Transient Buffer Congestion Detection

Transient buffer congestion can cause the dropping of packets and its cause is difficult to determine without a way to detect and log its occurrences. The SLX-OS device allows you to configure the monitoring of discarded packets for all VOQs that map to physical ports or multicast VOQs. You can also monitor the discarded and deleted packets on the device.

Note

• VOQs that are used by CPU traffic are not monitored.
• Transient buffer congestion detection is supported only on devices based on the DNX chipset family. For more information, see Supported Hardware.

When traffic flows from the ingress packet processor to the ingress TM, it is buffered in the destination VOQ which marks the destination port of the packet. If this VOQ does not receive enough credits from the egress TM, the VOQ buffer overflows and the packets are discarded or deleted (aged out) at the ingress TM.

By default, monitoring of these packets is not enabled. To enable logging, you must configure a threshold limit for the discarded or deleted packet count. The discarded packet count of a VOQ is logged based on the configured threshold value. You can also set the time interval that a RASlog message is generated if the discard count threshold limit is reached.

Note

To restrict logging other than the logging interval, the number of logs recorded within the logging interval is also set to a maximum of 50 logs for each line card. Within the expiry of logging time interval, only 50 logs are generated for 50 VOQs reporting the maximum discarded packets beyond the threshold limit.

When the discarded packet count exceeds the threshold value, an INFO severity RASlog is generated with information of the interface, ingress tower, ingress core and configured threshold limit, similar to the following example.

device# show logging raslog reverse count 100 | inc SYSD-1003
2017/01/05-11:03:59, [SYSD-1003], 793, M2 | Active | DCE, WARNING, SLX,  TM threshold 
2017/01/05-11:00:34, Discarded packets 930587727, interface 3/2 prio 0 on device 3.1.0

The RASlog is only generated if the logging-time interval has expired. If the discarded packet count threshold is constantly exceeded, the RASlog is only generated after the configured logging-time interval expires.

The time interval to monitor VOQ discarded packet counts is four minutes. Within every four minutes, the discarded packet count for all VOQs of all towers are monitored. If any VOQ discarded packet count has exceeded the threshold, a breach instance is recorded and a RASlog is generated. If the current recorded breach instance is before the logging time expiry, the RASLOG is not generated.

Although the configured RASlog interval is between 10-2880 minutes and the VOQ statistic monitoring time is four minutes, the logging time set is adjusted to the nearest multiple of 4. For instance, if you configure the monitoring time to 10 minutes, the log for the VOQ is generated every 12 minutes.

The SLX-OS device also allows the monitoring of the total discarded packet count and deleted packet count on the device. These are device counters and are monitored every five seconds. You can configure different threshold limits for the discarded and deleted packet counts. If a RASlog message for the TM device statistics is generated within the logging interval, the message includes the time stamp, the number of discarded packets, and ingress slot, tower and core, similar to the following example.

device# show logging raslog reverse count 10
2017/01/05-10:56:58, [SYSD-1005], 788, M2 | Active | DCE, WARNING, SLX,  TM threshold, 
Tail discarded packets 20734462 on device 3.1.1.