5 ** New OVN logical actions
9 Generates an ARP packet based on the current IPv4 packet and allows it
10 to be processed as part of the current pipeline (and then pop back to
11 processing the original IPv4 packet).
13 TCP/IP stacks typically limit the rate at which ARPs are sent, e.g. to
14 one per second for a given target. We might need to do this too.
16 We probably need to buffer the packet that generated the ARP. I don't
17 know where to do that.
19 *** icmp4 { action... }
21 Generates an ICMPv4 packet based on the current IPv4 packet and
22 processes it according to each nested action (and then pops back to
23 processing the original IPv4 packet). The intended use case is for
24 generating "time exceeded" and "destination unreachable" errors.
26 ovn-sb.xml includes a tentative specification for this action.
28 Tentatively, the icmp4 action sets a default icmp_type and icmp_code
29 and lets the nested actions override it. This means that we'd have to
30 make icmp_type and icmp_code writable. Because changing icmp_type and
31 icmp_code can change the interpretation of the rest of the data in the
32 ICMP packet, we would want to think this through carefully. If it
33 seems like a bad idea then we could instead make the type and code a
34 parameter to the action: icmp4(type, code) { action... }
36 It is worth considering what should be considered the ingress port for
37 the ICMPv4 packet. It's quite likely that the ICMPv4 packet is going
38 to go back out the ingress port. Maybe the icmp4 action, therefore,
39 should clear the inport, so that output to the original inport won't
44 Transforms the current TCP packet into a RST reply.
46 ovn-sb.xml includes a tentative specification for this action.
48 *** Other actions for IPv6.
50 IPv6 will probably need an action or actions for ND that is similar to
51 the "arp" action, and an action for generating
61 ** Dynamic IP to MAC bindings
63 Some bindings from IP address to MAC will undoubtedly need to be
64 discovered dynamically through ARP requests. It's straightforward
65 enough for a logical L3 router to generate ARP requests and forward
66 them to the appropriate switch.
68 It's more difficult to figure out where the reply should be processed
69 and stored. It might seem at first that a first-cut implementation
70 could just keep track of the binding on the hypervisor that needs to
71 know, but that can't happen easily because the VM that sends the reply
72 might not be on the same HV as the VM that needs the answer (that is,
73 the VM that sent the packet that needs the binding to be resolved) and
74 there isn't an easy way for it to know which HV needs the answer.
76 Thus, the HV that processes the ARP reply (which is unknown when the
77 ARP is sent) has to tell all the HVs the binding. The most obvious
78 place for this in the OVN_Southbound database.
80 Details need to be worked out, including:
82 *** OVN_Southbound schema changes.
84 Possibly bindings could be added to the Port_Binding table by adding
85 or modifying columns. Another possibility is that another table
88 *** Logical_Flow representation
90 It would be really nice to maintain the general-purpose nature of
91 logical flows, but these bindings might have to include some
92 hard-coded special cases, especially when it comes to the relationship
93 with populating the bindings into the OVN_Southbound table.
97 It's probably best to only record in the database responses to queries
98 actually issued by an L3 logical router, so somehow they have to be
99 tracked, probably by putting a tentative binding without a MAC address
102 *** Renewal and expiration.
104 Something needs to make sure that bindings remain valid and expire
105 those that become stale.
107 ** MTU handling (fragmentation on output)
113 *** ICMP error generation, TCP reset, UDP unreachable, protocol unreachable, ...
115 As a point of comparison, Linux doesn't ratelimit TCP resets but I
116 think it does everything else.
120 ** ovn-controller parameters and configuration.
122 *** SSL configuration.
124 Can probably get this from Open_vSwitch database.
128 *** Limiting the impact of a compromised chassis.
130 Every instance of ovn-controller has the same full access to the central
131 OVN_Southbound database. This means that a compromised chassis can
132 interfere with the normal operation of the rest of the deployment. Some
133 specific examples include writing to the logical flow table to alter
134 traffic handling or updating the port binding table to claim ports that are
135 actually present on a different chassis. In practice, the compromised host
136 would be fighting against ovn-northd and other instances of ovn-controller
137 that would be trying to restore the correct state. The impact could include
138 at least temporarily redirecting traffic (so the compromised host could
139 receive traffic that it shouldn't) and potentially a more general denial of
142 There are different potential improvements to this area. The first would be
143 to add some sort of ACL scheme to ovsdb-server. A proposal for this should
144 first include an ACL scheme for ovn-controller. An example policy would
145 be to make Logical_Flow read-only. Table-level control is needed, but is
146 not enough. For example, ovn-controller must be able to update the Chassis
147 and Encap tables, but should only be able to modify the rows associated with
148 that chassis and no others.
150 A more complex example is the Port_Binding table. Currently, ovn-controller
151 is the source of truth of where a port is located. There seems to be no
152 policy that can prevent malicious behavior of a compromised host with this
155 An alternative scheme for port bindings would be to provide an optional mode
156 where an external entity controls port bindings and make them read-only to
157 ovn-controller. This is actually how OpenStack works today, for example.
158 The part of OpenStack that manages VMs (Nova) tells the networking component
159 (Neutron) where a port will be located, as opposed to the networking
160 component discovering it.
164 ovsdb-server should have adequate features for OVN but it probably
165 needs work for scale and possibly for availability as deployments
166 grow. Here are some thoughts.
168 Andy Zhou is looking at these issues.
170 *** Reducing amount of data sent to clients.
172 Currently, whenever a row monitored by a client changes,
173 ovsdb-server sends the client every monitored column in the row,
174 even if only one column changes. It might be valuable to reduce
175 this only to the columns that changes.
177 Also, whenever a column changes, ovsdb-server sends the entire
178 contents of the column. It might be valuable, for columns that
179 are sets or maps, to send only added or removed values or
182 Currently, clients monitor the entire contents of a table. It
183 might make sense to allow clients to monitor only rows that
184 satisfy specific criteria, e.g. to allow an ovn-controller to
185 receive only Logical_Flow rows for logical networks on its hypervisor.
187 *** Reducing redundant data and code within ovsdb-server.
189 Currently, ovsdb-server separately composes database update
190 information to send to each of its clients. This is fine for a
191 small number of clients, but it wastes time and memory when
192 hundreds of clients all want the same updates (as will be in the
195 (This is somewhat opposed to the idea of letting a client monitor
196 only some rows in a table, since that would increase the diversity
201 If it turns out that other changes don't let ovsdb-server scale
202 adequately, we can multithread ovsdb-server. Initially one might
203 only break protocol handling into separate threads, leaving the
204 actual database work serialized through a lock.
206 ** Increasing availability.
208 Database availability might become an issue. The OVN system
209 shouldn't grind to a halt if the database becomes unavailable, but
210 it would become impossible to bring VIFs up or down, etc.
212 My current thought on how to increase availability is to add
213 clustering to ovsdb-server, probably via the Raft consensus
214 algorithm. As an experiment, I wrote an implementation of Raft
215 for Open vSwitch that you can clone from:
217 https://github.com/blp/ovs-reviews.git raft
219 ** Reducing startup time.
221 As-is, if ovsdb-server restarts, every client will fetch a fresh
222 copy of the part of the database that it cares about. With
223 hundreds of clients, this could cause heavy CPU load on
224 ovsdb-server and use excessive network bandwidth. It would be
225 better to allow incremental updates even across connection loss.
226 One way might be to use "Difference Digests" as described in
227 Epstein et al., "What's the Difference? Efficient Set
228 Reconciliation Without Prior Context". (I'm not yet aware of
229 previous non-academic use of this technique.)
231 ** Support multiple tunnel encapsulations in Chassis.
233 So far, both ovn-controller and ovn-controller-vtep only allow
234 chassis to have one tunnel encapsulation entry. We should extend
235 the implementation to support multiple tunnel encapsulations.
237 ** Update learned MAC addresses from VTEP to OVN
239 The VTEP gateway stores all MAC addresses learned from its
240 physical interfaces in the 'Ucast_Macs_Local' and the
241 'Mcast_Macs_Local' tables. ovn-controller-vtep should be
242 able to update that information back to ovn-sb database,
243 so that other chassis know where to send packets destined
244 to the extended external network instead of broadcasting.
246 ** Translate ovn-sb Multicast_Group table into VTEP config
248 The ovn-controller-vtep daemon should be able to translate
249 the Multicast_Group table entry in ovn-sb database into
250 Mcast_Macs_Remote table configuration in VTEP database.
252 * Consider the use of BFD as tunnel monitor.
254 The use of BFD for hypervisor-to-hypervisor tunnels is probably not worth it,
255 since there's no alternative to switch to if a tunnel goes down. It could
256 make sense at a slow rate if someone does OVN monitoring system integration,
259 When OVN gets to supporting HA for gateways (see ovn/OVN-GW-HA.md), BFD is
260 likely needed as a part of that solution.
262 There's more commentary in this ML post:
263 http://openvswitch.org/pipermail/dev/2015-November/062385.html
269 ** Support reject action.
271 ** Support log option.