[Bro-Dev] Bro 2.5 Packet Drop Issue
Azoff, Justin S
jazoff at illinois.edu
Thu Aug 30 13:28:37 PDT 2018
> On Aug 30, 2018, at 4:11 PM, Rajput, Jawad (CONTR) <Jawad.Rajput at hq.doe.gov> wrote:
>
> Hello Everyone,
>
> I am reaching out with the hope that someone will be able to help us with an issue we are having with Bro upgrade from 2.4.1 to 2.5.X.
>
> We have a system with 12 core (3Ghz) ,128GB RAM, and 10G NIC (Intel X520-SR2 10GbE Dual-port), monitoring between 1.5 - 2.5 Gbps traffic.
>
> Bro 2.4.1 is working great and periodically drops 2-5% when traffic peaks at ~ 2.5. However, when we upgrade to Bro 2.5.3/4 on the same exact system the drops go up to 90%.
>
> We are using CentOS-7 and tired installing Bro and Pfring from both rpm and source without any luck. I wonder if anyone has seen this issue and can give some clues to resolve this issue.
>
> Bro Node Conf:
> [manager]
> type=manager
> host=localhost
> #
> [proxy-1]
> type=proxy
> host=localhost
>
> #
> [worker-1]
> type=worker
> host=localhost
> interface=ens1f1
> lb_method=pf_ring
> lb_procs=11
> pin_cpus=1,2,3,4,5,6,7,8,9,10,11
You're missing a logger process, adding one will make the cluster run better:
[logger]
type=logger
host=localhost
> [root at bro-test ~]# cat /proc/net/pf_ring/info
> PF_RING Version : 7.3.0 (unknown)
> Total rings : 11
you should have 1, not 11...
> Standard (non ZC) Options
> Ring slots : 65534
> Slot version : 17
> Capture TX : No [RX only]
> IP Defragment : No
> Socket Mode : Standard
> Cluster Fragment Queue : 0
> Cluster Fragment Discard : 0
Looks like you are having the issue where bro is not actually use pf_ring load balancing if you installed it from rpms.
What you're effectively doing is running 11 workers that are all receiving 100% of the traffic, so you are doing 11 times the work.
You can further confirm that this is the problem you are having by running
broctl config | grep -i clusterid
and seeing if the id is set to 0:
pfringclusterid = 0
if so, edit /opt/bro/etc/broctl.cfg and add
PFRINGClusterID = 11
and broctl deploy to restart everything.
This is already fixed and won't happen again in bro >= 2.6... just keeps tripping people up on 2.5.x
You should also look into switching to the native bro pf_ring plugin or the bro af_packet plugin which are both better choices than using the pcap wrapper method.
—
Justin Azoff
More information about the bro-dev
mailing list