Chapter 4-12. Troubleshooting Congestion in Fibre Channel Fabrics

Error-statistics

The show logging onboard module error-stats command displays specific error-statistics, like TxWait, that are recorded every 20 seconds. Each module checks each of its interfaces’ error counters every 20 seconds to see if any of the applicable error counters have incremented (even by 1). It’s important to note that when an error counter increments the current value of the error counter is recorded not just the delta value (the amount it incremented in that 20 seconds). Consequently, to determine the delta value for a specific interface and error counter entry, an earlier (in time) entry must be located, and the two counts must be subtracted. If no earlier entry for the same interface and error counter is found, then it could be because this is the first entry for that specific interface and error counter combination, or it could mean that the prior entry was overwritten due to the OBFL cyclical file wrapping. show logging onboard module error-stats 命令显示每 20 秒记录一次的特定错误统计信息,如 TxWait。每个模块每 20 秒检查一次其每个接口的错误计数器,查看是否有适用的错误计数器递增(即使递增 1)。值得注意的是,当错误计数器递增时,记录的是错误计数器的当前值,而不仅仅是 delta 值(20 秒内的递增量)。因此,要确定特定接口和错误计数器条目的 delta 值,必须找到较早的(在时间上)条目,并将两个计数相减。如果没有找到同一接口和错误计数器的早期条目,可能是因为这是特定接口和错误计数器组合的第一个条目,也可能是由于 OBFL 周期性文件包装而覆盖了之前的条目。

Most MDS switches record their counter names with some hardware-specific information. Consequently, the same error or indication may look different on different types of switches or modules. 大多数 MDS 交换机在记录计数器名称时都会包含一些特定于硬件的信息。因此,相同的错误或指示在不同类型的交换机或模块上可能会有所不同。

The following are meaningful error-stats counters for troubleshooting congestion. 以下是对拥塞故障诊断有意义的错误统计计数器。

Tx-credit-not-available (100ms continuous)

Refer to Chapter 3, the section on Continuous Tx Credit Unavailability for 100 ms — Tx-credit-not-available for more details on this counter. The following are the counters that detect the continuous duration of zero remaining-Tx-B2B-credits on MDS switches. 有关该计数器的更多详情,请参阅第 3 "100 毫秒内连续 Tx 信元不可用 - Tx-credit-not-available "一节。以下是在 MDS 交换机上检测剩余 Tx-B2B 信元为零的持续时间的计数器。

 FCP_SW_CNTR_TX_WT_AVG_B2B_ZERO: MDS 9700 16G module, MDS 9148S, 9250i, 9396S

你可能感兴趣的:(Linux,kernel,网络,运维,服务器,kernel,linux,infiniband)