- ARP
- iStack
- stack build
- Stack new members to join
- stack merge
- Stack member exits
- stack split
- MAD detection
- stack configuration
- CSS
- Overview and Features
- CSS cluster establishment
- Joining and merging of cluster members
- cluster split
- MAD detection
- cluster configuration
1. ARP
ARP (Address Resolution Protocol) address resolution protocol is a TCP/IP protocol for obtaining MAC addresses based on IP addresses.
There are four types of commonly used ARP: Forward ARP, Reverse ARP, Proxy ARP, and Gratuitous ARP.
Among them, the forward ARP is to find the MAC address according to the IP.
Reverse ARP is to find the IP address based on the MAC address.
Proxy ARP means that when a data packet passes through a router, the router will generate a proxy ARP to find the MAC address.
Free ARP is the sending method to send an ARP packet whose destination IP address is its own IP address. The host sends a free ARP to detect whether there is an address conflict. At the same time, it also sends a VRRP when the master and backup switch, after the master fails. Gratuitous ARP packets with priority 0 are sent to backup.
Typical campus network - CSS + Eth-Trunk + iStack
Stacking features: simple, efficient and reliable.
Simple
The devices on each layer use stacking technology, with fewer logical devices and a simple network topology. The second layer is naturally loop-free, and there is no need to deploy the xSTP loop-breaking protocol.
Efficient
Eth-Trunk link aggregation technology is used between devices at each layer, with flexible load sharing algorithms and high link utilization.
Reliable
Stacking technology is used in combination with link aggregation technology, and physical devices at each layer form a dual-homing access network to improve the reliability of the entire network.
2. iStack - device stacking
Concept: Intelligent stack iStack (IntelligentStack) refers to the combination of multiple switch devices that support the stacking feature, logically combining into one switch device . As shown in the figure, SwitchA and SwitchB are connected through stack cables to form a stack system. For upstream and downstream devices, they are equivalent to a switch.
Stacking supports stacking of multiple switches.
Basic concepts in stacking:
1. Role
All single switches in the stack are called member switches, which can be divided into three roles according to different functions:
Master switch (Master): responsible for managing the entire stack. There is only one master switch in the stack.
Standby switch (Standby): It is the backup switch of the main switch. When the master switch fails, the standby switch will take over all the services of the original master switch. There is only one standby switch in the stack.
Slave switch (Slave): mainly used for service forwarding, the more the number of slave switches, the stronger the forwarding capability of the stack system.Note: Except for the master switch and the standby switch, all other member switches in the stack are slave switches.
2. Stack ID
That is, the slot number (Slot ID) of the member switch, which is used to identify and manage the member switch, and the stack ID of all member switches in the stack is unique.
3. Stack priority
The stack priority is an attribute of a member switch, which is mainly used to determine the role of the member switch during the role election process. The larger the priority value, the higher the priority, and the higher the priority, the greater the possibility of being elected as the master switch.
4. Rules for role election in the stack:
1. When the device is running, if no other switch joins in the specified time, it is the master switch.
2. Switch priority, the bigger the better.
3. MAC address, the smaller the better.
The election of the standby switch is based on the priority and MAC address.
2.1. Stack establishment process:
1. Configure stacking parameters, such as priority, slot ID, etc.
2. The device is powered off.
3. Connect stacking cables. There are two stacking methods. One is service port stacking, which uses the RJ45 interface, which is the port of the switch. The second is stacking through stacking ports, using stacking cables and stacking ports for stacking.
4. Power on the device.
5. Automatically complete stacking, such as device role election.
Stack connection method:
There are two ways to build a switch stack according to the different stack ports: stack card stack and service port stack .
Stacking with stack cards can be divided into the following two situations:
1. Switches are connected through dedicated stack cards and dedicated stack cables.
2. The stacking card is integrated into the rear panel of the switch, and the switch is connected through the integrated stacking port and a dedicated stacking cable.
Service port stacking refers to the connection between switches through physical member ports bound to logical stack ports, without dedicated stack cards.
2.2. New stack members join
When a stack system needs to expand its performance, it is necessary to add new members. The joining process of its stack members is as follows:
1. A stack system is running normally.
2. The new member enables stacking and configures stacking parameters.
3. The new member is powered off.
4. Connect the new members to the stack system.
5. New members power on.
6. The new member becomes the slave switch in the stack system.
Note: The stacking system may form a ring or a chain when new members join.
2.3. Stacking and merging
We need to merge two stack systems that are running normally. The merge process is as follows:
1. Two stacking systems in normal operation
2. Two stacking systems are connected through stacking cables.
3. The master switches in the two stack systems compete.
4. Assuming that the master switch in the B stack system fails to compete, all devices in the B stack system restart.
5. After the B stack system restarts successfully, all devices become slave switches.
2.4. Stack member exits
When a member exits in a normally running stack system, the process is as follows:
The triggers for member withdrawal are:
1. Unplug the stack cable;
2. Close the stack port or physical member port;3. The stack member device restarts;
4. Other reasons such as member device failure.Handling of stack exit:
1. When the master switch in the stack exits: the stack is upgraded to the master switch. After the stack system is updated, it continues to run.
2. When the standby switch in the stack exits, re-elect the standby switch, and continue to run after the stack system is updated.
3. When the stack exits from the switch, the stack system continues to run after being updated.
2.5, stack split
When a normally running stack system is split into two different stack systems, the process is as follows:
1. After splitting, if the original stack master and stack standby are in the same stack system, the removed member switch resets, reorganizes the stack, and elects the master and backup; 2. After splitting,
if the original stack master and stack standby are not in the same stack system , the standby switch becomes the master switch, and two stack systems with the same configuration appear on the network.Note: When the stack splits, two stack systems with the same configuration will be generated, such as IP addresses, which will conflict, so there will be network faults, and we need multi-master detection to detect faults and deal with them.
2.6. MAD detection
MAD---multi-master detection is divided into two methods: direct connection detection method and proxy detection method.
Direct connection detection method: It is detected by directly connecting cables. The cable connection methods are divided into direct connection through intermediate devices and full-mesh direct connection of stack member switches:
Proxy detection mode: According to the different proxy devices, the proxy detection mode can be divided into single machine as proxy and two sets of stacked systems as mutual proxy.
Note: If a stack split is detected in the network, the stack system will have two states: detect and recovery.
When a stack split is detected, two stack systems start to compete, and the stack system with the higher priority will have a successful status of the stack system as detect. The stack system that fails the competition enters recovery, and the stack system that enters recovery shuts down all physical ports except reserved ports to ensure that the failed stack system does not forward packets. If the stacking system of the recovery is recovered, the stacking will be merged.
2.7. Stack configuration
1. Set up a stack by connecting stack cards
[SwitchA] stack slot 0 priority 200
//Configure the stack priority of member switches. By default, the stack priority of member switches is 100[SwitchB] stack slot 0 renumber 1
//Configure the stack ID of the device
[SwitchC] stack slot 0 renumber 2
2. Set up a stack through a service port connection
[SwitchA] interface stack-port 0/1
[SwitchA-stack-portO/1] port interface gigabitethernet 0/0/27 enable\\Configure the service port as a physical member port and add it to the logical stack port. Switches B and C are the same.
[SwitchA] interface stack-port 0/2
[SwitchA-stack-port0/2] port interface gigabitethernet 0/0/28 enable[SwitchA] stack slot 0 priority 200
//Set the stack slot priority of SwitchA to 200[SwitchB] stack slot 0 renumber 1I
//Set the stack ID of SwitchB to 1
[SwitchC] stack slot 0 renumber 2
3. CSS -- cluster
3.1. Overview and features
Cluster switch system CSs (Cluster Switchsystem), also known as a cluster, refers to the combination of two switch devices that support the cluster feature, logically combined into one switch device.
Note: The cluster supports only two modular switches for clustering, so the roles of the switches are only active and standby.
Characteristics of the cluster:
1. Multiple virtual switches : a stacked switch appears as a logical switch externally, and the control plane is unified for unified management.
2. Unified forwarding plane : The forwarding plane of physical devices in the stack is integrated, and forwarding information is shared and synchronized in real time.3. Cross-device link aggregation : Links across physical devices in the stack are aggregated into an Eth-Trunk port to interconnect with downstream devices.
Note here:
Modular switches support cluster technology, common modular switches S9700 and S7700
Some fixed switches support stacking technology, common fixed switches that support stacking are S5700 and S6700
The names of stacking and clustering are different, but their technical principles are the same. They both logically virtualize multiple physical devices into one switch.
Some terms in the cluster:
The master switch
is responsible for managing the entire cluster. There is only one master switch in the cluster.The standby switch
is the backup switch of the master switch. When the master switch fails, the standby switch will take over all the services of the original master switch. There is only one standby switch in the cluster.
The cluster ID
is the CSS ID, which is used to identify and manage member switches. The cluster IDs of member switches in a cluster are unique.The priority of the cluster
is Priority, which is an attribute of a member switch. It is mainly used to determine the role of the member switch during the role election process. The larger the priority value, the higher the priority, and the higher the priority, the greater the possibility of being elected as the master switch. .Switch role election in the cluster: 1. Running status. 2. The higher the cluster priority, the better. 3. The smaller the MAC address, the better
3.2, CSS cluster establishment
The process of establishing a cluster is similar to the process of establishing a stack:
1. Establish a cluster switch to enable the cluster and configure the parameters required by the cluster.
2. The device is powered off
3. Connect cluster cables
4. Power on the device
5. The cluster system is automatically established.
The system automatically builds include:
1. Role election, based on running status, priority, MAC address
2. Version synchronization (the standby switch downloads the software from the main switch to restart and update),
3. Configuration synchronization, configuration file synchronization.
4. Configuration backup, when the switch enters the cluster from a non-cluster, some non-cluster files will be stored in the form of .bat, and these files will be restored when exiting the cluster.
How to connect to the cluster:
There are two connection methods for setting up clusters of devices, cluster card clusters and service port clusters.
Cluster card cluster mode: Cluster member switches are connected through dedicated cluster cards on the main control board and dedicated cluster cables.
Service port cluster mode: cluster member switches are connected through common service ports on service boards, and no dedicated cluster cards are required. Like iStack, the service port cluster involves two concepts of ports: physical member port and logical cluster port
3.3. Joining and merging of cluster members
Similar to stack member joining and merging.
Single-chassis CSS: After clustering is enabled on a modular switch, it is called single-chassis CSS.
Joining of cluster members:
When a modular switch is to be added to a running single-chassis CSS, the process is similar to stacking. After the power is off, the cables are connected and then started. After joining, it becomes the standby switch.
Merge of clusters:
Two single box CSS systems running, merged. After the election, the switch with higher priority becomes the master switch, and the switch with lower priority restarts. (Note that only the MPU of the main control board will be restarted, and the LPU of the interface board will work normally. The business and control are separated.)
3.4. Cluster split
After the cluster is established, the main control board of the system and the standby main control board of the system regularly send heartbeat messages to maintain the state of the cluster system. Failures of cluster cables, cluster cards, and main control boards, or power-off or restart of one of the switches will result in the loss of communication between the two switches. When the heartbeat message between the two switches times out (the timeout period is 8 seconds), the cluster system will split into two single-frame cluster systems, as shown in the figure:
After the cluster splits, since the member switches run the same configuration file, two cluster systems with the same IP and MAC will be generated. In order to prevent network failure caused by this, the conflict check of IP address and MAC address must be carried out.
3.5. MAD detection
The multi-master detection method of the cluster is the same as that of the stack.
3.6. Cluster configuration
1. Set up a cluster through the connection mode of the cluster card
[SwitchA] set css rijode css-card //Configure the connection mode of the cluster card[SwitchA] set css id 1 //Configure the cluster ID of the member switch
[SwitchA] set css priority 100 //Configure the cluster priority of the device
[SwitchA] css enable //Enable the cluster function of the switch
2. Build a cluster through the connection mode of the service port
[SwitchA] set css mode lpu //Configure the connection mode of the service port[SwitchA] set css id 1 //Configure the cluster ID of the member switch
[SwitchA] set css priority 100 //Configure the cluster priority of the device[SwitchA] interface css-port 1 //Enter the logical cluster port view
[SwitchA-css-port1] port interface xgigabitethernet 1/0/1 to xgigabitethernet 1/0/2 enable
//Configure the service port as a physical member port and set Add physical member ports to logical cluster ports
[SwitchA] interface css-port 2
[SwitchA-css-port2] port interface xgigabitethernet 2/0/1 to xgigabitethernet 2/0/2 enable[SwitchA] css enable //Enable the switch The cluster function of