in those connections in
human-readable format on stdout.
OpenFlow connections are distinguished by TCP port number.
Non-OpenFlow packets are ignored. By default, data on TCP
ports 6633 and 6653 are considered to be OpenFlow.
Specify one or more port arguments to override the
default.
This command cannot usefully print SSL encrypted traffic.
It does not understand IPv6.
Flow Syntax
Some ovs-ofctl commands accept an argument that describes a flow
or flows. Such flow descriptions comprise a series of
field=value assignments, separated by commas or white space.
(Embedding spaces into a flow description normally requires
quoting to prevent the shell from breaking the description into
multiple arguments.)
Flow descriptions should be in normal form. This means that a
flow may only specify a value for an L3 field if it also
specifies a particular L2 protocol, and that a flow may only
specify an L4 field if it also specifies particular L2 and L3
protocol types. For example, if the L2 protocol type dl_type is
wildcarded, then L3 fields nw_src, nw_dst, and nw_proto must also
be wildcarded. Similarly, if dl_type or nw_proto (the L3
protocol type) is wildcarded, so must be the L4 fields tcp_dst
and tcp_src. ovs-ofctl will warn about flows not in normal form.
ovs-fields(7) describes the supported fields and how to match
them. In addition to match fields, commands that operate on
flows accept a few additional key-value pairs:
table=table
For flow dump commands, limits the flows dumped to those
in table, which may be expressed as a number between 0 and
255 or (unless --no-names is specified) a name. If not
specified (or if 255 is specified as table), then flows in
all tables are dumped.
For flow table modification commands, behavior varies
based on the OpenFlow version used to connect to the
switch:
OpenFlow 1.0
OpenFlow 1.0 does not support table for modifying
flows. ovs-ofctl will exit with an error if table
(other than table=255) is specified for a switch
that only supports OpenFlow 1.0.
In OpenFlow 1.0, the switch chooses the table into
which to insert a new flow. The Open vSwitch
software switch always chooses table 0. Other Open
vSwitch datapaths and other OpenFlow
implementations may choose different tables.
The OpenFlow 1.0 behavior in Open vSwitch for
modifying or removing flows depends on whether
--strict is used. Without --strict, the command
applies to matching flows in all tables. With
--strict, the command will operate on any single
matching flow in any table; it will do nothing if
there are matches in more than one table. (The
distinction between these behaviors only matters if
non-OpenFlow 1.0 commands were also used, because
OpenFlow 1.0 alone cannot add flows with the same
matching criteria to multiple tables.)
OpenFlow 1.0 with table_id extension
Open vSwitch implements an OpenFlow extension that
allows the controller to specify the table on which
to operate. ovs-ofctl automatically enables the
extension when table is specified and OpenFlow 1.0
is used. ovs-ofctl automatically detects whether
the switch supports the extension. As of this
writing, this extension is only known to be
implemented by Open vSwitch.
With this extension, ovs-ofctl operates on the
requested table when table is specified, and acts
as described for OpenFlow 1.0 above when no table
is specified (or for table=255).
OpenFlow 1.1
OpenFlow 1.1 requires flow table modification
commands to specify a table. When table is not
specified (or table=255 is specified), ovs-ofctl
defaults to table 0.
OpenFlow 1.2 and later
OpenFlow 1.2 and later allow flow deletion
commands, but not other flow table modification
commands, to operate on all flow tables, with the
behavior described above for OpenFlow 1.0.
duration=...
n_packet=...
n_bytes=...
ovs-ofctl ignores assignments to these ``fields'' to allow
output from the dump-flows command to be used as input for
other commands that parse flows.
The add-flow, add-flows, and mod-flows commands require an
additional field, which must be the final field specified:
actions=[action][,action...]
Specifies a comma-separated list of actions to take on a
packet when the flow entry matches. If no action is
specified, then packets matching the flow are dropped.
See ovs-actions(7) for details on the syntax and semantics
of actions. K
An opaque identifier called a cookie can be used as a handle to
identify a set of flows:
cookie=value
A cookie can be associated with a flow using the add-flow,
add-flows, and mod-flows commands. value can be any
64-bit number and need not be unique among flows. If this
field is omitted, a default cookie value of 0 is used.
cookie=value/mask
When using NXM, the cookie can be used as a handle for
querying, modifying, and deleting flows. value and mask
may be supplied for the del-flows, mod-flows, dump-flows,
and dump-aggregate commands to limit matching cookies. A
1-bit in mask indicates that the corresponding bit in
cookie must match exactly, and a 0-bit wildcards that bit.
A mask of -1 may be used to exactly match a cookie.
The mod-flows command can update the cookies of flows that
match a cookie by specifying the cookie field twice (once
with a mask for matching and once without to indicate the
new value):
ovs-ofctl mod-flows br0 cookie=1,actions=normal
Change all flows' cookies to 1 and change their
actions to normal.
ovs-ofctl mod-flows br0
cookie=1/-1,cookie=2,actions=normal
Update cookies with a value of 1 to 2 and change
their actions to normal.
The ability to match on cookies was added in Open vSwitch
1.5.0.
The following additional field sets the priority for flows added
by the add-flow and add-flows commands. For mod-flows and
del-flows when --strict is specified, priority must match along
with the rest of the flow specification. For mod-flows without
--strict, priority is only significant if the command creates a
new flow, that is, non-strict mod-flows does not match on
priority and will not change the priority of existing flows.
Other commands do not allow priority to be specified.
priority=value
The priority at which a wildcarded entry will match in
comparison to others. value is a number between 0 and
65535, inclusive. A higher value will match before a
lower one. An exact-match entry will always have priority
over an entry containing wildcards, so it has an implicit
priority value of 65535. When adding a flow, if the field
is not specified, the flow's priority will default to
32768.
OpenFlow leaves behavior undefined when two or more flows
with the same priority can match a single packet. Some
users expect ``sensible'' behavior, such as more specific
flows taking precedence over less specific flows, but
OpenFlow does not specify this and Open vSwitch does not
implement it. Users should therefore take care to use
priorities to ensure the behavior that they expect.
The add-flow, add-flows, and mod-flows commands support the
following additional options. These options affect only new
flows. Thus, for add-flow and add-flows, these options are
always significant, but for mod-flows they are significant only
if the command creates a new flow, that is, their values do not
update or affect existing flows.
idle_timeout=seconds
Causes the flow to expire after the given number of
seconds of inactivity. A value of 0 (the default)
prevents a flow from expiring due to inactivity.
hard_timeout=seconds
Causes the flow to expire after the given number of
seconds, regardless of activity. A value of 0 (the
default) gives the flow no hard expiration deadline.
importance=value
Sets the importance of a flow. The flow entry eviction
mechanism can use importance as a factor in deciding which
flow to evict. A value of 0 (the default) makes the flow
non-evictable on the basis of importance. Specify a value
between 0 and 65535.
Only OpenFlow 1.4 and later support importance.
send_flow_rem
Marks the flow with a flag that causes the switch to
generate a ``flow removed'' message and send it to
interested controllers when the flow later expires or is
removed.
check_overlap
Forces the switch to check that the flow match does not
overlap that of any different flow with the same priority
in the same table. (This check is expensive so it is best
to avoid it.)
reset_counts
When this flag is specified on a flow being added to a
switch, and the switch already has a flow with an
identical match, an OpenFlow 1.2 (or later) switch resets
the flow's packet and byte counters to 0. Without the
flag, the packet and byte counters are preserved.
OpenFlow 1.0 and 1.1 switches always reset counters in
this situation, as if reset_counts were always specified.
Open vSwitch 1.10 added support for reset_counts.
no_packet_counts
no_byte_counts
Adding these flags to a flow advises an OpenFlow 1.3 (or
later) switch that the controller does not need packet or
byte counters, respectively, for the flow. Some switch
implementations might achieve higher performance or reduce
resource consumption when these flags are used. These
flags provide no benefit to the Open vSwitch software
switch implementation.
OpenFlow 1.2 and earlier do not support these flags.
Open vSwitch 1.10 added support for no_packet_counts and
no_byte_counts.
The dump-flows, dump-aggregate, del-flow and del-flows commands
support these additional optional fields:
out_port=port
If set, a matching flow must include an output action to
port, which must be an OpenFlow port number or name (e.g.
local).
out_group=group
If set, a matching flow must include an group action
naming group, which must be an OpenFlow group number.
This field is supported in Open vSwitch 2.5 and later and
requires OpenFlow 1.1 or later.
Table Entry Output
The dump-tables and dump-aggregate commands print information
about the entries in a datapath's tables. Each line of output is
a flow entry as described in Flow Syntax, above, plus some
additional fields:
duration=secs
The time, in seconds, that the entry has been in the
table. secs includes as much precision as the switch
provides, possibly to nanosecond resolution.
n_packets
The number of packets that have matched the entry.
n_bytes
The total number of bytes from packets that have matched
the entry.
The following additional fields are included only if the switch
is Open vSwitch 1.6 or later and the NXM flow format is used to
dump the flow (see the description of the --flow-format option
below). The values of these additional fields are approximations
only and in particular idle_age will sometimes become nonzero
even for busy flows.
hard_age=secs
The integer number of seconds since the flow was added or
modified. hard_age is displayed only if it differs from
the integer part of duration. (This is separate from
duration because mod-flows restarts the hard_timeout timer
without zeroing duration.)
idle_age=secs
The integer number of seconds that have passed without any
packets passing through the flow.
Packet-Out Syntax
ovs-ofctl bundle command accepts packet-outs to be specified in
the bundle file. Each packet-out comprises of a series of
field=value assignments, separated by commas or white space.
(Embedding spaces into a packet-out description normally requires
quoting to prevent the shell from breaking the description into
multiple arguments.). Unless noted otherwise only the last
instance of each field is honoured. This same syntax is also
supported by the ovs-ofctl packet-out command.
in_port=port
The port number to be considered the in_port when
processing actions. This can be any valid OpenFlow port
number, or any of the LOCAL, CONTROLLER, or NONE. This
field is required.
pipeline_field=value
Optionally, user can specify a list of pipeline fields for
a packet-out message. The supported pipeline fields
includes tunnel fields and register fields as defined in
ovs-fields(7).
packet=hex-string
The actual packet to send, expressed as a string of
hexadecimal bytes. This field is required.
actions=[action][,action...]
The syntax of actions are identical to the actions= field
described in Flow Syntax above. Specifying actions= is
optional, but omitting actions is interpreted as a drop,
so the packet will not be sent anywhere from the switch.
actions must be specified at the end of each line, like
for flow mods.
Group Syntax
Some ovs-ofctl commands accept an argument that describes a group
or groups. Such flow descriptions comprise a series field=value
assignments, separated by commas or white space. (Embedding
spaces into a group description normally requires quoting to
prevent the shell from breaking the description into multiple
arguments.). Unless noted otherwise only the last instance of
each field is honoured.
group_id=id
The integer group id of group. When this field is
specified in del-groups or dump-groups, the keyword "all"
may be used to designate all groups. This field is
required.
type=type
The type of the group. The add-group, add-groups and mod-
groups commands require this field. It is prohibited for
other commands. The following keywords designated the
allowed types:
all Execute all buckets in the group.
select Execute one bucket in the group, balancing across
the buckets according to their weights. To select
a bucket, for each live bucket, Open vSwitch hashes
flow data with the bucket ID and multiplies by the
bucket weight to obtain a ``score,'' and then
selects the bucket with the highest score. Use
selection_method to control the flow data used for
selection.
indirect
Executes the one bucket in the group.
ff
fast_failover
Executes the first live bucket in the group which
is associated with a live port or group.
command_bucket_id=id
The bucket to operate on. The insert-buckets and remove-
buckets commands require this field. It is prohibited for
other commands. id may be an integer or one of the
following keywords:
all Operate on all buckets in the group. Only valid
when used with the remove-buckets command in which
case the effect is to remove all buckets from the
group.
first Operate on the first bucket present in the group.
In the case of the insert-buckets command the
effect is to insert new bucets just before the
first bucket already present in the group; or to
replace the buckets of the group if there are no
buckets already present in the group. In the case
of the remove-buckets command the effect is to
remove the first bucket of the group; or do nothing
if there are no buckets present in the group.
last Operate on the last bucket present in the group.
In the case of the insert-buckets command the
effect is to insert new bucets just after the last
bucket already present in the group; or to replace
the buckets of the group if there are no buckets
already present in the group. In the case of the
remove-buckets command the effect is to remove the
last bucket of the group; or do nothing if there
are no buckets present in the group.
If id is an integer then it should correspond to the
bucket_id of a bucket present in the group. In case of
the insert-buckets command the effect is to insert buckets
just before the bucket in the group whose bucket_id is id.
In case of the iremove-buckets command the effect is to
remove the in the group whose bucket_id is id. It is an
error if there is no bucket persent group in whose
bucket_id is id.
selection_method=method
The selection method used to select a bucket for a select
group. This is a string of 1 to 15 bytes in length known
to lower layers. This field is optional for add-group,
add-groups and mod-group commands on groups of type
select. Prohibited otherwise. If no selection method is
specified, Open vSwitch up to release 2.9 applies the hash
method with default fields. From 2.10 onwards Open vSwitch
defaults to the dp_hash method with symmetric L3/L4 hash
algorithm, unless the weighted group buckets cannot be
mapped to a maximum of 64 dp_hash values with sufficient
accuracy. In those rare cases Open vSwitch 2.10 and later
fall back to the hash method with the default set of hash
fields.
dp_hash
Use a datapath computed hash value. The hash
algorithm varies accross different datapath
implementations. dp_hash uses the upper 32 bits of
the selection_method_param as the datapath hash
algorithm selector. The supported values are 0
(corresponding to hash computation over the IP
5-tuple) and 1 (corresponding to a symmetric hash
computation over the IP 5-tuple). Selecting
specific fields with the fields option is not
supported with dp_hash). The lower 32 bits are
used as the hash basis.
Using dp_hash has the advantage that it does not
require the generated datapath flows to exact match
any additional packet header fields. For example,
even if multiple TCP connections thus hashed to
different select group buckets have different
source port numbers, generally all of them would be
handled with a small set of already established
datapath flows, resulting in less latency for TCP
SYN packets. The downside is that the shared
datapath flows must match each packet twice, as the
datapath hash value calculation happens only when
needed, and a second match is required to match
some bits of its value. This double-matching
incurs a small additional latency cost for each
packet, but this latency is orders of magnitude
less than the latency of creating new datapath
flows for new TCP connections.
hash Use a hash computed over the fields specified with
the fields option, see below. If no hash fields
are specified, hash defaults to a symmetric hash
over the combination of MAC addresses, VLAN tags,
Ether type, IP addresses and L4 port numbers. hash
uses the selection_method_param as the hash basis.
Note that the hashed fields become exact matched by
the datapath flows. For example, if the TCP source
port is hashed, the created datapath flows will
match the specific TCP source port value present in
the packet received. Since each TCP connection
generally has a different source port value, a
separate datapath flow will be need to be inserted
for each TCP connection thus hashed to a select
group bucket.
This option uses a Netronome OpenFlow extension which is
only supported when using Open vSwitch 2.4 and later with
OpenFlow 1.5 and later.
selection_method_param=param
64-bit integer parameter to the selection method selected
by the selection_method field. The parameter's use is
defined by the lower-layer that implements the
selection_method. It is optional if the selection_method
field is specified as a non-empty string. Prohibited
otherwise. The default value is zero.
This option uses a Netronome OpenFlow extension which is
only supported when using Open vSwitch 2.4 and later with
OpenFlow 1.5 and later.
fields=field
fields(field[=mask]...)
The field parameters to selection method selected by the
selection_method field. The syntax is described in Flow
Syntax with the additional restrictions that if a value is
provided it is treated as a wildcard mask and wildcard
masks following a slash are prohibited. The pre-requisites
of fields must be provided by any flows that output to the
group. The use of the fields is defined by the lower-
layer that implements the selection_method. They are
optional if the selection_method field is specified as
``hash', prohibited otherwise. The default is no fields.
This option will use a Netronome OpenFlow extension which
is only supported when using Open vSwitch 2.4 and later
with OpenFlow 1.5 and later.
bucket=bucket_parameters
The add-group, add-groups and mod-group commands require
at least one bucket field. Bucket fields must appear after
all other fields. Multiple bucket fields to specify
multiple buckets. The order in which buckets are
specified corresponds to their order in the group. If the
type of the group is "indirect" then only one group may be
specified. bucket_parameters consists of a list of
field=value assignments, separated by commas or white
space followed by a comma-separated list of actions. The
fields for bucket_parameters are:
bucket_id=id
The 32-bit integer group id of the bucket. Values
greater than 0xffffff00 are reserved. This field
was added in Open vSwitch 2.4 to conform with the
OpenFlow 1.5 specification. It is not supported
when earlier versions of OpenFlow are used. Open
vSwitch will automatically allocate bucket ids when
they are not specified.
actions=[action][,action...]
The syntax of actions are identical to the actions=
field described in Flow Syntax above. Specifying
actions= is optional, any unknown bucket parameter
will be interpreted as an action.
weight=value
The relative weight of the bucket as an integer.
This may be used by the switch during bucket select
for groups whose type is select.
watch_port=port
Port used to determine liveness of group. This or
the watch_group field is required for groups whose
type is ff or fast_failover. This or the
watch_group field can also be used for groups whose
type is select.
watch_group=group_id
Group identifier of group used to determine
liveness of group. This or the watch_port field is
required for groups whose type is ff or
fast_failover. This or the watch_port field can
also be used for groups whose type is select.
Meter Syntax
The meter table commands accept an argument that describes a
meter. Such meter descriptions comprise a series field=value
assignments, separated by commas or white space. (Embedding
spaces into a group description normally requires quoting to
prevent the shell from breaking the description into multiple
arguments.). Unless noted otherwise only the last instance of
each field is honoured.
meter=id
The identifier for the meter. An integer is used to
specify a user-defined meter. In addition, the keywords
"all", "controller", and "slowpath", are also supported as
virtual meters. The "controller" and "slowpath" virtual
meters apply to packets sent to the controller and to the
OVS userspace, respectively.
When this field is specified in del-meter, dump-meter, or
meter-stats, the keyword "all" may be used to designate
all meters. This field is required, except for meter-
stats, which dumps all stats when this field is not
specified.
kbps
pktps The unit for the rate and burst_size band parameters.
kbps specifies kilobits per second, and pktps specifies
packets per second. A unit is required for the add-meter
and mod-meter commands.
burst If set, enables burst support for meter bands through the
burst_size parameter.
stats If set, enables the collection of meter and band
statistics.
bands=band_parameters
The add-meter and mod-meter commands require at least one
band specification. Bands must appear after all other
fields.
type=type
The type of the meter band. This keyword starts a
new band specification. Each band specifies a rate
above which the band is to take some action. The
action depends on the band type. If multiple
bands' rate is exceeded, then the band with the
highest rate among the exceeded bands is selected.
The following keywords designate the allowed meter
band types:
drop Drop packets exceeding the band's rate
limit.
The other band_parameters are:
rate=value
The relative rate limit for this band, in kilobits
per second or packets per second, depending on
whether kbps or pktps was specified.
burst_size=size
If burst is specified for the meter entry,
configures the maximum burst allowed for the band
in kilobits or packets, depending on whether kbps
or pktps was specified. If unspecified, the switch
is free to select some reasonable value depending
on its configuration.
