Run a RoCE fabric in production: know where every counter lives, work a symptom to a cause in a fixed order, configure the switch half on Cumulus and Dell SONiC, and plumb RDMA into Kubernetes.

1. 4.1 The counter map: sysfs, ethtool and rdma statisticThe dashboard that swears the rail is deadAp
2. 4.2 Triage in order: symptom to counter to cause01:40, and the bridge wants to change a DCQCN parameterE
3. 4.3 MTU, RoCE LAG, and what has to be installedThe bond that took the RDMA device with itAp
4. 4.4 The switch half: Cumulus NVUE and Dell Enterprise SONiCDoes this switch do RoCE, or only PFC?Ap
5. 4.5 RDMA in Kubernetes: the NVIDIA Network OperatorNobody proved the fabric, and acceptance is at nineAn
6. Checkpoint