Network Interface (NIC) CRC Errors and Cut-Through Switching

  • 18 September 2020
  • 0 replies
  • 88 views

Badge +1

 

Is it really possible that a networking issue, which exists on the other side of a large/vast network, could manifest locally on a host as NIC CRC errors (rx_crc_errors)?

Yes!

The way that frames are moved across a network with cut-through switching (which is the model used by current/high-performant data center switches) differs from the traditional store-and-forward model.

With store-and-forward switching, frames are entirely received (and error-checked) on a switch before being passed along to the next switch inline. If errors are found within a frame, the frame is not passed along to the next switch.

However, with cut-through switching, a switch is simultaneously receiving a frame and already passing it along to the next switch inline! Error checking is then only completed on a frame after it has already left a switch. Accordingly, any errors within the frame simply get passed along to the next switch inline until the frame reaches its final destination.

In this way, when troubleshooting CRC errors on a NIC, it is important to not only check the locally attached media to the NIC but also for any evidence of issues further upstream. More information regarding this can be found within the “NCC Health Check: host_nic_error_check” Knowledge Base article.


0 replies

Be the first to reply!

Reply