N-node virtual link trunking (VLT) systems fault management
Aspects of the present invention include an N-Node virtual link trunking (VLT) system comprising a set of N nodes collectively provide a logical fabric-level view that is consistent across the set of N nodes. Embodiments of the N-Node VLT system comprise a control plane mechanism to provide Layer 2 multipathing between access network devices (switches or servers) and the core network. The N-Node VLT system provides a loop-free topology with active-active load-sharing of uplinks from access to the core. Accordingly, the N-Node VLT system eliminates the disadvantage of Spanning Tree Protocol (STP) (active-standby links) by allowing link aggregation group (LAG) terminations on multiple separate distribution or core switches and also supporting a loop-free topology. Additional benefits of an N-Node VLT system include, but are not limited to, higher resiliency, improved link utilization, and improved manageability of the network.
1. A method for handing a failure of a link of a virtual link trunking (VLT) group in a n-node VLT system that comprises a plurality of nodes communicatively coupled via inter-node links (INLs) and at least some of the plurality of nodes of the n-node VLT system being communicatively coupled via one or more input/output (I/O) ports to one or more client devices to form one or more VLTs, the method comprising:
responsive to a VLT link failure of a VLT on a first node being detected:
assigning the first node as a topology-change-owner node, the first node performing the steps comprising:
determining one or more topology tables entries that require ownership change; and
sending one or more messages to one or more nodes in the n-node VLT system to assume ownership for a certain table entry or entries; and
responsive to the first node having no links for the VLT being operational on the first node as a result of the VLT link failure:
computing an assigned node for the first node, an assigned node being a node in the n-node VLT system has an active link or links for the VLT, wherein data traffic that is intended for the VLT and received at the first node is sent via an INL to the assigned node for egress; and
updating one or more port block masks for the n-node VLT system to avoid loops; and
synchronizing topology tables entry or entries that have changed as a result of the VLT link failure to maintain a consistent representation across all nodes in the n-node VLT system.
2. The method of claim 1 wherein the one or more port block masks are part of one or more topology tables and the one or more topology tables further comprising:
a virtual link trunking (VLT) membership table that comprises a listing of which nodes in the n-node VLT system participate in which of the one or more VLTs;
an assigned node table that comprises an assigned node for each node that may receive ingress traffic on the node that is destined for a VLT that does not have a link on the node but does have a link or links on one or more of the other nodes in the n-node VLT system, the assigned node having a link for the VLT;
a VLT-to-INL table that, for a node, maps a VLT to an inter-node link (INL) based upon the assigned node for the node; and
a system-wide egress mask table that comprises, for each assigned node assignment, a mapping of allowed VLT egress for data ingressing on a specific INL between network nodes in the n-node VLT system.
3. The method of claim 2 wherein the one or more topology tables, upon synchronization, are common to all nodes in the n-node VLT system.
4. The method of claim 1 wherein the first node performs the steps of:
computing an assigned node for the first node; and
updating one or more port block masks for the n-node VLT system to avoid loops.
5. The method of claim 1 wherein one or more of the nodes in the n-node VLT system pre-computes updates in anticipation of failure of the VLT link.
6. The method of claim 1 wherein one or more of the nodes sends a status update message to one or more of the other nodes in the n-node VLT system to detect a VLT link failure.
7. The method of claim 1 further comprising:
responsive to the VLT link on the first node being restored, assigning the first node as a topology-change-owner node, the first node performing the steps comprising:
reverting the updated one or more port block masks for the n-node VLT system and the assigned node back their states before the VLT link failure; and
opening existing port blocks that were affected after the reverted one or more port block masks have been successfully installed first.
8. A method for handing a failure of an inter-node link (INL) of a virtual link trunking (VLT) group in a n-node VLT system that comprises a plurality of nodes communicatively coupled via inter-node links (INLs) and at least some of the plurality of nodes of the n-node VLT system being communicatively coupled via one or more input/output (I/O) ports to one or more client devices to form one or more VLTs, the method comprising:
responsive to detecting an INL failure between a first node and a second node in the n-node VLT system:
selecting one of the first and second nodes to be removed from the n-node VLT system based upon one or more criteria;
disabling VLT ports at the node that is to be removed from the n-node VLT system according to rules comprising:
responsive to a VLT having as its only member the node that is to be removed from the n-node VLT system, retaining that VLT at that node; and
responsive to a VLT having more node members including the node that is to be removed from the n-node VLT system, disabling that VLT at the node that is to be removed from the n-node VLT system;
disabling all INL ports, except one, at the node that is to be removed from the n-node VLT system, the one INL port that is retained being selected based upon one or more criteria; and
removing the selected node from the n-node VLT system in order to operate, based upon table programming, including port block mask, as a normal switch connected to one of the nodes in the n-node VLT system.
9. The method of claim 8 further wherein the table programming comprises:
a virtual link trunking (VLT) membership table that comprises a listing of which nodes in the n-node VLT system participate in the one or more VLTs;
an assigned node table that comprises an assigned node for each node that may receive ingress traffic on the node that is destined for a VLT that does not have a link on the node but does have a link or links on one or more of the other nodes in the n-node VLT system, the assigned node having a link for the VLT;
a VLT-to-INL table that, for a node, maps a VLT to an inter-node link (INL) based upon the assigned node for the node; and
a system-wide egress mask table that comprises, for each assigned node assignment, a mapping of allowed VLT egress for data ingressing on a specific INL between network nodes in the n-node VLT system.
10. The method of claim 9 further comprising:
updating the assigned node table as a result of the INL failure and the removal of the selected node.
11. The method of claim 10 further comprising:
to avoid potential loops, opening existing port blocks only after new blocks are first installed.
12. The method of claim 9 further comprising:
assigning the node that is retained in the n-node VLT system as a topology-change-owner node.
13. The method of claim 12 wherein:
the topology-change-owner node changes ownership of affected table entries to a node in the n-node VLT system according to one or more criteria.
14. The method of claim 13 wherein the one or more criteria comprises selecting, for an affected table entry, the node that has a lowest MAC address.
15. The method of claim 9 wherein the table programming, upon synchronization, are common to all nodes in the n-node VLT system.
16. The method of claim 8 further comprising:
responsive to the INL between the first and second nodes being restored:
assigning a node as a topology-change-owner node;
reverting the VLT ports and table programming to their states as they were before the VLT link failure; and
returning the node that was removed from the n-node VLT system back to the n-node VLT system, wherein existing port blocks that were affected are opened after the reverted table programming has been successfully installed first.
17. A method for handing a failure of a node in a n-node virtual link trunking (VLT) system that comprises a plurality of nodes communicatively coupled via inter-node links (INLs) and at least some of the plurality of nodes of the n-node VLT system being communicatively coupled via one or more input/output (I/O) ports to one or more client devices to form one or more VLTs, the method comprising:
responsive to a node failure being detected:
assigning one of the remaining nodes in the n-node VLT system as a topology-change-owner node, the topology-change-owner node performing the steps comprising:
determining one or more topology tables entries that require ownership change; and
sending one or more messages to one or more nodes in the n-node VLT system to assume ownership for a certain table entry or entries;
computing assigned nodes for the remaining nodes in the n-node VLT system, an assigned node being a node in the n-node VLT system has an active link or links for a VLT, wherein data traffic that is intended for the VLT and received at the first node is sent via an INL to the assigned node for egress;
updating one or more port block masks for the n-node VLT system to avoid loops; and
synchronizing topology tables entry or entries that have changed as a result of the node failure to maintain a consistent representation across the remaining nodes in the n-node VLT system.
18. The method of claim 17 wherein the one or more port block masks are part of one or more topology tables and the one or more topology tables further comprise:
a virtual link trunking (VLT) membership table that comprises a listing of which nodes in the n-node VLT system participate in the one or more VLTs;
an assigned node table that comprises an assigned node for each node that may receive ingress traffic on the node that is destined for a VLT that does not have a link on the node but does have a link or links on one or more of the other nodes in the n-node VLT system, the assigned node having a link for the VLT;
a VLT-to-INL table that, for a node, maps a VLT to an inter-node link (INL) based upon the assigned node for the node; and
a system-wide egress mask table that comprises, for each assigned node assignment, a mapping of allowed VLT egress for data ingressing on a specific INL between network nodes in the n-node VLT system.
19. The method of claim 17 wherein one or more of the nodes in the n-node VLT system pre-computes updates in anticipation of failure of the VLT link.
20. The method of claim 17 further comprising:
responsive to the node being restored:
assigning a node as a topology-change-owner node;
reverting the updated one or more port block masks for the n-node VLT system and the assigned node back their states before the node failure; and
opening existing port blocks that were affected after the reverted one or more port block masks have been successfully installed first.