[j-nsp] Cosmetic bug? - mx80 12.2r7.7

Fabien DEDENON dafused.nog at gmail.com
Thu Sep 4 09:10:23 EDT 2014


Le 04/09/2014 14:52, Gavin Henry a écrit :
>> That looks to be it. I had regular flow monitoring yesterday. Changed to inline this morning and only noticed CPU then. Rolled back to regular, but CPU still at 100% - Looks like I need to restart FPC...
> We get this too when enabledin 1:1 IPFIX sampling with sampled using
> all the CPU. If 1000:10 then it's OK. Have you seen this?
>
Our conf was at 1:1000 ( rate 1000 ) sample ratio, and we tripped the
bug, on MX80 as on MX480.
For now we're back to RE based sampling. So many issues and feature
lacks on inline jflow Mx|Trio implementation that we still never really
used it.

For the lacks, i'm thinking to IPFIX attributes 304 thru 311, for
sampling ratio indication, source and des MacAddress attributes and
probably many others.

Fabien



More information about the juniper-nsp mailing list