Packet or frame loss - loss of bits of data causes an adverse effect on performance of applications. ![]() Typical performance problems in Layer 2 and IP networks would stem from: In complex environments with multiple responsible parties as typically seen in a FlexPod, you need to consider multiple aspects in order to troubleshoot the issue. This document discusses the most common FlexPod. Another solution, called the FlexPod Express, uses a UCS C-series chassis connected to Nexus 3000 switches. ![]() The most common FlexPod consists of a Cisco UCS B-series chassis connected via Fabric Interconnects (FIs) to Nexus 5500 switches to NetApp filers. This document was written as a result of issues seen by the Data Center Solutions Technical Assistance Center (TAC) team in recent months.Ī FlexPod consists of a Unified Computing System (UCS) computer connected via a Nexus switch to NetApp storage and IP networks. It is intended as starting point for customers who look to troubleshoot performance in a FlexPod environment. That will cause a network restart.This document describes common performance problems in FlexPod environments, provides a method to troubleshoot issues, and provides mitigation steps. Also note, once it comes back up, you will need to update the OS. I just tested this and click on "Evaluate" and it comes back saying it does cause a reboot. ![]() I've been told that it will just be a small outage while the Nic resets and then I've been told that the blades will reboot. What would the affect be of changing the MTU in the vNIC config? I have been told two different things on this as well. RAC DBs are the perfect example, the RAC application will not start unless there is a second NIC configured at 9000 for the heartbeat connection.Ģ. If everything requires 1500, then traffic should be able to go through those nics with no issues. If they require Jumbo Frames and it has not be set on the FI, then you will have network issues. This depends on the application and server requirements. I've been told yes it can and then also no it will not as the setting is just allowing the vNIC to use an MTU of 9000 as it's maximum if needed. If my vNIC's are set to 9000 can this cause an issue? I've been told two different things. The counters will be in hex, but if you have a decent number of MTU Violations, then your FIs are dropping frames that are likely larger than your QOS class "best effort" class allows by default (1500 unless you've changed it)ġ. (that will work on 6200 series FIs, if they are 6300, the command is different) Nxos# show hard internal carmel counters interrupt match mtu* ![]() Try running the following on each of your FIs: Personally I would set it back to 1500, unless you know jumbo frames are allowed end to end.Ģnd question, I got a prompt to reboot the HX rack servers I tried to test that on, so I'm guessing you have to reboot to apply the change.Īlso MTU violations will show up CRCs in your FIs/IOMs. VMware is a bit different due to the VMK port, and vswitch/port-groups defaulting to 1500. I've seen windows boot from iSCSI san cases where the installer started and due to windows baremetal OS reading the 9000mtu setting from the VNIC, it tried to communicate with iSCSI target LUN which of course lost communication due to frames being fragmented and dropped. For your first question about vnic being set to 9000 when overall network doesn't support it, it depends.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |