Hi Doug
the code in pfcount sets the cluster mode to round-robin,
for flow coherency you should change it to (for instance)
cluster_per_flow_2_tuple.
The daq-pfring code sets the cluster mode to cluster_per_flow_2_tuple by
default.
Best Regards
Alfredo
Index: pfcount.c
===================================================================
--- pfcount.c (revisione 6336)
+++ pfcount.c (copia locale)
@@ -924,7 +924,7 @@
#endif
if(clusterId > 0) {
- rc = pfring_set_cluster(pd, clusterId, cluster_round_robin);
+ rc = pfring_set_cluster(pd, clusterId, cluster_per_flow_2_tuple);
printf("pfring_set_cluster returned %d\n", rc);
}
On Jun 2, 2013, at 2:54 PM, Doug Burks <[email protected]> wrote:
> I copied the clusterId code from pfcount and pasted into pfdump and
> compiled it. Then tested with a fresh pcap of "curl testmyids.com":
>
> tcpdump -nnr testmyids.pcap
> reading from file testmyids.pcap, link-type EN10MB (Ethernet)
> 12:37:21.846561 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [S],
> seq 2183306783, win 42340, options [mss 1460,sackOK,TS val 13599714
> ecr 0,nop,wscale 11], length 0
> 12:37:21.963023 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [S.], seq 3354284181, ack 2183306784, win 64240, options [mss 1460],
> length 0
> 12:37:21.963070 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 1, win 42340, length 0
> 12:37:21.963268 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags
> [P.], seq 1:166, ack 1, win 42340, length 165
> 12:37:21.963423 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags [.],
> ack 166, win 64240, length 0
> 12:37:22.083864 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [P.], seq 1:260, ack 166, win 64240, length 259
> 12:37:22.083906 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 260, win 42081, length 0
> 12:37:22.084118 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags
> [F.], seq 166, ack 260, win 42081, length 0
> 12:37:22.085362 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags [.],
> ack 167, win 64239, length 0
> 12:37:22.202741 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [FP.], seq 260, ack 167, win 64239, length 0
> 12:37:22.202786 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 261, win 42081, length 0
>
> I then started the two instances of pfdump using the same clusterId
> and then replayed the 11 packets with tcpreplay:
> sudo tcpreplay -i eth1 -M10 testmyids.pcap
> sending out eth1
> processing file: testmyids.pcap
> Actual: 11 packets (1062 bytes) sent in 0.01 seconds
> Rated: 106200.0 bps, 0.81 Mbps, 1100.00 pps
> Statistics for network device: eth1
> Attempted packets: 11
> Successful packets: 11
> Failed packets: 0
> Retried packets (ENOBUFS): 0
> Retried packets (EAGAIN): 0
>
> FIRST INSTANCE OF PFDUMP
>
> sudo ./pfdump -l77 -i eth1 -w instance1.pcap
> Using PF_RING v.5.5.3
> Capturing from eth1 [00:0C:29:5F:58:D8][ifIndex: 3]
> # Device RX channels: 1
> pfring_set_cluster returned 0
> <snip>
> 241 sec pkts 6 drop 0 bytes 500 | pkts 6 bytes 500 drop 0
> <snip>
>
> tcpdump -nnr instance1.pcap
> reading from file instance1.pcap, link-type EN10MB (Ethernet)
> 12:38:55.886037 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [S],
> seq 2183306783, win 42340, options [mss 1460,sackOK,TS val 13599714
> ecr 0,nop,wscale 11], length 0
> 12:38:55.886889 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 3354284182, win 42340, length 0
> 12:38:55.887325 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags [.],
> ack 165, win 64240, length 0
> 12:38:55.887986 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 260, win 42081, length 0
> 12:38:55.888306 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags [.],
> ack 166, win 64239, length 0
> 12:38:55.888741 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags [.],
> ack 261, win 42081, length 0
>
> SECOND INSTANCE OF PFDUMP
>
> sudo ./pfdump -l77 -i eth1 -w instance2.pcap
> Using PF_RING v.5.5.3
> Capturing from eth1 [00:0C:29:5F:58:D8][ifIndex: 3]
> # Device RX channels: 1
> pfring_set_cluster returned 0
> <snip>
> 16 sec pkts 5 drop 0 bytes 826 | pkts 5 bytes 826 drop 0
> 17 sec pkts 0 drop 0 bytes 0 | pkts 5 bytes 826 drop 0
> 18 sec pkts 0 drop 0 bytes 0 | pkts 5 bytes 826 drop 0
> 19 sec pkts 0 drop 0 bytes 0 | pkts 5 bytes 826 drop 0
> ^CLeaving...
> 20 sec pkts 0 drop 0 bytes 0 | pkts 5 bytes 826 drop 0
>
> tcpdump -nnr instance2.pcap
> reading from file instance2.pcap, link-type EN10MB (Ethernet)
> 12:38:55.886499 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [S.], seq 3354284181, ack 2183306784, win 64240, options [mss 1460],
> length 0
> 12:38:55.887129 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags
> [P.], seq 1:166, ack 1, win 42340, length 165
> 12:38:55.887666 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [P.], seq 1:260, ack 166, win 64240, length 259
> 12:38:55.888117 IP 172.16.116.128.44229 > 217.160.51.31.80: Flags
> [F.], seq 166, ack 260, win 42081, length 0
> 12:38:55.888530 IP 217.160.51.31.80 > 172.16.116.128.44229: Flags
> [FP.], seq 260, ack 167, win 64239, length 0
>
> As you can see, the first instance sees 6 packets and the second
> instance sees 5 packets. Shouldn't all 11 packets in that TCP stream
> be sent to the same instance?
>
> Thanks!
>
> Doug
>
>
> On Sun, Jun 2, 2013 at 7:11 AM, Doug Burks <[email protected]> wrote:
>> Hi Luca,
>>
>> I can repeat the test with pfdump when I'm back at my computer, but is there
>> something in particular you're looking for that wasn't in the pfcount output
>> I provided? Shouldn't all the traffic from that one TCP stream be sent to
>> one instance of pfcount?
>>
>> Thanks,
>> Doug
>>
>>
>> On Sunday, June 2, 2013, Luca Deri wrote:
>>>
>>> Hi
>>> You're right. We need to add it: you can c&p the code from pfcount in the
>>> meantime
>>>
>>> Luca
>>>
>>> On Jun 2, 2013, at 1:54 AM, Doug Burks <[email protected]> wrote:
>>>
>>>> I have pfdump now but I don't see a cluster-id option. Did you mean
>>>> pfcount? If I run 2 instances of pfcount with the same cluster-id and
>>>> then replay a pcap with 10 packets all belonging to the same TCP
>>>> stream, I get 5 packets being sent to each pfcount instance.
>>>> Shouldn't all 10 packets be sent to 1 instance?
>>>>
>>>> First instance:
>>>>
>>>> sudo ./pfcount -c77 -i eth1
>>>> <snip>
>>>> =========================
>>>> Absolute Stats: [5 pkts rcvd][5 pkts filtered][0 pkts dropped]
>>>> Total Pkts=5/Dropped=0.0 %
>>>> 5 pkts - 434 bytes [0.38 pkt/sec - 0.00 Mbit/sec]
>>>> =========================
>>>> Actual Stats: 5 pkts [1'000.75 ms][5.00 pps/0.00 Gbps]
>>>> =========================
>>>>
>>>> Second instance:
>>>>
>>>> sudo ./pfcount -c77 -i eth1
>>>> <snip>
>>>> =========================
>>>> Absolute Stats: [5 pkts rcvd][5 pkts filtered][0 pkts dropped]
>>>> Total Pkts=5/Dropped=0.0 %
>>>> 5 pkts - 834 bytes [0.62 pkt/sec - 0.00 Mbit/sec]
>>>> =========================
>>>> Actual Stats: 5 pkts [1'001.39 ms][4.99 pps/0.00 Gbps]
>>>> =========================
>>>>
>>>> The replayed pcap is just ten packets that result from "curl
>>>> testmyids.com":
>>>>
>>>> tcpdump -nnr testmyids.pcap
>>>> reading from file testmyids.pcap, link-type EN10MB (Ethernet)
>>>> 11:46:11.691648 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [S], seq 3840903154, win 42340, options [mss 1460,sackOK,TS val
>>>> 20137183 ecr 0,nop,wscale 11], length 0
>>>> 11:46:11.808833 IP 217.160.51.31.80 > 192.168.111.111.50154: Flags
>>>> [S.], seq 2859277445, ack 3840903155, win 5840, options [mss
>>>> 1460,nop,wscale 7], length 0
>>>> 11:46:11.808854 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [.], ack 1, win 21, length 0
>>>> 11:46:11.809083 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [P.], seq 1:166, ack 1, win 21, length 165
>>>> 11:46:11.927518 IP 217.160.51.31.80 > 192.168.111.111.50154: Flags
>>>> [.], ack 166, win 54, length 0
>>>> 11:46:12.036708 IP 217.160.51.31.80 > 192.168.111.111.50154: Flags
>>>> [P.], seq 1:260, ack 166, win 54, length 259
>>>> 11:46:12.036956 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [.], ack 260, win 21, length 0
>>>> 11:46:12.037206 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [F.], seq 166, ack 260, win 21, length 0
>>>> 11:46:12.154641 IP 217.160.51.31.80 > 192.168.111.111.50154: Flags
>>>> [F.], seq 260, ack 167, win 54, length 0
>>>> 11:46:12.154888 IP 192.168.111.111.50154 > 217.160.51.31.80: Flags
>>>> [.], ack 261, win 21, length 0
>>>>
>>>> Any ideas?
>>>>
>>>> Thanks,
>>>> Doug
>>>>
>>>> On Sat, Jun 1, 2013 at 5:48 PM, Doug Burks <[email protected]> wrote:
>>>>> On Sat, Jun 1, 2013 at 10:24 AM, Luca Deri <[email protected]> wrote:
>>>>>> Hi Doug
>>>>>>
>>>>>> On Jun 1, 2013, at 6:59 AM, Doug Burks <[email protected]> wrote:
>>>>>>
>>>>>>> Hello all,
>>>>>>>
>>>>>>> I recently packaged PF_RING 5.5.3 for my Security Onion distro:
>>>>>>>
>>>>>>> http://securityonion.blogspot.com/2013/05/pfring-553-packages-now-available.html
>>>>>>>
>>>>>>> Perhaps I'm missing something, but I'm seeing some behavior I don't
>>>>>>> remember seeing in 5.5.2 or previous versions of PF_RING.
>>>>>>>
>>>>>>> Here are my testing parameters:
>>>>>>> - starting off with a good test, if I run just one instance of snort,
>>>>>>> I get an alert from rule 2100498 for EACH time I run "curl
>>>>>>> testmyids.com"
>>>>>>> - if I increase to two instances of snort with the same cluster-id, I
>>>>>>> get NO alerts when running "curl testmyids.com"
>>>>>>> - if I set the daq clustermode to 2, I get NO alerts when running
>>>>>>> "curl > _______________________________________________
>>>
>>>> Ntop-misc mailing list
>>>> [email protected]
>>>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>>>
>>> _______________________________________________
>>> Ntop-misc mailing list
>>> [email protected]
>>> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
>>
>>
>>
>> --
>> Doug Burks
>> http://securityonion.blogspot.com
>>
>
>
>
> --
> Doug Burks
> http://securityonion.blogspot.com
> _______________________________________________
> Ntop-misc mailing list
> [email protected]
> http://listgateway.unipi.it/mailman/listinfo/ntop-misc
_______________________________________________
Ntop-misc mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop-misc
