03-03-2017 08:32 AM - edited 03-03-2017 08:38 AM
Got this interesting issue in only one of my fabrics, where port 33 is higlighted on GUI as bottlnect,
Warning Severe latency bottleneck detected at E-Port 33. AN-1010
error wise, showing pretty much identical number of disc c3 and c3timeout tx (all other counters being absolutelly clean) errors when using porterrshow:
33: 2.4g 1.2g 0 0 0 0 0 0 0 1.5k 0 0 0 0 0 1.5k 0 0
1: 33-> 13 10:00:00:05:33:c6:ef:18 19 deskew 15 MASTER 32-> 12 10:00:00:05:33:c6:ef:18 19 deskew 16 Tx: Bandwidth 32.00Gbps, Throughput 1.41Gbps (5.15%) Rx: Bandwidth 32.00Gbps, Throughput 995.14Mbps (3.62%) Tx+Rx: Bandwidth 64.00Gbps, Throughput 2.41Gbps (4.38%)
Am wondering firs and foremost how such robust ISL link can become bottleneck (I blame here my lack of knowledge as am pretty sure there is a good explanation to it)?
Am suspecting a dodgy fibre but this may seem as red herring (still not best in narrowing down paticular combinations of errors).
Pattern wise, this seems to be happening fairly ocassionally, not constantly, happening for example most of today, a little bit yesterday, and then like three weeks ago ....
Is it some downstream switch or host causing some issue?
Any hints or pointers as to how to toroubleshoot this would be greately appreciated as this sprung to live fairly unexpectedly (I assume bottleneckmon would be the first port of call - had no need to use it till now - unless above actually manifest some obvious issue I am not able to pick up on).