CN114223182B - Method and system for isolating leaf switches in a network - Google Patents

Abstract

The present disclosure provides methods and systems for isolating a leaf switch in a network that is connected to a server in the network. The method may include: in response to receiving a request to isolate a leaf switch in a network, sending a notification to a server through the leaf switch, wherein the notification instructs the server to stop sending egress traffic to the leaf switch; determining whether an acknowledgement of the notification is received from the server; and stopping ingress traffic towards the server in response to the determination that the acknowledgement has been received.

Description

Method and system for isolating leaf switches in a network

Background technique

Server centers (such as large-scale scalable data centers) can include multiple network servers and switches to provide zero-offline services, including remote storage services, cloud processing services, large-scale data distribution, etc. High availability (HA) is critical in massively scalable data centers (MSDC) due to zero downtime requirements.

To provide high-bandwidth, low-latency, and non-blocking connections, MSDC makes extensive use of Clos network topology. Clos networks can be based on a spine-leaf topology, including multiple spine switches and multiple leaf switches. In a spine-leaf topology, a leaf switch can connect to all spine switches for increased resiliency and scalability, as well as to multiple servers. A server can also be connected to multiple leaf switches. Due to hardware failure or software upgrades, leaf switches may have to be isolated from the Clos network for maintenance or upgrades. However, isolation of leaf switches can cause real-time traffic drops, resulting in undesirable service outages.

Contents of the invention

Embodiments of the present invention provide a method for isolating a first leaf switch in a network that is connected to a server in the network. The method may include: in response to receiving a request from a first leaf switch in the isolation network, sending a notification to the server via the first leaf switch, the notification instructing the server to stop sending egress traffic to the first leaf switch; determining whether to receive to an acknowledgment of the notification; and in response to a received determination that the acknowledgment has been accepted, ceasing ingress traffic toward the server.

The disclosed embodiments further provide a first leaf switch connected to servers in the network. The first leaf switch may include a memory storing a set of instructions; and at least one processor coupled to the memory and configured to execute the set of instructions, such that the first leaf switch performs: in response to receiving the isolation a request from a switch in the network to send a notification to the server instructing the server to stop sending egress traffic to the first leaf switch; determine whether an acknowledgment of the notification is received from the server; and in response to a determination that the acknowledgment has been received, stop Ingress traffic towards said server.

Embodiments of the present invention also provide a non-volatile computer-readable medium that stores a set of instructions executable by at least one processor of the leaf switch to cause the leaf switch to perform a method of isolating leaves in a network. switch method. Leaf switches can be connected to servers in the network. The method may include: in response to receiving a request to isolate the first leaf switch in the network, sending a notification to the server through the first leaf switch, the notification instructing the server to stop sending egress traffic to the first leaf switch; determining whether a response to the request is received from the server. acknowledgment of the notification; and in response to a determination that the acknowledgment has been received, ceasing ingress traffic toward the server.

Description of drawings

Embodiments and various aspects of the present disclosure are set forth in the following detailed description and accompanying drawings. The various features shown are not to scale.

Figure 1 shows a schematic diagram of the Clos network.

Figure 2 shows a schematic diagram of an example network in accordance with certain embodiments of the present disclosure.

Figure 3 shows a schematic diagram of the isolated network according to certain embodiments of the present disclosure.

Figure 4 is a flowchart of a method of isolating a first leaf switch in a network according to certain embodiments of the present disclosure.

Figure 5 illustrates a block diagram of an example leaf switch in accordance with some embodiments of the present disclosure.

Detailed ways

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. Where possible, use the same numbering throughout the drawing to refer to the same or similar parts.

As mentioned above, with legacy systems, isolating leaf switches from the Clos network can disrupt the network, causing live traffic to be lost, for example. The techniques described in this disclosure can minimize these types of disruptions.

As used herein, the terms "comprises," "comprises," or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, composition, article, or apparatus that includes a list of elements not only includes those elements, but may include other Other elements expressly listed or inherent in such process, method, composition, article or apparatus. The word "exemplary" means "example" rather than "ideal".

Figure 1 shows a schematic diagram of network 100. Although network 100 is envisioned as a Clos network, it is noted that any network with at least a three-layer architecture may be used.

As shown in Figure 1, the Clos network 100 is a three-layer architecture, including a spine layer 110, a leaf layer 120 and a server layer 130. The spine layer 110 is the backbone of the Clos network 100 and is responsible for interconnecting all leaf switches in the leaf layer 120. It may include multiple spine switches (such as spine switches 112, 114, 116, and 118). Leaf layer 120 may provide access to devices such as servers and include a plurality of leaf switches (eg, leaf switches 122, 124, 126, and 128). Server tier 130 may include multiple servers (eg, servers 132, 134, 136, and 138).

In this three-tier architecture, multiple leaf switches can be connected to multiple spine switches in a network-wide topology. In other words, each leaf switch (eg, leaf switch 122) is connected to each spine switch (eg, spine switches 112, 114, 116, and 118) in spine layer 110 to create multiple links. Links between leaf switches (eg, 122) and spine switches may be randomly selected among the plurality of links so that the traffic load between leaf layer 120 and spine layer 130 is evenly distributed. The connections between these leaf switches and spine switches may also be called L3 connections.

Each leaf switch (eg, leaf switch 122) may also be connected to at least one server (eg, server 132 and server 134) in server tier 130. On the other hand, each server (eg, server 132) may be connected to at least two leaf switches (eg, leaf switches 122 and 124) to ensure connectivity. In other words, server 132 may use leaf switch 122 to establish a first link and leaf switch 124 to establish a second link, for example. The first link and the second link between the server and the leaf switch may be called L2 links.

Under this three-layer architecture of the Clos network 100, if the Clos network 100 is overloaded, the process of expanding the capacity of the Clos network 100 can be very simple. For example, an additional spine switch can be added and connected to each leaf switch to provide additional inter-layer bandwidth between spine layer 110 and leaf layer 120 to reduce the overload.

Similarly, you can add a new leaf switch simply by connecting it to each spine switch. However, when existing leaf switches are isolated from the Clos network, this isolation can cause unnecessary service interruptions. For example, according to existing technology, the L2 link between a leaf switch and a server can be shut down from one side of the leaf switch without the server being aware of the shutdown. Therefore, the server may continuously send traffic to the leaf switch until the server detects the shutdown and switches the traffic to another L2 link. Therefore, traffic sent to the leaf switches before the switchover will never be processed and will have to be dropped, causing undesirable business interruption.

Embodiments of the present disclosure further provide methods and systems for isolating leaf switches in a network to minimize traffic disruption.

Figure 2 shows a schematic diagram of an example network 200 in accordance with some embodiments of the invention. As shown in FIG. 2 , network 200 may include spine switches 212 and 214 , leaf switches 222 and 224 , and server 232 . Each leaf switch 222 and 224 is connected to a spine switch 212 and 214. Server 232 is connected to leaf switches 222 and 224. The connection between server 232 and leaf switch 222 may be referred to as a first L2 link, and the connection between server 232 and leaf switch 224 may be referred to as a second L2 link.

In some embodiments, leaf switch 222 may receive a request to isolate leaf switch 222 from network 200 . For example, an administrator of network 200 may make a request for maintenance, software upgrades, etc. Note that requests can also be made via the network 200 itself. For example, when the network 200 detects that the leaf switch 222 fails, the network 200 can automatically request to isolate the switch 222 to avoid further service interruption.

In some embodiments, when leaf switch 222 receives a quarantine request, leaf switch 222 can determine whether the second L2 link associated with leaf switch 224 has sufficient bandwidth to handle the traffic associated with leaf switch 222 .

If the second L2 link associated with leaf switch 224 cannot handle the additional traffic associated with leaf switch 222, leaf switch 222 may indicate that isolation cannot be performed. For example, leaf switch 222 may generate a message informing the administrator of network 200 that quarantine is temporarily unavailable. It should be noted that due to the decrease in traffic on the second L2 link, the bandwidth of the second L2 link can be released. For example, the second L2 link may not have sufficient bandwidth to handle the traffic associated with leaf switch 222 at the first moment, but may have appropriate bandwidth at the second moment. Therefore, in some embodiments, the message generated by leaf switch 222 may further indicate another time to perform this isolation.

If the second L2 link associated with leaf switch 224 is unable to handle the additional traffic associated with leaf switch 222, leaf switch 222 may continue to process quarantine requests. For example, in response to a request to isolate leaf switch 222 , leaf switch 222 may send notification 202 to server 232 . Notification 202 may be used to notify server 232 that egress traffic to leaf switch 222 should stop. In some embodiments, notification 202 may include an identification of leaf switch 222 (eg, a media access control (MAC) address). After receiving the notification 202, the server 232 may stop sending egress traffic to the leaf switch 222. Egress traffic that would have been sent to leaf switch 222 can now be sent to another leaf switch (such as leaf switch 224) and, therefore, ultimately reach the spine.

Notably, when egress traffic to leaf switch 222 ceases, ingress traffic toward server 232 may be continuously sent by leaf switch 222.

Server 232 may then send an acknowledgment 204 of notification 202 to leaf switch 222. Acknowledgment 204 may notify leaf switch 222 that server 232 is aware of the request. In some embodiments, acknowledgment 204 may further notify leaf switch 222 that egress traffic has been sent to another leaf switch (eg, leaf switch 224).

Link Aggregation Control Protocol (LACP) can be used to manage L2 links and communication between leaf switches and servers. LACP allows network devices (such as leaf switches 222) to negotiate automatic link binding by sending LACP data units (LACPDUs) to peer devices (such as servers 232) that also perform LACP. . In some embodiments, notifications 202 and acknowledgments 204 may be transmitted between leaf switches 222 and servers 232 using LACPDUs.

Accordingly, leaf switch 222 may determine whether the acknowledgment 204 was received from server 232 . In certain embodiments, leaf switch 222 may further determine whether the acknowledgment 204 is received from server 232 within a given time (eg, 3 seconds). For various reasons, the leaf switch 222 may not receive the acknowledgment 204 within a given time. For example, these reasons may include at least one of leaf switch 222 not being able to send the notification 202 , server 232 not being able to receive the notification 202 , server 232 not being able to send the acknowledgment 204 , leaf switch 222 not being able to receive the acknowledgment 204 , and so forth.

In response to the determination that the acknowledgment 204 was not received, the leaf switch 222 may resend the notification 202 to the server 232 to further notify the server 232 of the quarantine request.

In response to determining that the acknowledgment 204 was received, the leaf switch 222 may stop ingress traffic toward the server 232 . Therefore, the server 232 has stopped egress traffic to the leaf switch 222 before determining that the acknowledgment 204 has been received. And upon determining that the acknowledgment 204 is received, the leaf switch 222 may stop ingress traffic to the server 232 . In other words, after receiving the acknowledgment 204, the traffic between the leaf switch 222 and the server 232 (including the egress traffic of the server 232 and the ingress traffic of the server 232) may be completely stopped.

It is worth noting that before the acknowledgment 204 is received, some ingress traffic towards the server 232 may still be sent and become live traffic. Therefore, leaf switch 222 preferably does not disconnect from server 232 until real-time traffic toward server 232 is processed. If leaf switch 222 immediately disconnects from server 232 after receiving acknowledgment 204, the live traffic may not be processed thoroughly.

Because the duration of real-time traffic processing may be short, leaf switch 222 may disconnect from server 232 for a period of time after ingress traffic toward the server at leaf switch 222 is stopped. The time period is configurable and can be set to a few milliseconds. It is worth noting that although leaf switch 222 has been isolated from server 232, traffic between the spine switch and server 232 of the spine layer can communicate through leaf switch 224 after leaf switch 222 is disconnected.

In some embodiments, leaf switch 222 may also use the last packet of traffic to server 232 as an acknowledgment. In response to receiving the confirmation, server 232 may further confirm that all traffic has been processed and that it is safe to disconnect leaf switch 222 .

It can be understood that when the leaf switch 222 comes back online, the L2 link can be re-established between the leaf switch 222 and the server 232 .

As can be seen from the above, during the isolation period of the leaf switch 222, the notification 202 and the confirmation 204 can be used to coordinate the sequential termination of the egress traffic and the ingress traffic of the server 232, so that the real-time traffic between the leaf switch 222 and the server 232 can be terminated before isolation. Processed to avoid traffic interruption.

As mentioned above, LACP can be used to manage links and communications between leaf switches and servers. In some embodiments, LACP may further be used to perform sequential termination of egress traffic and ingress traffic to the server.

In some embodiments, the LACP Port Status field of the LACPDU may be used as a synchronization field to coordinate leaf switches and servers during isolation. The LACP port status field contains at least three bits, each bit is a flag indicating the specific status of the sending port. Table 1 below shows example meanings of the three bits of the LACP port status field, including "synchronize", "collect", and "distribute".

Table 1

The "sync" bit can be used to indicate whether the sending device and the receiving device are in sync. As shown in Table 1 above, if the "synchronization" bit is "0", it means that the receiving end and the transmitting end are not synchronized, and the receiving end device can resynchronize multiple receiving end and transmitting end physical ports. Resynchronization may also be called "flapping". After synchronization, physical ports can be aggregated into a high-bandwidth data path to provide better connectivity. The aggregated physical port may also be called a link aggregation group (LAG).

If the "Synchronization" bit is "1", it means that the receiving end and the transmitting end are synchronized, and at least one collection and distribution can be performed. As shown in Table 1, when the three bits of the port status field are "101", it means that the sending device is sending traffic to the receiving end and expects the sending end to stop sending traffic. More specifically, in this embodiment of the present disclosure, the three port status bits of the leaf switch are "101", indicating that the leaf switch is still sending ingress traffic to the servers connected to the leaf switch and expects each server to stop Send egress traffic to the leaf switch. For example, when leaf switch 222 sends notification 202 to server 232 using LACPDU, leaf switch 222 may also set the LACP port status field to "101". Therefore, upon receiving the notification 202, the server 232 can read the port status of the leaf switch and stop egress traffic to the leaf switch and send a confirmation (confirmation 204 in Figure 2).

In response to notification 202, when server 232 sends acknowledgment 204, the three bits of server 232's port status field may be set to "110", indicating that server 232 is still receiving traffic from leaf switch 222 and expects no traffic from leaf switch 222 Confirmation of transmission. Because server 232 continuously handles real-time traffic on the link between leaf switch 222 and server 232, real-time traffic can be processed from the link, thereby avoiding traffic interruption, or minimizing traffic when leaf switch 222 is isolated. Interrupt.

After leaf switch 222 receives acknowledgment 204, leaf switch 222 may further confirm that there is no traffic transmission on the link between leaf switch 222 and server 232. For example, the last packet of real-time traffic sent by leaf switch 222 may serve as the acknowledgment.

At this point, leaf switch 222 is ready to disconnect from the server. According to some embodiments of the present invention, FIG. 3 shows a schematic diagram of the isolated network 300. As shown in Figure 3, after leaf switch 222 receives the acknowledgment, leaf switch 222 has disconnected from server 232. It can be understood that traffic sent by server 232 can still reach spine switches 212 and 214 through leaf switch 224. Therefore, users of network 300 are not aware of the isolation of leaf switch 222.

Referring again to FIG. 2 and Table 1, when the three bits of the synchronization field of the leaf switch 222 and the server 232 are "111", it may indicate that the leaf switch 222 and the server 232 are conducting bidirectional communication.

It can be understood that when a certain leaf switch receives a request to isolate the leaf switch from the network, the above multiplexing of the LACP port status field of the LACPDU can be activated. It is also understood that the synchronization field may be transmitted between the leaf switch and the server using three different bits of the LACP Port Status field of the LACPDU. In some embodiments, the synchronization field may be transmitted using data units other than LACPDU.

Figure 4 is a flowchart of a method 400 of isolating a first leaf switch in a network according to some embodiments of the invention. In addition to the first leaf switch, the network may also include second leaf switches and spine switches. Both the first and second leaf switches may be connected to a server (eg, server 134 of server tier 130), with the spine switch connected to the first and second leaf switches. Method 400 may be performed by an electronic device. The electronic device may include a memory that stores a set of instructions and at least one processor that executes the set of instructions to cause the electronic device to perform method 400 . For example, the electronic device may be a leaf switch (eg, leaf switch 222 in Figure 2-3) of a leaf layer (eg, leaf layer 120). Referring to Figure 4, method 400 may include the following steps.

At step 402, in response to receiving a request to isolate a first leaf switch in the network, the first leaf switch may send a notification to the server. The notification may instruct the server to stop sending egress traffic to the first leaf switch. In some embodiments, the notification may also include the first port status of the first leaf switch. For example, the notification may be carried by the first LACPDU (Link Aggregation Control Protocol Data Unit), and the first port status may be represented by the LACP port status field of the first LACPDU. In some embodiments, the LACP port status field may include 3 bits to indicate the port status of the sending end (eg, the first leaf switch). In this step 402, the first LACP port status field of the first LACPDU may be "101", indicating that the port of the first leaf switch is still distributing services but not receiving services, and the first leaf switch expects the server to stop moving towards the first Egress traffic from the leaf switch. Therefore, after sending the notification, the first leaf switch can continuously send ingress traffic to the server for processing.

At step 404, the first leaf switch may determine whether an acknowledgment of the notification was received from the server. When the server receives the notification, the server can stop sending egress traffic to the first leaf switch and send an acknowledgment back to the first leaf switch. The confirmation information can be carried by the second LACPDU. Similarly, the server's second port status can also be represented by the LACP port status field of the second LACPDU. In this step 404, the LACP port status field of the second LACPDU may be "101", indicating that the server's port is still receiving traffic from the first leaf switch and expects confirmation of no traffic transmission from the first leaf switch.

As mentioned earlier, the servers are also connected to the second leaf switch. In some embodiments, the notification may further cause egress traffic from the server to be sent to the spine switch through the second leaf switch.

In step 406, in response to a determination that the acknowledgment was received, the first leaf switch may stop ingress traffic toward the server.

In some embodiments, in response to the determination that the acknowledgment was not received within the first period of time in step 404, the first leaf switch may send another notification to the server in step 402. It is worth noting that if notifications are sent a given number of times, the first leaf switch can generate an error code indicating that isolation has failed.

In step 408, the first leaf switch may disconnect the first leaf switch from the server. Traffic between the spine switch and the server communicates through the second leaf switch after disconnecting the first leaf switch. In some embodiments, the first leaf switch may disconnect the first leaf switch from the server within a second period of time after ingress traffic toward the server is stopped at the first leaf switch. In some embodiments, when the server receives the last packet of ingress traffic, the server may further confirm that the last packet has been processed. In response to the acknowledgment, the first leaf switch may disconnect the first leaf switch from the server.

Figure 5 shows a block diagram of an example leaf switch 500 in accordance with some embodiments of the present disclosure. Leaf switch 500 may be connected to a server in the network and configured to perform method 400. The network may also include a spine switch.

Leaf switch 500 may include a plurality of network ports 502a-502n, memory 504, and a processor 506 coupling the plurality of network ports 502a-502n and memory 504.

Network ports 502a-502n can be used to send and receive traffic from spine switches and servers. Memory 504 may store a set of instructions for performing method 400. In addition, the memory 504 may also store an address lookup table including the addresses of devices in the network and the addresses of corresponding ports. Processor 506 may execute a set of instructions causing leaf switch 500 to perform method 400.

Embodiments of the present disclosure also provide a computer program product. The computer program product may include a non-volatile computer-readable storage medium having computer-readable program instructions for causing the processor to perform the above method.

A computer-readable storage medium may be a tangible device that stores instructions for use by an instruction execution device. For example, the computer-readable storage medium may be, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of computer-readable storage media includes the following more specific examples: portable computer disks, hard drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), Static random access memory (SRAM), flash memory, portable compact disc read-only memory (CD-ROM), digital versatile disk (DVD), memory stick, floppy disk, a mechanical encoding device such as a punched card or bumps in a slot A structure on which instructions are recorded, and any appropriate combination of the foregoing.

Computer-readable program instructions for performing the above methods may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or use one or more Source code or object code written in any combination of programming languages, including object-oriented programming languages and conventional procedural programming languages. The computer-readable program instructions may execute entirely on a computer system, as a stand-alone software package, or execute partially on a first computer and partially on a second computer remote from the first computer. In the latter scenario, the second remote computer can be connected to the first computer over any type of network, including a local area network (LAN) or a wide area network (WAN).

Computer-readable program instructions may be provided to one or more processors of a computer or other programmable data processing apparatus to make a machine such that the instructions execute by the one or more processors of the computer or other programmable data processing apparatus , create a method that implements the above method.

The flowcharts and schematic diagrams in the Figures illustrate exemplary architecture, functionality, and operations of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present specification. In this regard, a block in the flowchart or schematic diagram may represent a software program, segment, or portion of code, which includes one or more executable instructions for implementing the specified functions. It should also be noted that in some alternative implementations, functions documented in blocks may appear out of the order documented in the figures. For example, two blocks shown in succession may actually be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order depending on the functionality involved. It will also be noted that each block of the figures and/or flowchart illustrations, and combinations of blocks in the figures and flowchart illustrations, can be constructed by special purpose hardware-based systems that perform the specified functions or actions, or by combinations of special purpose hardware and computer instructions. accomplish.

It is to be understood that, for clarity, certain features of the specification that are described in the context of separate embodiments can also be provided in combination in a single embodiment. Conversely, various features of the specification that are, for the sake of brevity, described in the context of a single embodiment, may also be provided separately, or in any suitable subcombination of the specification, or in any other described embodiment of the specification. Certain features described in the context of various embodiments should not be considered essential features of those embodiments, except that the embodiment would not function without these elements.

Although specific embodiments have been described in connection with the specification, it will be apparent that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, this disclosure includes all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.

Claims (16)

1. A method for isolating a first leaf switch in a network connected to a server in the network, comprising: In response to receiving a request to isolate a first leaf switch in the network, sending a notification to the server through the first leaf switch, instructing the server in the notification to stop sending egress traffic to the first leaf switch; Determine whether acknowledgment of the notification is received from the server; and In response to the determination that the acknowledgment has been received, stopping ingress traffic toward the server; Disconnecting the first leaf switch from the server includes: disconnecting the first leaf switch from the server within a second period of time after ingress traffic toward the server stops at the first leaf switch. Connection. 2. The method of claim 1, wherein the network further includes a second leaf switch connected to the server and a spine switch connected to the first leaf switch and the second leaf switch. 3. The method of claim 2, wherein the notification further causes egress traffic from the server to be sent to the spine switch through the second leaf switch. 4. The method according to any one of claims 1-3, wherein After sending the notification, the first leaf switch sends the ingress traffic to the server for processing. 5. The method of any one of claims 1-4, further comprising: In response to the determination that the acknowledgment was not received within the first time period, another notification is sent to the server. 6. The method of claim 3, wherein disconnecting the first leaf switch from the server causes traffic between the spine switch and the server to pass through the second leaf switch. communicate. 7. The method according to any one of claims 1-6, wherein the notification is carried by a first Link Aggregation Control Protocol Data Unit LACPDU and the confirmation is carried by a second LACPDU. 8. The method of any one of claims 1-7, wherein the notification includes a first port status of the first leaf switch and the acknowledgment includes a second port status of the server. 9. First leaf switches connected to servers in the network, including: A memory that stores a set of instructions; and At least one processor coupled to the memory and configured to execute a set of instructions to cause the first leaf switch to: In response to receiving the request to isolate the switch in the network, sending a notification to the server, the notification instructing the server to stop sending egress traffic to the first leaf switch; Determine whether an acknowledgment of the notification is received from the server; and Responsive to a determination that the acknowledgment has been received, stopping ingress traffic toward the server; Causing the first switch to disconnect the first leaf switch from the server includes disconnecting the first leaf switch from the server within a second period of time after ingress traffic toward the server stops at the first leaf switch. 10. The first leaf switch of claim 9, wherein the network further includes a second leaf switch connected to the server and a spine switch connected to the first leaf switch and the second leaf switch. 11. The first leaf switch of claim 10, wherein This notification further causes egress traffic from the server to be sent to the spine switch through the second leaf switch. 12. The first leaf switch of any one of claims 9-11, wherein at least one processor is further configured to execute a set of instructions to cause the first leaf switch to further: After sending the notification, the ingress traffic is sent to the server for processing. 13. The first leaf switch of any one of claims 9-12, wherein the at least one processor is further configured to execute a set of instructions to cause the first switch to further: In response to the determination that the acknowledgment was not received within the first time period, another notification is sent to the server. 14. The first leaf switch of claim 10, wherein disconnecting the switch from the server further comprises: After the first leaf switch is disconnected, traffic between the spine switch and the server is conveyed through the second leaf switch. 15. A first leaf switch according to any one of claims 9-14, wherein said notification is carried by a first Link Aggregation Control Protocol Data Unit LACPDU and said acknowledgment is carried by a second LACPDU. 16. A non-volatile computer-readable medium storing a set of instructions executable by at least one processor of a leaf switch to cause the leaf switch to perform a method for isolating the leaf switch in a network to which the leaf switch is connected server, the method includes: In response to receiving the request to isolate the first leaf switch in the network, sending a notification to the server via the first leaf switch, wherein the notification instructs the server to stop sending egress traffic to the first leaf switch; Determine whether an acknowledgment of the notification is received from the server; and Responsive to a determination that said acknowledgment is received, stopping ingress traffic toward the server; Disconnecting the first leaf switch from the server includes: disconnecting the first leaf switch from the server within a second period of time after ingress traffic toward the server stops at the first leaf switch. Connection.