F5 is upgrading its customer support chat feature on My.F5.com. Chat support will be unavailable from 6am-10am PST on 1/20/26. Refer to K000159584 for details.

Forum Discussion

Josh_41258's avatar
Josh_41258
Icon for Nimbostratus rankNimbostratus
Sep 17, 2013

Installation of new blade in VIPRION/vCMP System

Let's say I have a VIPRION chassis running vCMP with 2 blades installed. The vCMP guests on this host are set to "Use all slots."   As I understand it, if I install a new blade into the chassis, ...
deployment
design
virtualization
Josh_41258's avatar
Josh_41258
Icon for Nimbostratus rankNimbostratus
Sep 18, 2013

I plan on having the switch interfaces configured and ready to be plugged in. They just won't be plugged in IMMEDIATELY as the blade is inserted.. perhaps 30-60seconds later. I just want to make sure there the blade won't try to start processing traffic before I can get the interfaces cabled.

 

@Josh - thanks for the detailed response! Ours are cabled directly to Nexus 7k's. I typically have 4x10GB for each blade (2x10 for the internal side, 2x10 for the external). Each blade also has a switchport on the "sync" or "ha" VLAN that is used for sync and network failover traffic. I don't typically interconnect individual blades together.

 

Thanks,

 

Josh

 

  • JoshBecigneul's avatar
    JoshBecigneul
    Icon for MVP rankMVP
    Sep 18, 2013
    That sounds solid. Are you splitting trunks on different line cards with VPC? Blades can connect to other blades through the chassis. To my knowledge you do not need to cable every blade to the network in order to gain their compute power, but you end up with SPOFs in certain failure scenarios. Also, you can setup trunks across slots, so you can have trunk INTERNAL running on 1/1.1 and 2/1.1 and trunk EXTERNAL on 1/1.2 and 2/1.2, and so on. This mode works great for ours, allowing us to "yield" blades for downtime. Be very careful to always yield the blade during downtime as you can otherwise end up with a traffic interruption. I found that when removing a slot 2 blade without any guests, it was still handling traffic for guests on the other blade, likely due to how LACP had organized my links.
  • JoshBecigneul's avatar
    JoshBecigneul
    Icon for MVP rankMVP
    Sep 18, 2013
    Also, it will likely be 15-30 minutes before the new blade is patched, rebooted, firmware updated and joined to the cluster before it is ready to process traffic. Once its fully operational, then you can configure its network interfaces. Up until the network config, this happens more or less automatically.
  • Josh_41258's avatar
    Josh_41258
    Icon for Nimbostratus rankNimbostratus
    Sep 18, 2013
    Yes, we do use VPC. So, if we are talking about a single blade: 1.1 -> internal-trunk -> NEXUS-7K-01 -> VPC 100 / 1.2 -> internal-trunk -> NEXUS-7K-02 -> VPC 100 / 1.3 -> external-trunk -> NEXUS-7K-01 -> VPC 200 / 1.4 -> external-trunk -> NEXUS-7K-02 -> VPC 200 / 1.8 -> ha-trunk -> private/non-routed VLAN I didn't know blades could connect to other blades through the chassis. I thought every blade had to be cabled individually. I will continue to do it this way anyways, as it provides higher throughput and more redundancy, I suppose. Um, yeah! I have also noticed that when I remove a blade with no active guests, I get traffic interruption. (sorry, this forum is mangling my line breaks)