Chapter 4-23. Troubleshooting Congestion in Fibre Channel Fabrics - 指南

Anomaly — End Device Reflecting Congestion in the Fabric in Opposite Direction

The second anomaly is even more interesting. In the graphs shown inFigure 4-18 and Figure 4-19, for both Fabrics, there seems to be a time correlation between the Tx-datarate > 80% (high utilization) events on host port (fc2/45) and TxWait < 30% (Tx-Lvl-1 on the Y-Axis) on both target ports (such as fc3/7 and fc4/7 in Fabric A). This is interesting because the Tx-datarate > 80% (high utilization) events indicate traffic going towards the host on fc2/45 and the TxWait < 30% entries to the two targets on fc3/7 and fc4/7 indicate congestion in traffic going toward the two arrays which is the opposite direction of the Tx-datarate entries. So, somehow, the large number of Read I/O operations, which is generating a large amount of traffic towards the host is causing some minor congestion even in the flow of command frames (which initiate I/O operations) going in the opposite direction. These details are covered inChapter 5, the section onI/O Operations and Network Traffic Patterns.第二个异常更为有趣。在图 4-18 和图 4-19 所示的图中，对于两个Fabric，主机端口(fc2/45)上的Tx-datarate > 80%（高利用率）事件与两个目标端口（如Fabric A中的 fc3/7 和 fc4/7）上的TxWait < 30%（Y 轴上的Tx-Lvl-1）之间似乎存在时间相关性。这很有趣，因为Tx-datarate > 80%（高利用率）事件表明流量流向fc2/45上的主机，而 fc3/7 和 fc4/7 上两个目标的TxWait < 30%条目表明流向两个阵列的流量拥塞，这与Tx-datarate条目方向相反。因此，大量读取 I/O 操作会产生大量流向主机的流量，甚至会导致相反方向的命令帧流（启动 I/O 操作）出现轻微拥塞。这些细节将在第 5 章 "I/O 操作和网络流量模式 "一节中介绍。

Figure 4-21is the same asFigure 4-18, except that the focus is different. Notice the two circled times of high utilization on fc2/45 and the corresponding TxWait entries on fc3/7 and fc4/7 (inside circles).图 4-21 与图 4-18 相同，只是焦点不同。请注意fc2/45上两个圈内的高利用率时间，以及 fc3/7 和 fc4/7