No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>


To have a better experience, please upgrade your IE browser.


A problem about the N8300 high reliability redundant testing and validation

Publication Date:  2012-12-04 Views:  84 Downloads:  0
Issue Description
There are two nodes, and 8 physical service interfaces all together, bonding the 4 interfaces of each nodes as the bond0 respectively, and supports 2 VIP for the external, configure the DNS resolution in the DNS server, all the clients access the N8000 system via the DNS resolution, there has 2 VIP which is used to resolution. After completing the configuration, implement the high reliability testing, for the first time, there remains only one network cable in the node, and draw out the other, ping DNS domain name from the client; draw out/in the network cable in turn, and connect only one network cable each time, sometimes it can ping through, but sometimes can’t ping through, and there appears some problem. The network is as followed:

Alarm Information
Handling Process
Re-modify the configuration and modify the bonding (Network> bond remove bond0), then bonding the 3 network interfaces except the Pubeth0 interface (Network> bond create pubeth1,pubeth2,pubeth3  6), as the following picture displayed, there also provide 2 VIP for the out. The configuration of the DNS server hasn’t changed. Separate the Pubeth0 management interface and use to manage, not draw out the network cable, take the other 6 network cables for testing operation, the client can ping through the DNS domain name and haven’t discarded packet, the problem has been solved basically.

Root Cause
In the above two nodes, each node fixedly has a Pubeth0 network interface, this interface is the management interface of the cluster, meanwhile, the management IP of the console VIP is drifting between the two Pubeth0 network interfaces, either in the Pubeth0 of node1 or the Pubeth0 of the node2, some mutual information of the cluster must go through the two management interfaces, therefore, once the network cable of the management interface has been drawn out, it will influence the N8000 system’s services, and cause to can’t ping through, even cause the N8000 system collapse.

In the beginning of the test, it can’t ping through, we can’t confirm where happens the problem, a feedback from the research & development is not to draw in/out the network cable randomly, firstly, don’t begin to test until connect all the cables, they haven’t tell the root reason. From this case, we can understand the relationship between the management interface and the service interface, the design of the management interface is not reasonable, if the controller management interface of the S2600 doesn’t draw in the network cable, the device runs normally, the management interface will breakdown once the N8000 management interface hasn’t connected the network cable. We suggest advancing the N8000’s management interface, and make it low coupling with the N8000 system (we have submitted the OR).