Classifications

• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L12/00—Data switching networks
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
  • H04L41/08—Configuration management of networks or network elements
  • H04L41/0803—Configuration setting
  • H04L41/0813—Configuration setting characterised by the conditions triggering a change of settings
  • H04L41/0816—Configuration setting characterised by the conditions triggering a change of settings the condition being an adaptation, e.g. in response to network events
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
  • H04L41/06—Management of faults, events, alarms or notifications
  • H04L41/0654—Management of faults, events, alarms or notifications using network fault recovery
  • H04L41/0659—Management of faults, events, alarms or notifications using network fault recovery by isolating or reconfiguring faulty entities
  • H04L41/0661—Management of faults, events, alarms or notifications using network fault recovery by isolating or reconfiguring faulty entities by reconfiguring faulty entities
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L43/00—Arrangements for monitoring or testing data switching networks
  • H04L43/08—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
  • H04L43/0805—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability
  • H04L43/0811—Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters by checking availability by checking connectivity
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
  • H04L45/00—Routing or path finding of packets in data switching networks
  • H04L45/24—Multipath
  • H04L45/245—Link aggregation, e.g. trunking
• • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
  • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
  • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
  • Y02D30/00—Reducing energy consumption in communication networks
  • Y02D30/50—Reducing energy consumption in communication networks in wire-line communication networks, e.g. low power modes or reduced link rate

Definitions

• the present invention generally relates to link aggregation, and more particularly relates to methods and apparatus for detecting and handling split brain conditions in a Link Aggregation Group.
• Link Aggregation is widely used to aggregate multiple links between a pair of nodes, in order to be able to transmit user data on each of the links participating a Link Aggregation Group (LAG) (see, e.g., IEEE 802.1AX). Aggregating multiple network connections in this fashion can increase throughput beyond what a single connection can sustain, and/or can be used to provide redundancy in case of a failure of one of the links.
• LAG Link Aggregation Group
• the "Distributed Resilient Network Interconnect” (see Clause 8 of IEEE 802.1AX-REV/D0.2) specifies extensions to Link Aggregation in order to be able to use link aggregation on a network interface even between more than two nodes, for example between four nodes A, B, C and D as illustrated in Fig. 1.
• the label "DRNI” is used to indicate a LAG that includes the four nodes A, B, C and D.
• a LAG is formed between Network 1 and Network 2. More specifically, a LAG is formed between LAG virtual nodes 30, 32.
• First LAG virtual node 30 includes a first node (A) and a second node (B).
• Second LAG virtual node 32 includes a third node (C) and a fourth node (D).
• LAG Nodes A and C are connected as peer nodes, and LAG Nodes B and D are also connected as peer nodes.
• nodes A and B are connected as “fellow nodes,” and similarly within virtual node 32 nodes C and D are connected as “fellow nodes.”
• a "LAG virtual node” refers to a DRNI portal in the IEEE documentation discussed above (i.e., two nodes that appear as a single node to their respective peers).
• the statement that virtual node 30 "includes" two nodes A, B means that the virtual node 30 is emulated by the nodes A, B.
• the statement that virtual node 32 "includes" two nodes C, D means that the virtual node 32 is emulated by the nodes C, D.
• the System ID is used to identify each node (e.g., node A, node B, node C, node D).
• the System ID is included in Link Aggregation Control Protocol Data Units (LACPDUs) sent between the individual nodes of the LAG. It is practical to use the System ID of one of the fellow nodes as a common System ID for their corresponding LAG virtual node. Thus, as shown in Fig.
• node A and node B belong to the same Network 1 and they are part of the same DRNI Portal (i.e., the same LAG virtual node 30), and use a common System ID of "A" for the emulated LAG virtual node 30.
• Nodes C and D of Network 2 are seen as a single LAG virtual node 32 with a System ID "C" by Nodes A and B.
• Fig. 1 also shows the DRNI hand-off of a particular service (see bold "service" line in Fig. 1 ).
• the service handed-off on an interface may be a Virtual Local Area Network (VLAN), and an identifier for the service may be a VLAN Identifier (VID), such as a Service VID (i.e., VLAN Identifier (VID), such as a Service VID (i.e., VLAN Identifier
• S-VID Service to Network Interfaces
• C-VID Customer VID
• UNIs User to Network Interfaces
• the service is handed off on the upper link (between nodes A, C), as both Networks 1 and 2 have selected the upper nodes as “active gateway nodes” and have selected the upper link for the service hand-off.
• active gateway nodes are shown as having a bold boundary. This gateway functionality is introduced by DRNI for loop prevention.
• the nodes B and D block the service from being handed-off between the DRNI and their own respective networks.
• a "portal node failure” illustrated in Fig. 2 ("portal node” and “LAG node” are being used synonymously in this context).
• Node A experiences a failure and can no longer communicate with Node B or Node C.
• Node B would start to use its own System ID for the LAG instead of the formerly used common System ID, which in the Example of Fig. 1 was the System ID of Node A.
• Node C is aware of the unreachability of node A, and node D may be aware of it too.
• Node C and Node D have to accept the new partner System ID (B) in order to provide LAG connectivity.
• B System ID
• Graceful name change from the old System ID (A) to the new System ID (B) can be applied for smoother transition and for not dropping and reestablishing an active aggregation (see, e.g., N. Finn, Graceful Name Change in LACP, Std. contrib. 201 1 , http://www.ieee802.Org/1/files/public/docs201 1/axbq-nfinn-graceful-name- change-051 1-v1.pdf).
• Fig. 2 The behavior illustrated in Fig. 2 is problematic though, because it is based on changing the System ID of the portal based on the System IDs of the individual nodes that comprise the portal.
• this prior art solution provides visibility to the individual systems (i.e., Nodes C and D learn that node A has failed), which goes against a main design principle of the DRNI, which is to hide the details of its internal systems that provide an external view of a single LAG virtual node to its peers.
• DRNI main design principle of the DRNI principles, it is desirable to avoid System ID change even if a node failure occurs.
• FIG 3 shows another failure event, in the case when connectivity between nodes on the same portal (i.e., the "portal link") is broken, causing the link between fellow Nodes A and B to fail.
• nodes cannot distinguish between portal link and portal node failures, and Node B's reaction to the portal link failure is the same as to the portal node failure explained above (i.e., Node B starts using its own System ID instead of the common System ID).
• Node A is up and running and also uses its own System ID in LACPDUs, which is the same as the common System ID, as illustrated in Fig. 2.
• Nodes C and D then only maintain the links towards the LAG virtual node 30 that use the common System ID (i.e., the links to Node A in this example).
• the links to the other node are deactivated by nodes C and D by deactivating the links from LAG as illustrated in Fig. 4 (see dotted line between Node B and Node D - this notation will be used throughout this application to indicate a deactivated link).
• a split brain The situation caused by the portal link failure of Figs. 3-4 is referred to as a "split brain" (SB), because both nodes A and B emulating a single LAG virtual node 30 are up and running but not connected to each other. If both of the split brain nodes had become active gateway, then a loop would appear. Nevertheless, the peering partner nodes C and D are able to inform the split brain nodes A and B that they are in a split brain situation as shown by Fig. 4. That is, both nodes C and D use an LACPDU to inform their respective peer node that a split brain situation has occurred at the LAG virtual node 30.
• SB split brain
• Node B does not become the active gateway for the service of Fig. 1 .
• the connectivity between nodes C and D works properly, they emulate a single node LAG virtual node 32 and both of them are aware of the different System IDs received from their respective peer nodes A and B.
• a "Double Split Brain” (DSB) condition also known as a “Dual Split Brain”
• DSB Double Split Brain
• both sides of the LAG are experiencing split brain, then neither LAG virtual node 30, 32 is able to detect the split brain situation of the other LAG virtual node 30, 32, because there is no connection within either portal.
• the nodes of the same portal cannot notify each other of the fact that they receive different System IDs in their respective LACPDUs, which was the basis of prior art single split brain handling. Therefore, all the nodes will consider their fellow node within the portal to be down, and all the nodes become active gateway for all services. This results in forwarding loop of data frames as illustrated in Fig. 5. No method is available for handling Double split brain situations.
• a number of methods for operating a virtual node in a Link Aggregation Group (LAG) comprising a first virtual node and a second virtual node are disclosed.
• LAG Link Aggregation Group
• a method of operating a virtual node in a LAG including a first virtual node and a second virtual node is disclosed.
• the first virtual node includes at least a first fellow node and a second fellow node.
• the first fellow node determines whether the first fellow node is in communication with the second fellow node.
• the first fellow node also transmits control information to the second virtual node based on the determining such that: the control information comprises a system ID and first configuration information associated with the LAG when the determining indicates that the first fellow node is in communication with the second fellow node; and the control information comprises the system ID and second configuration information associated with the LAG when the determining indicates that the first fellow node is not in communication with the second fellow node, the second configuration information being different than the first configuration information.
• the first configuration information and the second configuration information are one of configuration keys and service digests.
• the first configuration information and the second configuration information are carried in Link Aggregation Control Protocol Data Unit (LACPDU) messages.
• LACPDU Link Aggregation Control Protocol Data Unit
• a method of operating a virtual node in a LAG including a first virtual node and a second virtual node is disclosed.
• the first virtual node includes at least a first fellow node and a second fellow node.
• the first fellow node receives, from the second virtual node, first control information comprising a system ID and first configuration information associated with the LAG.
• the first configuration information is compared against reference configuration information associated with the LAG.
• a split brain condition is determined to exist in the LAG in response to at least one of: detecting that the first fellow node is not in communication with the second fellow node; and the second fellow node being in communication with the second virtual node and the first configuration information not matching second configuration information provided to the second fellow node by the second virtual node.
• the reference configuration represents previously established expected configuration information associated with the LAG.
• the determining that the split brain condition exists includes determining that a dual split brain condition exists in the LAG in response to detecting that the first fellow node is not in communication with the second fellow node.
• a non-primary or non-default connecting link that connects the first and second virtual nodes may be excluded from the LAG if a dual split brain condition is determined to exist in the LAG.
• the determining that the split brain condition exists includes determining that a single split brain condition exists in the LAG in response to the second fellow node being in communication with the second virtual node and the first configuration information not matching second configuration information provided to the second fellow node by the second virtual node.
• the first, second and reference configuration information may be one of a configuration key and a configuration digest.
• the configuration information may be carried in a Link Aggregation Control Protocol Data Unit (LACPDU) message.
• LACPDU Link Aggregation Control Protocol Data Unit
• Fig. 1 is a block diagram illustrating a LAG supporting a service.
• Fig. 2 is a block diagram of a portal node failure in a LAG.
• Fig. 3 is a block diagram of a portal link failure in a LAG, causing a split brain condition.
• Fig. 4 is a block diagram of a prior art solution for handling the split brain condition of
• Fig. 5 is a block diagram of a double split brain condition.
• Figs. 6-7 are block diagrams of a normal LAG operating state using a novel System ID naming arrangement.
• Fig. 8 is a block diagram of a node failure in the LAG of Figs. 6-7.
• Fig. 9 is a block diagram of a normal operating condition for LAG nodes using a common key.
• Fig. 10 is a block diagram of a non-primary LAG node failure in the LAG of Fig 9.
• Fig. 1 1 is a block diagram of a primary LAG node failure in the LAG of Fig 9.
• Figs. 12-13 are block diagrams of single split brain handling in the LAG of Fig. 9
• Fig. 14 is a block diagram of a normal operating condition for LAG nodes using service assignment digests.
• Fig. 15 is a block diagram of a LAG node failure in the LAG of Fig. 14.
• Figs. 16-17 are block diagrams of single split brain handling in the LAG of Fig. 14.
• Fig. 18 illustrates an exemplary method of operating a virtual node in a LAG to detect split brain conditions.
• Fig. 19 illustrates another exemplary method of operating a virtual node in a LAG to detect split brain conditions.
• Figs. 20-21 are block diagrams of double split brain handling in a LAG.
• Fig. 22 is a block diagram of double split brain handling in a fully meshed DRNI.
• Fig. 23 is a block diagram of broken LAG connectivity due to double split brain mishandling.
• Fig. 24 illustrates an exemplary method of operating a virtual node in a LAG.
• Fig. 25 illustrates another exemplary method of operating a virtual node in a LAG.
• Fig. 26 illustrates an exemplary LAG node.
• Figs. 6-7 illustrate a System ID naming method, where a same node-identifying System
• each LAG virtual node 30, 32 is used at each LAG virtual node 30, 32.
• Nodes A and B claim to be the single virtual Node A' in LAG virtual node 30 towards Network 2. That is, both Node C and Node D receive LACPDUs such that their partner's System ID is A' as shown in Fig. 6, and they perceive themselves as being connected to a single Node A'.
• the operation on the other side is the same, as Node C and Node D claim to be a single LAG virtual node 32 toward Network 1 , with a System ID of C.
• the existing split brain handling solution cannot be applied if the DRNI operational principles are aimed to be applied and maintained all the time, as the LAG nodes A, B, C, D in Figs. 6-7 do not change the System ID they advertise for LAG.
• One potential step to minimize the chance of the appearance of a split brain situation is to apply protected links within the portal for the interconnection of the fellow portal nodes (i.e., have protected portal links between A-B, and between C-D).
• the fellow portal nodes could be connected to each other by multiple physical links grouped by LAG.
• Other methods that may be used in combination with or as an alternative to this method are described below.
• Figs. 9-13 illustrate an exemplary method of using key change to detect and address single split brain conditions.
• a double circle around a node indicates that the node is a "primary node" within its respective LAG virtual node 30 or 32.
• Figs. 9-13 The method of Figs. 9-13 is proposed for single split brain handling if System IDs are not changed when failures occur, and they are not changed under any circumstances.
• the proposed method relies on changing an Aggregation Key if a fellow portal node is perceived to be down.
• the Aggregation Key (or "Aggregator Key” or “Actor Key” or simply “Key”) is used in LACPDUs transmitted between Nodes A and C, and between Nodes B and D. Referring to Fig. 9, in a normal operating condition each of nodes A and B use a key "K" in their
• the non-primary node B does change its Aggregation Key as shown in Fig. 1 1 . That is, if Node A goes down then Node B changes the key in its LACPDUs to "L" instead of the formerly used common key "K.” In order to avoid disruption due to the key change, the old and the new key values may be carried in LACPDUs during the change and for a while after it. Thus the LAG can be maintained if there is a match with either the old or the new key. Notably, the same System ID of A' continues to be used by Node B, despite Node B perceiving Node A as having failed.
• Figs. 12-13 The handling of a single split brain condition by key change, according to one exemplary embodiment, is shown in Figs. 12-13.
• Fig. 12 if the connectivity within the portal of virtual node 30 is broken, then both portal nodes A, B perceive that the fellow portal node is down. Neither of Node A or Node B changes its System ID though, as both continue using System ID A' towards their peer nodes C, D.
• the primary node (i.e., Node A in the example), also does not change its key and it keeps using the common key "K.”
• Node B which is a non-primary node, changes its key value to key "L.”
• the peer nodes receive different key values in LACPDUs, and use this to detect that a single split brain has occurred at LAG virtual node 30.
• the link on which the expected key K is received is kept in the LAG (i.e., the link between nodes A and C).
• Those links where the received key differs from the expected one are excluded from LAG as shown in Fig. 13 (i.e. the link between B and D is excluded - as shown by the dotted line between B and D).
• the difference between the node failure and the split brain conditions is that in the case of a node failure a single key is received by the peer even if the key is changed.
• the peer is able to recognize a split brain situation and inform the nodes being in split brain about the fact they are in split brain as illustrated in Fig. 13.
• Figs. 14-17 illustrate an exemplary method of using service assignment digests to detect and address single split brain conditions.
• there is no "primary” and “non-primary” node and thus there are no double circles around any nodes.
• these embodiments utilize the notion of a "default link” and a “non-default link” in a similar fashion.
• information of the actual service assignments to the individual links that are aggregated in a LAG could be carried in the exchanged LACPDUs, in the form of digests (see, e.g., Ericsson provisional patent application Serial No. 61/605,818, docket number 4015-8027 / P361 13, entitled "Method for Ensuring Congruency in Link Aggregation").
• certain services are allocated to specific LAG peer links, and digests are transmitted between LAG nodes to indicate these service assignments. Comparison of the exchanged assignments at the two ends of the LAG enables verification of the congruency of the carried services.
• a split brain avoidance method relies on detecting change in the service to link assignment (i.e., differences compared to service assignment information expected to be received in a digest). As described below in greater detail, by taking service assignment digests to represent LAG service assignments in LACPDUs, split brain conditions may be detected based on certain digest changes.
• Fig. 14 is a block diagram of a normal operating condition for LAG that use service assignment digests. Because all links are operational in Fig. 14, a same service assignment is used throughout the LAG, and a same service digest "G" is transmitted between peer nodes in the LAG. That is, on each aggregated link an additional service digest is exchanged carrying the information on, for example, VLAN Identifier (VID) assignments to physical links. As the VIDs are handed-off in a congruent manner, all the digest values are the same and a same service digest "G" is exchanged between both LAG virtual nodes 30, 32 as shown in Fig. 14. The assignment is consistent if a VID is assigned to the same physical link on both sides of the LAG, and the digest covers the assignments of all VIDs. Thus, the same digest is expected to be received as the one that is sent.
• VID VLAN Identifier
• Fig. 15 illustrates operation in case of a node failure.
• all of the services that were carried by peer link B-D are now being carried by peer link A-C.
• This means a new Digest H will be communicated to Network 2 indicating the service transfer.
• the LAG virtual node 32 of Network 2 is also aware of the failure because of losing a link of LAG, therefore, virtual node 32 also makes changes in the appropriate arrangements as to move all the services to link A-C and update the digest.
• the digest exchange becomes as depicted in Fig. 15.
• FIG. 16 and 17 An exemplary method of handling a single split brain condition based on changes in the service assignments is shown in Figs. 16 and 17.
• the connectivity within the portal is broken (e.g., link between Node A and Node B is broken)
• both portal nodes A and B perceive that their fellow portal node is down.
• Each of the aggregated links takes over all the services that used to be carried on the other link.
• Digest "H” sent by Node A is calculated such that all services are carried over the A-C link.
• Node B also sends Digest "J” which is calculated such that all services carried over the B-D link.
• the portal nodes in the split brain send different digests as shown in Fig. 16.
• each of Nodes C and D receive different digests from their peer nodes A and B over the A-C and B-D links.
• Nodes C and D can use this information to determine that their peer is experiencing a split brain condition, and can take appropriate action so that the LAG is split by separating the lower ID link (i.e., deactivating the "non-default link” between Node B and Node D), and keeping the "default link” in LAG (which is the A-C link between Node A and Node C in the example of Figs 16-17).
• Nodes C and D inform their respective peer nodes on A and B about the split brain condition at the LAG virtual node 30 (see "SB" notifications in Fig. 17).
• Fig. 24 illustrates an exemplary method 300 of operating a virtual node in a LAG comprising a first virtual node (e.g., LAG virtual node 30) and a second virtual node (e.g., LAG virtual node 32).
• the method 300 will be described with the "first virtual node" being LAG virtual node 30, which includes at least a first fellow node and a second fellow node (e.g., nodes A and B).
• the first fellow node determines whether the first fellow node is in communication with the second fellow node (step 302), and transmits control information to the second virtual node based on the determining (steps 304, 306).
• the control information includes a system ID and first configuration information associated with the LAG when the determining indicates that the first fellow node is in communication with the second fellow node (step 304).
• the control information comprises the system ID and second configuration information associated with the LAG when the determining indicates that the first fellow node is not in communication with the second fellow node, with the second configuration information being different than the first configuration information (step 306).
• the first configuration information and the second configuration information may be configuration keys and/or service digests. Also, the first configuration information and the second configuration information may optionally but advantageously be carried in LACPDU messages.
• Figs. 18-19 illustrate exemplary detailed methods of operating a virtual node in a LAG to detect split brain conditions.
• each LAG node (A, B, C, D) is configured to expect certain information from its LAG peer, and once detected, the double split brain condition may be addressed by excluding a connecting peer link.
• Fig. 18 illustrates an exemplary method 100 that facilitates single or double split brain handling, based on the reception of a LACPDU from a peer node.
• the method 100 will be described from the standpoint of Node D of LAG virtual node 32.
• a check is performed to determine if control information has been received from the other virtual node (e.g. from node 30) (step 102).
• the information may be received, for example, in a LACPDU.
• the control information includes a system ID and also first configuration information.
• a check is performed to determine if the first configuration information matches reference configuration information (step 104).
• the reference configuration represents previously established expected configuration information associated with the LAG (e.g., an expected key).
• step 106 If the first configuration information does not match the reference configuration information, a check is performed to determine if Node D is able to communicate with its fellow Node C (i.e., whether node C is reachable) (step 106). If Nodes C and D can communicate with each other, then a check is performed to determine if second configuration information is received at fellow Node C (step 107). If the Second Configuration Information is not received after a defined period of time, then it may be determined that peer Node A has failed (step 108).
• a single split brain condition is determined to have occurred at LAG virtual node 30 (step 109). Based on this, a check is performed to determine if the link in question is still a member of the LAG (step 1 12). In the example of Node D, this would be the peer link connecting Nodes B and D. If the link is still a member of the LAG, then it is excluded from the LAG (step 1 14), and a LACPDU with a single SB notification is sent to the LAG peer, which would be Node B (step 1 16).
• step 1 10 if Nodes C and D are unable to communicate with each other, then a double split brain condition is detected (step 1 10), and steps 1 12-1 16 are performed in a similar fashion, except that a DSB notification would be provided in step 1 16 instead of a single SB notification.
• a check is performed to determine if the link in question is still a member of the LAG (step 1 12), which in the example of Node D would be the peer link connecting Nodes B and D. If the link is still a member of the LAG, then it is excluded from the LAG (step 1 14), and a LACPDU with a DSB notification is sent to the LAG peer, which would be Node B (step 1 16). Note that the method proposed here does not modify or spoil the handling of other failure events.
• a check is performed to determine if the fellow node is reachable (step 1 18). Thus, a check is performed to determine if Node D can communicate with Node C. If the nodes C, D can communicate, then a previous double split brain condition no longer exists. A check is therefore performed to determine if the peer link between Node B and Node D is currently part of the LAG (step 120), and if it is not a part of the LAG it is added to the LAG once again (step 122).
• the prior art single brain method in which Node B would change its System ID in the event of a failure
• the method 400 in which a single split brain condition is detected by key change
• the method 100 has been described as being implemented by the Node D (with Node D and Node C both expecting the same reference configuration information from their respective fellow nodes), for disabling of the non-primary connecting peer link between Nodes B and D (such that the primary peer link between Nodes A and C would be maintained), it is understood that this is only an example. If the method 100 was also performed by Node C, for example, Node C would still receive its expected information in steps 102-104. This is because Nodes A and C would continue using the same information (e.g., a same system ID and/or key), and it is only Nodes C and D that would be using different information. Therefore, the peer link between Node A and Node C would not be disabled even in the event of a double split brain. In one embodiment, all four of the nodes A, B, C and D perform the method 100 to provide single and double split brain detection at both LAG virtual nodes 30, 32.
• the reference configuration information could include an aggregation key, for example, to be used for information verification.
• configuration information could be determined, for example, by an initial configuration action, or it could be based on the values carried by LACPDU under normal operation (e.g., during initial LACPDU exchanges).
• Nodes C and D are both configured to store the expected key K of their peer. Information mismatch of the method described in Fig. 18 would occur if the key received in the LACPDU differs from the local expected key.
• Fig. 19 illustrates an exemplary method of operating a virtual node in a LAG to detect split brain conditions, in which the Configuration Information is a service assignment digest that indicates the assignment of services to physical LAG links.
• the digest may be received in a LACPDU, and the service digest may differ from an expected service digest.
• the double split brain handling mechanism in Fig. 18 relies on detection of mismatch in the received and the expected service-to-link assignment.
• Steps 202-212 and 214-222 of Fig. 19 are performed in a similar same fashion as steps 102-1 12 and 214-222 from Fig. 18, except that the expected information corresponds to a service assignment digest (as opposed to a key or System ID).
• an additional step 213 is included in the method 200 for determining whether the LAG peer link in question (e.g., the link B-D connecting Node B and Node D) is the default LAG link.
• This step ensure that only the non-default link (i.e., that between Nodes B, D) will be deactivated, and the primary link (i.e., that between Nodes A, C) will not be deactivated.
• the link has to be excluded from the LAG unless it is the default LAG link.
• the default LAG link is the link having the numerically lowest link ID value, unless another link is configured as the Default LAG link.
• Fig. 25 illustrates a generalized method 400 of operating a virtual node in a LAG comprising a first virtual node and a second virtual node.
• LAG virtual node 32 is the illustrative "first virtual node”
• LAG virtual node 30 will be the illustrative "second virtual node", although these could reversed in different situations.
• the first fellow node e.g., node C
• receives, from the second virtual node 30, first control information comprising a system ID and first configuration information associated with the LAG step 402).
• the first configuration information is compared against reference configuration information associated with the LAG (step 404).
• Step 404 there is a determination of whether the node (node C) is in communication with its fellow node (e.g., node D) (step 406). If not, then a double split brain condition is determined to exist in the LAG (step 410). If the fellow nodes C,D are in communication, then a single split brain condition is determined to exist in response to the second fellow node being in communication with the second virtual node (step 407) and the first configuration information not matching second configuration information provided to the second fellow node by the second virtual node (step 409).
• Step 408 represents detection of a node failure (e.g., node A).
• the reference configuration may represent previously established expected configuration information associated with the LAG.
• the determining that the split brain condition exists may include determining that a dual split brain condition exists in the LAG in response to detecting that the first fellow node is not in communication with the second fellow node (step 406).
• a non-primary or non- default connecting link that connects the first and second virtual nodes from the LAG may be excluded if a dual split brain condition is determined to exist in the LAG.
• the determining that the split brain condition exists may include determining that a single split brain condition exists in the LAG in response to the second fellow node being in communication with the second virtual node and the first configuration information not matching second configuration information provided to the second fellow node by the second virtual node (step 407, 409).
• the Configuration Information may be either a configuration key or a configuration digest, and may be carried in a Link Aggregation Control Protocol Data Unit (LACPDU) message if desired.
• LACPDU Link Aggregation Control Protocol Data Unit
• Figs. 20-21 are block diagrams of double split brain handling in a LAG, based on System IDs, using the method 100.
• the link connecting Node A and Node B has been lost, and Node B starts using a new System ID.
• the connection between Nodes C and D has been lost, and Node D starts using a new System ID.
• Nodes B and D therefore do not receive their expected information (steps 102, 104).
• each of Nodes B and D are unable to reach their fellow Nodes A and C, respectively (step 106).
• a double split brain condition is detected (step 1 10), and a LACPDU with a double split brain indication is sent from Node B to Node D, and a similar LACPDU is sent from Node D to Node B (step 1 16) (see “DSB" notifications of Fig. 21 ).
• each of these nodes receives the same System ID from their LAG peer (i.e., from each other), that is the locally expected basic System ID. Therefore, the peer link between nodes A and C is kept in the LAG group, and the peer link A-C continuously provides connectivity between Network 1 and Network 2.
• the System ID received in LACPDUs by nodes B and D does not match with the locally expected basic System ID, and Nodes B and C have lost connectivity to their fellow portal node. Therefore, nodes B and D declare a double split brain situation and remove the peer link B-D between them from the LAG group (see steps 1 12-1 14 discussed above).
• Nodes B and D may optionally send a DSB notification to each other (step 1 16). Note that the removal of the link B-D from the LAG group is a local decision, and does not require any agreement with any other node.
• Fig. 22 illustrates the handling of double split rain by the proposed methods 100, 200 in another example, where a full mesh topology is used among the nodes participating in the LAG. In this example, it is only the link between nodes A and C that is kept in the LAG group due to the methods proposed here. All the other links are deactivated by removing them from the LAG group, thus ensuring loop-free connectivity (see dotted lines between the nodes A, B, C, D).
• the connectivity formerly provided by the LAG can be broken due to double split brain if the proposed method is applied without careful configuration.
• the System ID received in the LACPDUs does not match the locally expected basic System ID in either of the LAG virtual nodes, and neither node can reach its fellow portal node. Therefore, both links are removed from the LAG group, and no connectivity is provided between Network 1 and Network 2.
• the nodes A, B and also the nodes C, D may be configured to use a common System ID as discussed above that does not change in the event of failures.
• Fig. 26 illustrates an exemplary LAG node 800 operative to implement the methods described above.
• the LAG node 800 includes a processor 802 operable to implement one or more of the methods described above.
• the processor 802 is operatively connected to non- transitory memory 804 and at least one input/output device 806 for communicating with its fellow LAG node (within the same LAG virtual node 30 or 32), and with its peer LAG node (in the remote other LAG virtual node).
• the memory 804 may be used to store the Configuration Information discussed above, for example.
• the processor 802 may be implemented by one or more processing circuits (e.g., microprocessors, hardware, firmware, or a combination thereof).
• the memory 804 may include program instructions for Split Brain Handling, that when run on a LAG node configures the LAG node to perform one or more of the methods described above.
• Some exemplary single split brain handling methods proposed above can avoid changing the System ID in case of any failure event, and can therefore meet DRNI requirements by hiding portal internal details from a LAG peer.
• Some exemplary double split brain handling methods proposed above prevent frame looping that would otherwise be caused by a DSB condition in a LAG, where at least four nodes participate in the LAG. Furthermore, the proposed methods do not influence the handling of other failure scenarios.

Landscapes

• Engineering & Computer Science (AREA)
• Computer Networks & Wireless Communication (AREA)
• Signal Processing (AREA)
• Environmental & Geological Engineering (AREA)
• Data Exchanges In Wide-Area Networks (AREA)
• Debugging And Monitoring (AREA)

Priority Applications (7)

Applications Claiming Priority (2)

Related Child Applications (2)

Publications (1)

Family

ID=47137985

Family Applications (1)

Country Status (6)

Cited By (7)

Families Citing this family (78)

Citations (2)

Family Cites Families (12)

• 2012
  • 2012-08-29 EP EP12781160.2A patent/EP2850787B1/en not_active Not-in-force
  • 2012-08-29 WO PCT/IB2012/054444 patent/WO2013171552A1/en not_active Ceased
  • 2012-08-29 IN IN8942DEN2014 patent/IN2014DN08942A/en unknown
  • 2012-08-29 KR KR1020147035112A patent/KR102043071B1/ko not_active Expired - Fee Related
  • 2012-08-29 US US14/400,619 patent/US9602348B2/en not_active Expired - Fee Related
  • 2012-08-29 JP JP2015512140A patent/JP6046241B2/ja not_active Expired - Fee Related
• 2017
  • 2017-03-20 US US15/463,292 patent/US9906411B2/en active Active
• 2018
  • 2018-02-08 US US15/891,875 patent/US10567224B2/en active Active

Patent Citations (2)

Non-Patent Citations (3)

Also Published As

Similar Documents

Legal Events