Another workaround, what I found - make remote NetFlow sources routable. I
dont know why, but disabling RP_filter has no effect on NFCAPD.
Thx Adrian for this.

2015-08-21 19:05 GMT+03:00 Roman Mavrichev <roman.mavric...@gmail.com>:

> NFCAPD is listening for all configured sources, but don`t write any data.
> I make reconfiguration about 2 hours ago. :
>
> root@msk-nms-1:/home/rmavrichev# nfdump -r
> /srv/nfsen/profiles-data/live/SPB-c3945-PE1/2015/08/21/nfcapd.201508211850
> Date first seen          Duration Proto      Src IP Addr:Port          Dst
> IP Addr:Port   Packets    Bytes Flows
> No matched flows
> root@msk-nms-1:/home/rmavrichev#
> root@msk-nms-1:/home/rmavrichev#
> root@msk-nms-1:/home/rmavrichev# netstat -ap | grep nfcapd
> udp        0      0 *:9996                  *:*
>       1310/nfcapd
> udp        0      0 *:9997                  *:*
>       1353/nfcapd
> udp        0      0 *:9998                  *:*
>       1261/nfcapd
> unix  2      [ ]         DGRAM                    10857    1310/nfcapd
>
> unix  2      [ ]         DGRAM                    10893    1353/nfcapd
>
> unix  2      [ ]         DGRAM                    10095    1261/nfcapd
>
> root@msk-nms-1:/home/rmavrichev#
>
> 2015-08-21 16:55 GMT+03:00 Adrian Popa <adrian.popa...@gmail.com>:
>
>> Then check that NFCAPD processes are listening on ports 9996 and 9997,
>> and also, if you've recently reconfigured your netflow sources you need to
>> wait a bit for them to send a template packet to describe the data format
>> they are sending. I found this can take up to 30 minutes sometimes...
>>
>> On Fri, Aug 21, 2015 at 3:47 PM, Roman Mavrichev <
>> roman.mavric...@gmail.com> wrote:
>>
>>> I tryed to disable rp_filter on interface, that recives flow`s, and
>>> again use non-local ip as source, but without success.
>>>
>>> I can see recieved traffic:
>>> root@msk-nms-1:/home/rmavrichev# tcpdump -i eth0.152  udp port 9996 or
>>> 9997 or 9998
>>> tcpdump: verbose output suppressed, use -v or -vv for full protocol
>>> decode
>>> listening on eth0.152, link-type EN10MB (Ethernet), capture size 65535
>>> bytes
>>> 15:39:19.065175 IP 10.78.19.1.57893 > 10.77.27.12.9996: UDP, length 1416
>>> 15:39:20.077726 IP 10.77.19.1.61026 > 10.77.27.12.9997: UDP, length 1416
>>> 15:39:22.077653 IP 10.77.19.1.61026 > 10.77.27.12.9997: UDP, length 1416
>>> 15:39:22.657511 IP 10.77.19.3.40000 > 10.77.27.12.9998: UDP, length 1316
>>> 15:39:23.077736 IP 10.77.19.1.61026 > 10.77.27.12.9997: UDP, length 1416
>>>
>>> rp_filter for eth0/152 is disabled:
>>> root@msk-nms-1:/home/rmavrichev# sysctl -a | grep 152.rp_filter
>>> net.ipv4.conf.eth0/152.rp_filter = 0
>>>
>>> But nfcapd write emty files after reconfiguration:
>>> root@msk-nms-1:/home/rmavrichev# nfdump -r
>>> /srv/nfsen/profiles-data/live/MSK-c3945-PE1/2015/08/21/nfcapd.201508211525
>>> Date first seen          Duration Proto      Src IP Addr:Port
>>>  Dst IP Addr:Port   Packets    Bytes Flows
>>> No matched flows
>>>
>>> Before reconfiguration, all was ok:
>>> root@msk-nms-1:/home/rmavrichev# nfdump -r
>>> /srv/nfsen/profiles-data/live/MSK-c3945-PE1/2015/08/21/nfcapd.201508211500
>>> | tail -n 10
>>> 2015-08-21 15:04:41.752     0.000 TCP     109.232.108.94:65083 ->
>>> 5.45.249.59:443          1       40     1
>>> 2015-08-21 15:04:41.920     0.000 TCP     109.232.108.94:65086 ->
>>> 87.250.247.193:443          1       40     1
>>> 2015-08-21 15:04:41.944     0.000 TCP     109.232.108.94:65085 ->
>>> 87.250.247.193:443          1       40     1
>>> 2015-08-21 15:04:50.476     5.008 TCP       185.3.142.64:80    ->
>>> 109.232.108.94:51590       17    19040     1
>>> 2015-08-21 15:04:50.476     5.036 TCP       185.3.142.64:80    ->
>>> 109.232.108.94:51591       35    45774     1
>>> 2015-08-21 15:04:50.476     5.424 TCP       185.3.142.64:80    ->
>>> 109.232.108.94:51593      105   149712     1
>>> Summary: total flows: 11241, total bytes: 75457199, total packets:
>>> 121691, avg bps: 1692074, avg pps: 341, avg bpp: 620
>>> Time window: 2015-08-21 14:58:59 - 2015-08-21 15:04:55
>>> Total flows processed: 11241, Blocks skipped: 0, Bytes read: 719528
>>> Sys: 0.083s flows/second: 135410.9   Wall: 0.081s flows/second: 137120.5
>>>
>>>
>>>
>>>
>>> 2015-08-21 14:03 GMT+03:00 Adrian Popa <adrian.popa...@gmail.com>:
>>>
>>>> If your server doesn't have routes in the routing table back to the
>>>> subnets of the source ips, then most likely the packets are dropped by
>>>> Linux's rp_filter protection mechanism. You can disable rp_filter on all
>>>> interfaces and see if there are any differences.
>>>>
>>>> On Fri, Aug 21, 2015 at 11:37 AM, Roman Mavrichev <
>>>> roman.mavric...@gmail.com> wrote:
>>>>
>>>>> The are workaround exists - for example, I set up on my routers /32
>>>>> adresses on Loopback`s from same subnet, where nfsen located, as source 
>>>>> for
>>>>> netflow traffic.
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>>
>>>>> _______________________________________________
>>>>> Nfsen-discuss mailing list
>>>>> Nfsen-discuss@lists.sourceforge.net
>>>>> https://lists.sourceforge.net/lists/listinfo/nfsen-discuss
>>>>>
>>>>>
>>>>
>>>
>>
>
------------------------------------------------------------------------------
_______________________________________________
Nfsen-discuss mailing list
Nfsen-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfsen-discuss

Reply via email to