Mirror port - Dropping all packets

[bibawa]

Hi,

I just installed a new Fasnetmon installation on a Ubuntu 14.04 server, the installation itself went smoothly.

We want to mirror our 2 core switches to this fastnetmon box. If I start the fasnetmon daemon I see a lot of packets coming into it but all packates are being dropped:

Total amount of IPv6 packets related to our own network: 0 Not processed packets: 0 pps Packets received: 915 Packets dropped: 148713 Packets dropped: 16252.8 %

###
### Main configuration params
###

### Logging configuration

# enable this option if you want to send logs to local syslog facility
logging:local_syslog_logging = off

# enable this option if you want to send logs to a remote syslog server via UDP
logging:remote_syslog_logging = off

# specify a custom server and port for remote logging
logging:remote_syslog_server = 10.10.10.10
logging:remote_syslog_port = 514

# Enable/Disable any actions in case of attack
enable_ban = on

# disable processing for certain direction of traffic
process_incoming_traffic = on
process_outgoing_traffic = on

# How many packets will be collected from attack traffic
ban_details_records_count = 500

# How long (in seconds) we should keep an IP in blocked state
# If you set 0 here it completely disables unban capability
ban_time = 1900

# Check if the attack is still active, before triggering an unban callback with this option
# If the attack is still active, check each run of the unban watchdog
unban_only_if_attack_finished = on

# enable per subnet speed meters
# For each subnet, list track speed in bps and pps for both directions
enable_subnet_counters = off

# list of all your networks in CIDR format
networks_list_path = /etc/networks_list

# list networks in CIDR format which will be not monitored for attacks
white_list_path = /etc/networks_whitelist

# redraw period for client's screen
check_period = 1

# Connection tracking is very useful for attack detection because it provides huge amounts of information,
# but it's very CPU intensive and not recommended in big networks
enable_connection_tracking = off

# Different approaches to attack detection
ban_for_pps = on
ban_for_bandwidth = on
ban_for_flows = off

# Limits for Dos/DDoS attacks
threshold_pps = 20000
threshold_mbps = 1000
threshold_flows = 3500

# Per protocol attack thresholds
# We don't implement per protocol flow limits, sorry :(
# These limits should be smaller than global pps/mbps limits

threshold_tcp_mbps = 100000
threshold_udp_mbps = 100000
threshold_icmp_mbps = 100000

threshold_tcp_pps = 100000
threshold_udp_pps = 100000
threshold_icmp_pps = 100000

ban_for_tcp_bandwidth = off
ban_for_udp_bandwidth = off
ban_for_icmp_bandwidth = off

ban_for_tcp_pps = off 
ban_for_udp_pps = off
ban_for_icmp_pps = off

###
### Traffic capture methods
###

# PF_RING traffic capture, fast enough but the wirespeed version needs a paid license
mirror = on

# Port mirroring sample rate
pfring_sampling_ratio = 1

# Netmap traffic capture (very fast but needs patched drivers)
mirror_netmap = off

# SnabbSwitch traffic capture
mirror_snabbswitch = off

# AF_PACKET capture engine
# Please use it only with modern Linux kernels (3.6 and more)
# And please install birq for irq ditribution over cores
mirror_afpacket = off

# use PCI-e addresses here instead of OS device names. You can find them in "lspci" output
interfaces = 02:00.0

# Port mirroring sampling ratio
netmap_sampling_ratio = 1

# This option should be enabled if you are using Juniper with mirroring of the first X bytes of packet: maximum-packet-length 110;
netmap_read_packet_length_from_ip_header = off

# Pcap mode, very slow and thus not suitable for production
pcap = off
# Netflow capture method with v5, v9 and IPFIX support
netflow = off
# sFLOW capture suitable for switches
sflow = off

# PF_RING configuration
# If you have a license for PF_RING ZC, enable this mode and it might achieve wire speed for 10GE
enable_pf_ring_zc_mode = off

# Configuration for netmap, mirror, pcap modes
# For pcap and PF_RING we could specify "any"
# For netmap and PF_RING we could specify multiple interfaces = em1
interfaces = any

# We use average values for traffic speed to certain IP and we calculate average over this time slice
average_calculation_time = 5

# We use average values for traffic speed for subnet and we calculate average over this time slice
average_calculation_time_for_subnets = 20

# Netflow configuration

# it's possible to specify multiple ports here, using commas as delimiter
netflow_port = 2055
netflow_host = 0.0.0.0

# To bind to all interfaces = em1
# To bind to all interfaces = em1
# To bind to localhost for a specific protocol:      ::1 or 127.0.0.1

# Netflow v9 and IPFIX agents use different and very complex approaches for notifying about sample ratio
# Here you could specify a sampling ratio for all this agents
# For NetFLOW v5 we extract sampling ratio from packets directely and this option not used
netflow_sampling_ratio = 1

# In some cases with NetFlow we could get huge bursts related to aggregated data nature
# We could try to get smoother data with this option, i.e. we will divide counters on collection interval time
netflow_divide_counters_on_interval_length = off

# Process each netflow packet with LUA
# This option is not default and you need build it additionally
# netflow_lua_hooks_path = /usr/src/fastnetmon/src/netflow_hooks.lua

# sFLOW configuration

# It's possible to specify multiple ports here, using commas as delimiter
sflow_port = 6343
# sflow_port = 6343,6344
sflow_host = 0.0.0.0

# process each sFLOW packet with LUA
# This option is not default and you need build it additionally
# sflow_lua_hooks_path = /usr/src/fastnetmon/src/sflow_hooks.lua

###
### Actions when attack detected
###

# This script executed for ban, unban and attack detail collection
notify_script_path = /usr/local/bin/notify_about_attack.sh

# pass attack details to notify_script via stdin
# Pass details only in case of "ban" call
# No details will be passed for "unban" call
notify_script_pass_details = on

# collect a full dump of the attack with full payload in pcap compatible format
collect_attack_pcap_dumps = off

# Execute Deep Packet Inspection on captured PCAP packets
process_pcap_attack_dumps_with_dpi = off

# Save attack details to Redis
redis_enabled = off

# Redis configuration
redis_port = 6379
redis_host = 127.0.0.1

# specify a custom prefix here
redis_prefix = mydc1

# We could store attack information to MongoDB
mongodb_enabled = off
mongodb_host = localhost
mongodb_port = 27017
mongodb_database_name = fastnetmon

# If you are using PF_RING non ZC version you could block traffic on host with hardware filters
# Please be aware! We can not remove blocks with this action plugin
pfring_hardware_filters_enabled = off

# announce blocked IPs with BGP protocol with ExaBGP
exabgp = off
exabgp_command_pipe = /var/run/exabgp.cmd
exabgp_community = 65001:666

# specify multiple communities with this syntax:
# exabgp_community = [65001:666 65001:777]

# specify different communities for host and subnet announces
# exabgp_community_subnet = 65001:667
# exabgp_community_host = 65001:668

exabgp_next_hop = 10.0.3.114

# In complex cases you could have both options enabled and announce host and subnet simultaneously

# Announce /32 host itself with BGP
exabgp_announce_host = on

# Announce origin subnet of IP address instead IP itself
exabgp_announce_whole_subnet = off

# Announce Flow Spec rules when we could detect certain attack type
# Please we aware! Flow Spec announce triggered when we collect some details about attack,
# i.e. when we call attack_details script
# Please disable exabgp_announce_host and exabgp_announce_whole_subnet if you want to use this feature
# Please use ExaBGP v4 only (Git version), for more details: https://github.com/FastVPSEestiOu/fastnetmon/blob/master/docs/BGP_FLOW_SPEC.md
exabgp_flow_spec_announces = off

# GoBGP intergation
gobgp = off
gobgp_next_hop = 0.0.0.0
gobgp_announce_host = on
gobgp_announce_whole_subnet = off

# Graphite monitoring
# InfluxDB is also supported, please check our reference:
# https://github.com/FastVPSEestiOu/fastnetmon/blob/master/docs/INFLUXDB_INTEGRATION.md
graphite = off
graphite_host = 127.0.0.1
graphite_port = 2003

# Default namespace for Graphite data
graphite_prefix = fastnetmon

# Add local IP addresses and aliases to monitoring list
# Works only for Linux
monitor_local_ip_addresses = on

# Create group of hosts with non-standard thresholds
# You should create this group before (in configuration file) specifying any limits
hostgroup = my_hosts:10.10.10.221/32,10.10.10.222/32

# Configure this group
my_hosts_enable_ban = off

my_hosts_ban_for_pps = off
my_hosts_ban_for_bandwidth = off
my_hosts_ban_for_flows = off

my_hosts_threshold_pps = 20000
my_hosts_threshold_mbps = 1000
my_hosts_threshold_flows = 3500

# Path to pid file for checking "if another copy of tool is running", it's useful when you run multiple instances of tool
pid_path = /var/run/fastnetmon.pid

# Path to file where we store information for fastnetmon_client
cli_stats_file_path = /tmp/fastnetmon.dat

# Enable gRPC api (required for fastnetmon_api_client tool)
enable_api = off

###
### Client configuration
###

# Field used for sorting in client, valid values are: packets, bytes or flows
sort_parameter = packets
# How much IPs will be listed for incoming and outgoing channel eaters
max_ips_in_list = 7 

What am I missing ?

Thank for your help !

[pavel-odintsov]

Hello

How much traffic do you have in mbps / packets per second?

[bibawa]

Hi,

Thanks for your fast response.. That will be around 500Mbps..

[pavel-odintsov]

What's your cpu? Do you have any other tools on same server?

Pfring works fine for 2-6GE.

[bibawa]

CPU= E3-1220V2 This server is only used for fastnetmon,

I changed our switch to sflow and now it is working, but wouldn't it be better to use por mirror to be more faster ?

brg,

[pavel-odintsov]

Sflow is best option ;) I recommend it.

[winkmichael]

In my experience pf_ring_zc (the commercial version) is faster than sflow, however if you just use the free version then go with sflow. One of the benefits with pf_ring_zc is you don’t loose data when things get very busy.

Cheers, Mike

-- Michael McConnell WINK Streaming; email: michael@winkstreaming.com mailto:michael@winkstreaming.com phone: +1 312 281-5433 x 7400 cell: +506 8706-2389 skype: wink-michael web: http://winkstreaming.com http://winkstreaming.com/

On Mar 18, 2017, at 12:05 PM, bibawa notifications@github.com wrote:

CPU= E3-1220V2 This server is only used for fastnetmon,

I changed our switch to sflow and now it is working, but wouldn't it be better to use por mirror to be more faster ?

brg,

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/pavel-odintsov/fastnetmon/issues/639#issuecomment-287563621, or mute the thread https://github.com/notifications/unsubscribe-auth/AE8hYZ3IJM3iwab9xgd5K9gOSisIltKiks5rnBzVgaJpZM4MhdcJ.

[fire002]

Use sudo perl fastnetmon_install.pl --use-modern-pf-ring --it's build a new version of pf_ring. It's helped me on Debian Jessie with 3.16.0-4-amd64 kernel