Hi Inge,

Glad to read back from you. I think you are hitting a classic limit of
pcap filters: to match something a packet/flow within a VLAN the filter
should be 'vlan and ip' (to say: any IP packet/flow within any vlan).
pcap filters write a BPF program: i've not been able to find a way to
make a single filter to comprise both vlan-tagged and non vlan-tagged
traffic so the only solution that appears to work to me is:

id=220 ip=a.b.c.d filter='ip'
id=230 ip=a.b.d.e filter='ip'
id=220 ip=a.b.c.d filter='vlan and ip'
id=230 ip=a.b.d.e filter='vlan and ip'

Cheers,
Paolo

On Thu, Apr 16, 2015 at 11:07:56AM +0000, Inge Bjørnvall Arnesen wrote:
> Hi all,
> 
> I've been using this software for a number of years, but this year is the 
> first time I've tried to set it (here 1.5.0 and later 1.5.1) up with Ipfix 
> (V10) reporting (Juniper) and I've come across a strange problem. When 
> looking at flows with:
> 
> aggregate[inmem]: dst_host, in_iface,out_iface,vlan
> 
> and no aggregate_filter or filtering in pretag.map, I see all flows coming in 
> from all reporting interfaces and with their various VLAN tags. As soon as I 
> create a filter (any positive filter that require IP- either in pretag.map or 
> using aggregate_filter), all flows where the VLAN ID field is non-zero is 
> filtered away from the flows reported by nfacctd. E.g. with a pretag.map with 
> (variables used here to hide real IPs):
> 
> 
> -----------------------------
> id=220 ip=a.b.c.d
> id=230 ip=a.b.d.e
> -----------------------------
> 
> and nfacctd.conf containing:
> 
> -----------------------------
> pre_tag_filter[outmem]: 220,230
> aggregate[outmem]: src_host, in_iface,out_iface,vlan
> -----------------------------
> 
> 
> I see all traffic for all flows, but if I change pretag.map to:
> 
> -----------------------------
> id=220 ip=a.b.c.d filter='ip'
> id=230 ip=a.b.d.e filter='ip'
> -----------------------------
> 
> Only flows where VLAN is 0 is shown:
> 
> [root@bf-netflow01 sbin]#  pmacct -s -p /tmp/pmacct_out.pipe  | head -n 6
> IN_IFACE    OUT_IFACE   VLAN   SRC_IP           PACKETS               BYTES
> 556         577         0      141.136.112.53   500                   20000
> 556         577         0      85.114.2.251     500                   750000
> 556         577         0      95.153.164.152   500                   26000
> 556         577         0      84.18.126.21     500                   254000
> 556         577         0      79.110.18.161    500                   41500
> .....
> 
> All others appear to be somehow lost in filtering. Of course, using 'ip' as a 
> filter is just to illustrate the problem - I usually filter on src net (which 
> also requires positive IP filtering).
> 
> If I change the pretag filter to:
> -----------------------------
> id=220 ip=a.b.c.d filter='not ip'
> id=230 ip=a.b.d.e filter='not ip'
> -----------------------------
> 
> I only see flows where the VLAN number is non-zero:
> 
> [root@bf-netflow01 sbin]#  pmacct -s -p /tmp/pmacct_out.pipe  | head -n 6
> IN_IFACE    OUT_IFACE   VLAN   SRC_IP           PACKETS               BYTES
> 554         511         3334   194.76.246.132   500                   730000
> 554         511         3334   109.68.89.199    500                   26000
> 511         554         3335   109.74.179.138   500                   750000
> 511         556         3335   194.76.246.18    18000                 12336000
> 511         556         3335   79.171.81.64     39500                 2618500
> .....
> 
> The src_host field strangely contains the correct source IP even though I 
> filter for 'not ip'. Note that if I substitute the filter 'ip' with 'src net 
> 128.0.0.0/1', the same situation occurs - only flows with VLAN ID zero are 
> shown, but if I substitute 'not ip'  with e.g. 'not src net 128.0.0.0/1', 
> flows are shown regardless of VLAN ID.
> 
> Any idea on what might be the issue here or how I might patch around this 
> issue in the source?
> 
> Regards,
> 
> 
> n  Inge
> 
> 

> _______________________________________________
> pmacct-discussion mailing list
> http://www.pmacct.net/#mailinglists


_______________________________________________
pmacct-discussion mailing list
http://www.pmacct.net/#mailinglists

Reply via email to