Configuring sFlow Agents on the Network

---

You must first configure the sFlow Collector before configuring the sFlow Agent using the sflow-create command. You must configure and enable sFlow Agent on each switch that you want to be part of the sFlow monitoring system to monitor the network traffic.

To configure and enable an sFlow Agent on a switch, use the sflow-create command:

CLI (network-admin@switch) >  sflow-create name name-string

Note: While creating the sFlow, do not configure both the sample-rate and counter-polling-interval to zero (0). If you do so, an error message: sflow-create: Sample-rate/counter-polling-interval both can't be 0 is displayed. However, later, you can change both values to zero (0) using the sflow-modify command if desired.

Note: When you configure the sample-rate and the counter-polling-interval to zero (0) using the sflow-modify command, then no packets are sent out for sFlow monitoring. 

Below is an example configuration on a switch to enable the sFlow Agent, net-monitor, on the ingress ports 57-59, with sample-type raw, sample-rate 4096, counter-polling-interval of 5 seconds (i.e., 1 sample in every 5 seconds), and trunc-length of 160 bytes:

CLI (network-admin@switch) >  sflow-create name net-monitor type ingress sample-type raw ports 57-59 sample-rate 4096 counter-polling-interval 5 trunc-length 160

For deleting an existing sFlow Agent, use the command:

CLI (network-admin@switch) >  sflow-delete name name-string

To view the sFlow Agent details, use the command:

CLI (network-admin@switch) >  sflow-show

Note: To view the binding between the sFlows and the nv-message-queues, use the sflow-thread-binding-show command. This command helps during debugging process and should be run with the support from Pluribus TAC team. See the Example: A Sample Configuration below.

To modify the configuration details for an sFlow Agent, use the command:

CLI (network-admin@switch) >  sflow-modify name name-string

For example, to modify and view the counter-polling-interval of the sFlow, net-monitor, use the command:

CLI (network-admin@switch*) >  sflow-modify name net-monitor counter-polling-interval 120

CLI (network-admin@switch*) > sflow-show format all layout vertical

name:                                net-monitor

type:                                ingress

sample-type:                        raw        

ports:                                1,3,53,57-59

sample-rate:                        4096

counter-polling-interval:120

sample-interval:                5

trunc-length:                        160 

agent id:                                10.1.1.200

sample-pkt-cnt:                2845

sample-drops:                        87

Based on the the configuration following details are displayed for the sflow-show command:

CLI (network-admin@switch*) > sflow-show

If all the command parameters are configured for an sFlow Agent, then the sflow-show command displays an output as given here:

CLI (network-admin@switch*) > sflow-show format all layout vertical

switch:                                         switch

name:                                        sf25

type:                                        ingress

sample-type:                                raw

ports:                                        25

sample-rate:                                4096

trunc-length:                                160

agent-id:                                 10.14.22.100

sample-pkt-cnt:                        66798

sample-drops:                                 0

malformed-packet-drops:                 0

malformed-vlan-packet-drops:         0

malformed-ip-packet-drops:         0 

internal-q-drops:                         0

oversize-packet-drops:                 0 

internal-nq-drops:                         0

Example: A Sample Configuration

Below is a set of sample configuration details for the sflows created in Pluribus lab for further understanding.

CLI (network-admin@switch*) > sflow-show

The sflow-show command displays 15 sFlows Instances (s1-s15) configured using the sflow-create command. In NetVisor OS, when you configure the sFlows, every six sFlows are

bound to one thread (sFlow Agent). To view the details, use the command:

CLI (network-admin@switch*) > sflow-thread-binding-show

name thread-id       nv-msg-queue-name nv-cache-queue-name

---- --------------- ----------------- -------------------

s1   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s2   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s3   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s4   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s5   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s6   140323777259264 sflow-packet-q-9  sflow-cache-q-9

s7   140322640197376 sflow-packet-q-10 sflow-cache-q-10

s8   140322640197376 sflow-packet-q-10 sflow-cache-q-10

s9   140322640197376 sflow-packet-q-10 sflow-cache-q-10

s10  140322640197376 sflow-packet-q-10 sflow-cache-q-10

s11  140322640197376 sflow-packet-q-10 sflow-cache-q-10

s12  140322640197376 sflow-packet-q-10 sflow-cache-q-10

s13  140322665375488 sflow-packet-q-11 sflow-cache-q-11

s14  140322665375488 sflow-packet-q-11 sflow-cache-q-11

s15  140322665375488 sflow-packet-q-11 sflow-cache-q-11

From the above output, you can see that the first six sFlows Instances (s1-s6) have the same thread-id,nv-msg-queue-name, and  nv-cache-queue-name parameters. Similarly, the next six sFlows Instances (s7-s12) have a new set of parameters and so on. This helps in understanding the bindings between the sFlows and the parameters.

In the above output:

• thread-id: is the unique identifier of the CPU resources associated to sFlow Instances (in groups of 6)
• nv-msg-queue-name: is the actual software queue used for sFlow data
• nv-cache-queue-name: is a pool of buffers. NetVisor OS allocates one memory block from a cache_queue whenever a sample is received from the hardware, which is then queued to the nv_queue. 

To view the nv_queue details, use the command:

CLI (network-admin@switch*) > nv-queue-stats-show | grep sflow