I am looking for best practices or steps to enable Jumbo frames on running Cisco Hyperflex cluster with production VMs running on it.
From Cisco UCS & ESXi perspective I can enable jumbo frames. However, I am not sure if enabling jumbo frames on already deployed cluster has any impact on running workload ? Does it require downtime of entire cluster along with running virtual machines on the HX cluster ?
Solved! Go to Solution.
HyperFlex already configure the UCS-Manager (FIs) with QoS and the right MTU sizes.
If you want to change those sizes, you can do that on the UCS Manager without any impact.
It is important to have an end-to-end solution regarding the MTU size. So if the traffic is also going over the network, those switches should be having Jumbo frames as wel. This is a disruptive change.
Also make sure the DF (Don't Fragment) bit is not set on the switches.
Yes, Hyperflex will set the policies accordingly. However, could you please elaborate more about the steps that I need to follow. I assume I need to follow below, please correct if its wrong.
1) Enable the Jumbo Frames on Network. If its disruptive then how much downtime is required to do this change ? Will my Hyperflex cluster be down for few minutes due to this activity.
2) Enable Jumbo Frames in Cisco UCS. I believe, this is not disruptive with FI Model 6454. Does it have outage for Hyperflex ?
3) Enable Jumbo Frames in VMware. Does it cause outage for Hyperflex ?
1) When changing the MTU size on the switches will bring down this switch. Depending on your network, it can be disruptive.
Every MTU Change project I did was within a maintenance windows with possibility of downtime.
2) Changing the MTU size of any of the Fabric Interconnects (61xx, 62xx, 63xx, 64xx) doesn't have any impact of the flow of the packets.
3) Changing the MTU size of the vSwitch doesn't have any impact. I don't have any experience with vDS. You will have to verify on the VMware site.
Thanks for the suggestion. Changing on VMware also does not have much impact. However, can you enlighten with the sequence of steps that need to be followed. For example - which layer do I need to first perform this change first, like is it first on network -> UCS -> VMware or is it some other sequence ?
Also, do I need to bring Hyperflex cluster down for this change ? If yes then does it have any impact on the running VMs on it ?