[j-nsp] Netflow v9 sampling rate configuration in 10.4R3

prd::S prd.stee at gmail.com
Thu Apr 7 07:44:05 EDT 2011


Thanks for the reply, Chris.

On 4/6/11, Chris Tracy <ctracy at es.net> wrote:
> Our Juniper rep gave us a 4-slide PPT showing how cflowd v9 was configured
> before and after 9.6 -- you might want to ask if they can share this with
> you.  One thing that has been added >9.6 is the ability to do per-FPC
> sampling instances.  e.g. you can do forwarding-options { sampling {
> instance { test { input { ... } family [whatever] { output { ... } }, then
> apply this to an FPC using chassis { fpc X { sampling-instance test; } }
>
> Unfortunately, I don't see a way to do the per-family sampling rate like you
> noted, and the sampling instance stuff doesn't look like it would solve your
> problem _unless_ you can apply multiple sampling instances to an FPC.  I
> only briefly tested cflowdv9 w/ MS-DPC on the MX platform, and never with
> 10.x, but am interested to hear what you find out.
>
> Hope this helps!
>
> -Chris

Tried to apply multiple sampling-instances on an fpc, still no luck:

[edit chassis]
fpc 0 {
    /* IPv4 instance */
    sampling-instance sampling-inet;
    /* IPv6 instance */
    sampling-instance sampling-inet6;
}

Commit failed:

[edit chassis fpc 0]
  'sampling-instance'
    number of elements exceeds limit of 1

However if I take out instance sampling-inet and configure IPv4
sampling at global level, like:

[edit forwarding-options]
sampling {
    input { ... }
    family inet { ... }
    instance {
        si-inet6 {
            family inet6 { ... }
        }
    }
}

[edit chassis]
fpc 0 {
    /* IPv6 instance */
    sampling-instance sampling-inet6;
}

This configuration passes commit check, but the sampling does not work.
As I observed, there is no sampled packet sent to the sp interface.
Deactivating either global or non-global sampling-instance makes
sampling functional again (with only one family).

It seems that per-family sampling rate is no longer configurable on
JUNOS 10.4R3, or do I miss something?

Regards,
Nat.


More information about the juniper-nsp mailing list