Hi everyone, I'm reviewing the statistics of my trunk on the CMTS C40g and would like to know if this data is a cause for concern. Thanks!
CASA-C40G(config)#show interface trunk 1 stat
Interface Trunk 1 statistics
IfIndex 9000001
interface status UP
link status UP
mac address 0017.101f.b41d
link up member xgige 2/1 3/0
IfInOctets 204159293235994
IfInUcastPkts 162623202181
IfInNUcastPkts 375413
IfInDiscards 167644679 <-------------------------------- THIS
IfInErrors 0
IfInOversizePkts 0
IfInUnknownProtos 40729
IfOutOctets 18816505656799
IfOutUcastPkts 48269041894
IfOutNUcastPkts 14920
IfOutErrors 0
IfOutDiscards 0
IpForwDatagrams 0
IpInReceives 162623031908
IpInDiscards 1
IfSpeed 20000000000
Last clearing of interface stat: 5 d, 1 h, 16 m, 38 s
unfortunately that is not enough information to answer your question. These are counters, counters are only useful in relation to time. you need two counters start and end , and the time between the two collection points, this will give you a rate.
You only have a single counter IfInDiscards 167644679 which the only thing you can glean from this is, that in the last 5d there have been 167,644,679 discards, during the same period there were IfInOctets 204159293235994 octets (bytes) recieved and IfInUcastPkts 162623202181 packets received. so 162,623,202,181 / 167,644,679 = 970.05 meaning on average for every 970 packets received there were 1 packets discarded. or in decimal a 0.0097 discard rate.
But what you can't tell is if the discards happened regularly, or if they only happened at peak time, or if there was a problem initially but it has since disappeared
This is why you need two counter values and the time difference between the two to calculate a rate, and the rate is what you want to graph.
you can run
show clock
show interface trunk 1 stat
wait about 30 secs then
show clock
show interface trunk 1 stat
That will tell you the rate that the ifDiscards are happening.
Another important note, counters reset when they hit their max value, they are only guaranteed to not have reset if you collect the two counter values within 1 hour of each other.
I created this graph in Grafana to show the rate of IfInDiscards for the Trunk1 interface.
The raw value from the device is a cumulative counter, so I used InfluxDB’s non_negative_derivative() function to calculate the rate of change over time.
Applied non_negative_derivative(5m) to get the rate of change per 5 minutes, avoiding negative spikes when the counter resets.
This is part of a deeper investigation, because I’ve been noticing some latency in the metrics during prime time (ping latency — which I understand doesn’t have priority). When checking with the cpuinfo command, I see peaks up to 91% on the upstream line cards, but it’s confusing because when I run cpu-history during prime time, it never exceeds 5%. I’m attaching information related to this as well.
What that means is tasks your upstream does, map requests, registration, etc... are making the processor to be fully loaded. The smm processor handles routing and other system level tasks.
It is most likely that you are out of minislots can you run show controller upstream when you are experience issues? vccap#show controller upstream
Interface upstream 1:0/0.0/0 information:
IfIndex : 50001280
Admin status: UP
Operating status: UP
Frequency: 17000000 Hz
Channel Width: 6400000 Hz
Power Level: 14
Modulation profile 3
SNR: 42.1
Statistics:
Received 665 broadcasts, 4150 multicasts, 18635111 unicasts
73 discards, 0 errors, 0 unknown protocol
12907120 Unerroreds, 14 Correcteds, 0 Uncorrectables
Req Mslots 2481039606, Used Req Mslots 6326367
Init Mtn Mslots 577738646, Used Init Mtn Mslots 0
Total Mslots 3185829677, Ucast Granted Mslots 127051425
There are 3 different kinds of minslots, Requests (bandwidth requests), Initial Maintenance (modems coming online) and Everything else. you need to compare Used versus total for all three and make sure that there is not any that are getting full. You can adjust the ratio if Request or initial maintenance are running full, but if the other is full then you are just out of bandwidth and would signify a capacity issue. You can run out of minislots before you reach bandwidth capacity on a channel.
CASA-C40G(config)#show interface upstream 5/3.0 stat
Interface upstream 5/3.0 statistics
Admin status: UP
Channel utilization interval: 30
Avg upstream channel utilization: 48
Avg percent contention slots: 51
Avg percent initial ranging slots: 1
Avg percent minislots lost on late MAPs: 0
Logical-channel 0
Received 116360 broadcasts, 289305 multicasts, 744346934 unicasts
24182 discards, 2080649 errors, 5046 unknown protocol
1550195650 Unerroreds, 90362 Correcteds, 4467 Uncorrectables
Total Modems On This Upstream Channel: 71 ,70 active cm,144 secondary cm
Req Mslots 2966573346, Used Req Mslots 318764274
Init Mtn Mslots 209239356, Used Init Mtn Mslots 0
Total Mslots 2393314686, Ucast Granted Mslots 3441414862
Admin status: UP
Last clearing of interface stat: 8 d, 5 h, 23 m, 25 s