A Zeek Cluster is a set of systems jointly analyzing the traffic of a network link in a coordinated fashion. You can operate such a setup from a central manager system easily using ZeekControl because it hides much of the complexity of the multi-machine installation.
Zeek is not multithreaded, so once the limitations of a single processor core are reached the only option currently is to spread the workload across many cores, or even many physical computers. The cluster deployment scenario for Zeek is the current solution to build these larger systems. The tools and scripts that accompany Zeek provide the structure to easily manage many Zeek processes examining packets and doing correlation activities but acting as a singular, cohesive entity. This document describes the Zeek cluster architecture. For information on how to configure a Zeek cluster, see the documentation for ZeekControl.
The figure below illustrates the main components of a Zeek cluster.
For more specific information on the way Zeek processes are connected, how they function, and how they communicate with each other, see the :ref:`Broker Framework Documentation <broker-framework>`.
The tap is a mechanism that splits the packet stream in order to make a copy available for inspection. Examples include the monitoring port on a switch and an optical splitter on fiber networks.
The frontend is a discrete hardware device or on-host technique that splits traffic into many streams or flows. The Zeek binary does not do this job. There are numerous ways to accomplish this task, some of which are described below in Frontend Options.
The manager is a Zeek process that has two primary jobs. It receives log messages and notices from the rest of the nodes in the cluster using the Zeek communications protocol (note that if you use a separate logger node, then the logger receives all logs instead of the manager). The result is a single log instead of many discrete logs that you have to combine in some manner with post-processing. The manager also supports other functionality and analysis which requires a centralized, global view of events or data.
A logger is an optional Zeek process that receives log messages from the rest of the nodes in the cluster using the Zeek communications protocol. The purpose of having a logger receive logs instead of the manager is to reduce the load on the manager. If no logger is needed, then the manager will receive logs instead.
A proxy is a Zeek process that may be used to offload data storage or any arbitrary workload. A cluster may contain multiple proxy nodes. The default scripts that come with Zeek make minimal use of proxies, so a single one may be sufficient, but customized use of them to partition data or workloads provides greater cluster scalability potential than just doing similar tasks on a single, centralized Manager node.
Zeek processes acting as proxies don't tend to be extremely hard on CPU or memory and users frequently run proxy processes on the same physical host as the manager.
The worker is the Zeek process that sniffs network traffic and does protocol analysis on the reassembled traffic streams. Most of the work of an active cluster takes place on the workers and as such, the workers typically represent the bulk of the Zeek processes that are running in a cluster. The fastest memory and CPU core speed you can afford is recommended since all of the protocol parsing and most analysis will take place here. There are no particular requirements for the disks in workers since almost all logging is done remotely to the manager, and normally very little is written to disk.
There are many options for setting up a frontend flow distributor. In many cases it is beneficial to do multiple stages of flow distribution on the network and on the host.
If you are monitoring one or more 10G physical interfaces, the recommended solution is to use either a cFlow or cVu device from cPacket because they are used successfully at a number of sites. These devices will perform layer-2 load balancing by rewriting the destination Ethernet MAC address to cause each packet associated with a particular flow to have the same destination MAC. The packets can then be passed directly to a monitoring host where each worker has a BPF filter to limit its visibility to only that stream of flows, or onward to a commodity switch to split the traffic out to multiple 1G interfaces for the workers. This greatly reduces costs since workers can use relatively inexpensive 1G interfaces.
We are currently exploring the use of OpenFlow based switches to do flow-based load balancing directly on the switch, which greatly reduces frontend costs for many users. This document will be updated when we have more information.
The PF_RING software for Linux has a "clustering" feature which will do flow-based load balancing across a number of processes that are sniffing the same interface. This allows you to easily take advantage of multiple cores in a single physical host because Zeek's main event loop is single threaded and can't natively utilize all of the cores. If you want to use PF_RING, see the documentation on :ref:`how to configure Zeek with PF_RING <pf-ring-config>`.
On Linux, Zeek supports AF_PACKET sockets natively. Currently, this is provided by including the external Zeek::AF_Packet plugin in default builds of Zeek for Linux. Additional information can be found in the project's README file.
To check the availability of the af_packet packet source, print its information using zeek -N:
zeek -N Zeek::AF_Packet Zeek::AF_Packet - Packet acquisition via AF_Packet (dynamic, version 3.2.0)
On FreeBSD, MacOSX, or if Zeek was built with --disable-af-packet, the
plugin won't be available.
For the most basic usage, prefix the interface with af_packet:: when invoking Zeek:
zeek -i af_packet::eth0
Generally, running Zeek this way requires a privileged user with CAP_NET_RAW and CAP_NET_ADMIN capabilities. Linux supports file-based capabilities: A process executing an executable with capabilities will receive these. Using this mechanism allows to run Zeek as an unprivileged user once the file capabilities have been added:
sudo setcap cap_net_raw,cap_net_admin=+eip /path/to/zeek
While not specific to AF_PACKET, it is recommended to disable any offloading features provided by the network card or Linux networking stack when running Zeek. This allows to see network packets as they arrive on the wire. See this blog post <https://blog.securityonion.net/2011/10/when-is-full-packet-capture-not-full.html>_ for more background
Toggling these features can be done with the ethtool -K command, for example:
IFACE=eth0 for offload in rx tx sg tso ufo gso gro lro; do ethtool -K $IFACE $offload off done
Detailed statistics about the interface can be gathered via ethtool -S.
For more details around the involved offloads consult the ethtool manpage.
The more interesting use-case is to use AF_PACKET to run multiple Zeek workers and have their packet sockets join what is called a fanout group. In such a setup, the network traffic is load-balanced across Zeek workers. By default load balancing is based on symmetric flow hashes [1].
For example, running two Zeek workers listening on the same network interface, each worker analyzing approximately half of the network traffic, can be done as follows:
zeek -i af_packet::eth0 & zeek -i af_packet::eth0 &
The fanout group is identified by an id and configurable using the
AF_Packet::fanout_id constant which defaults to 23. In the example
above, both Zeek workers join the same fanout group.
Note
As a caveat, within the same Linux network namespace, two Zeek processes can not use the same fanout group id for listening on different network interfaces. If this is a setup you're planning on running, configure the fanout group ids explicitly. For illustration purposes, the following starts two Zeek workers each using a different network interface and fanout group id:
zeek -i af_packet::eth0 AF_Packet::fanout_id=23 & zeek -i af_packet::eth1 AF_Packet::fanout_id=24 &
Warning
Zeek workers crashing or restarting due to running out of memory can, for a short period of time, disturb load balancing due to their packet sockets being removed and later rejoining the fanout group. This may be visible in Zeek logs as gaps and/or duplicated connection entries produced by different Zeek workers.
Netmap is a framework for fast packet I/O that is natively supported on FreeBSD since version 10. On Linux it can be installed as an out-of-tree kernel module.
FreeBSD's libpcap library supports netmap natively. This allows to prefix
interface names with netmap: to instruct libpcap to open the interface
in netmap mode. For example, a single Zeek worker can leverage netmap
transparently using Zeek's default packet source as follows:
zeek -i netmap:em0
Warning
Above command will put the em0 interface into kernel-bypass mode. Network packets will pass directly to Zeek without being interpreted by the kernel. If em0 is your primary network interface, this effectively disables networking, including SSH connectivity.
If your network card supports multiple rings, individual Zeek workers can be attached to these as well (this assumes the NIC does proper flow hashing in hardware):
zeek -i netmap:em0-0 zeek -i netmap:em0-1
For software load balancing support, the FreeBSD source tree includes the
lb tool to distribute packets into netmap pipes doing flow hashing
in user-space.
To compile and install lb, ensure /usr/src is available on your
FreeBSD system, then run the following commands:
cd /usr/src/tools/tools/netmap/ make # Installs lb into /usr/local/bin cp /usr/obj/usr/src/`uname -m`.`uname -m`/tools/tools/netmap/lb /usr/local/bin/
To load-balance packets arriving on em0 into 4 different netmap pipes named
zeek}0 through zeek}3, run lb as follows:
lb -i em0 -p zeek:4
410.154166 main [634] interface is em0
411.377220 main [741] successfully opened netmap:em0
411.377243 main [812] opening pipe named netmap:zeek{0/xT@1
411.379200 main [829] successfully opened pipe #1 netmap:zeek{0/xT@1 (tx slots: 1024)
411.379242 main [838] zerocopy enabled
...
Now, Zeek workers can attach to these four netmap pipes. When starting Zeek
workers manually, the respective invocations would be as follows. The /x
suffix specifies exclusive mode to prevent two Zeek processes consuming packets
from the same netmap pipe:
zeek -i netmap:zeek}0/x zeek -i netmap:zeek}1/x zeek -i netmap:zeek}2/x zeek -i netmap:zeek}3/x
For packet-level debugging, you can attach tcpdump to any of the netmap
pipes in read monitor mode even while Zeek workers are consuming from them:
tcpdump -i netmap:zeek}1/r
In case libpcap's netmap support is insufficient, the external Zeek netmap plugin can be installed.
Warning
When using the zeek-netmap plugin on FreeBSD, the interface specification given to Zeek
needs to change from netmap:zeek}0/x to netmap::zeek}0/x - a single colon more.
In the first case, Zeek uses the default libpcap packet source and passes netmap:zeek}0
as interface name. In the second case, netmap:: is interpreted by Zeek and
the netmap packet source is instantiated. The zeek}0/x part is used as
interface name.
While netmap isn't included in the Linux kernel, it can be installed as
an out-of-tree kernel module.
See the project's GitHub repository
for detailed instructions. This includes the lb tool for load balancing.
On Linux, the external zeek-netmap packet source plugin is required, or the system's libpcap library as used by Zeek needs to be recompiled with native netmap support. With the netmap kernel module loaded and the Zeek plugin installed, running a Zeek worker as follows will leverage netmap on Linux:
zeek -i netmap::eth1
For using lb or libpcap with netmap support, refer to the commands shown
in the FreeBSD section - these are essentially the same.
Click! can be used for flow based load balancing with a simple configuration. This solution is not recommended on Linux due to Zeek's PF_RING support and only as a last resort on other operating systems since it causes a lot of overhead due to context switching back and forth between kernel and userland several times per packet.
A Zeek Cluster is a set of systems jointly analyzing the traffic of a network link in a coordinated fashion. You can operate such a setup from a central manager system easily using ZeekControl because it hides much of the complexity of the multi-machine installation.
This section gives examples of how to setup common cluster configurations using ZeekControl. For a full reference on ZeekControl, see the ZeekControl documentation.
In this document we refer to the user account used to set up the cluster as the "Zeek user". When setting up a cluster the Zeek user must be set up on all hosts, and this user must have ssh access from the manager to all machines in the cluster, and it must work without being prompted for a password/passphrase (for example, using ssh public key authentication). Also, on the worker nodes this user must have access to the target network interface in promiscuous mode.
Additional storage must be available on all hosts under the same path,
which we will call the cluster's prefix path. We refer to this directory
as <prefix>. If you build Zeek from source, then <prefix> is
the directory specified with the --prefix configure option,
or /usr/local/zeek by default. The Zeek user must be able to either
create this directory or, where it already exists, must have write
permission inside this directory on all hosts.
When trying to decide how to configure the Zeek nodes, keep in mind that there can be multiple Zeek instances running on the same host. For example, it's possible to run a proxy and the manager on the same host. However, it is recommended to run workers on a different machine than the manager because workers can consume a lot of CPU resources. The maximum recommended number of workers to run on a machine should be one or two less than the number of CPU cores available on that machine. Using a load-balancing method (such as PF_RING) along with CPU pinning can decrease the load on the worker machines. Also, in order to reduce the load on the manager process, it is recommended to have a logger in your configuration. If a logger is defined in your cluster configuration, then it will receive logs instead of the manager process.
With all prerequisites in place, perform the following steps to setup a Zeek cluster (do this as the Zeek user on the manager host only):
Edit the ZeekControl configuration file,
<prefix>/etc/zeekctl.cfg, and change the value of any options to be more suitable for your environment. You will most likely want to change the value of theMailToandLogRotationIntervaloptions. A complete reference of all ZeekControl options can be found in the ZeekControl documentation.Edit the ZeekControl node configuration file,
<prefix>/etc/node.cfgto define where logger, manager, proxies, and workers are to run. For a cluster configuration, you must comment-out (or remove) the standalone node in that file, and either uncomment or add node entries for each node in your cluster (logger, manager, proxy, and workers). For example, if you wanted to run five Zeek nodes (two workers, one proxy, a logger, and a manager) on a cluster consisting of three machines, your cluster configuration would look like this:[logger] type=logger host=10.0.0.10 [manager] type=manager host=10.0.0.10 [proxy-1] type=proxy host=10.0.0.10 [worker-1] type=worker host=10.0.0.11 interface=eth0 [worker-2] type=worker host=10.0.0.12 interface=eth0
For a complete reference of all options that are allowed in the
node.cfgfile, see the ZeekControl documentation.Edit the network configuration file
<prefix>/etc/networks.cfg. This file lists all of the networks which the cluster should consider as local to the monitored environment.Install Zeek on all machines in the cluster using ZeekControl:
> zeekctl install
See the ZeekControl documentation for information on setting up a cron job on the manager host that can monitor the cluster.
PF_RING allows speeding up the packet capture process by installing a new type of socket in Linux systems. It supports 10Gbit hardware packet filtering using standard network adapters, and user-space DNA (Direct NIC Access) for fast packet capture/transmission.
Download and install PF_RING for your system following the instructions here. The following commands will install the PF_RING libraries and kernel module (replace the version number 5.6.2 in this example with the version that you downloaded):
cd /usr/src tar xvzf PF_RING-5.6.2.tar.gz cd PF_RING-5.6.2/userland/lib ./configure --prefix=/opt/pfring make install cd ../libpcap ./configure --prefix=/opt/pfring make install cd ../tcpdump-4.1.1 ./configure --prefix=/opt/pfring make install cd ../../kernel make make install modprobe pf_ring enable_tx_capture=0 min_num_slots=32768
Refer to the documentation for your Linux distribution on how to load the pf_ring module at boot time. You will need to install the PF_RING library files and kernel module on all of the workers in your cluster.
Download the Zeek source code.
Configure and install Zeek using the following commands:
./configure --with-pcap=/opt/pfring make make install
Make sure Zeek is correctly linked to the PF_RING libpcap libraries:
ldd /usr/local/zeek/bin/zeek | grep pcap libpcap.so.1 => /opt/pfring/lib/libpcap.so.1 (0x00007fa6d7d24000)Configure ZeekControl to use PF_RING (explained below).
Run "zeekctl install" on the manager. This command will install Zeek and required scripts to all machines in your cluster.
In order to use PF_RING, you need to specify the correct configuration
options for your worker nodes in ZeekControl's node configuration file.
Edit the node.cfg file and specify lb_method=pf_ring for each of
your worker nodes. Next, use the lb_procs node option to specify how
many Zeek processes you'd like that worker node to run, and optionally pin
those processes to certain CPU cores with the pin_cpus option (CPU
numbering starts at zero). The correct pin_cpus setting to use is
dependent on your CPU architecture (Intel and AMD systems enumerate
processors in different ways). Using the wrong pin_cpus setting
can cause poor performance. Here is what a worker node entry should
look like when using PF_RING and CPU pinning:
[worker-1] type=worker host=10.0.0.50 interface=eth0 lb_method=pf_ring lb_procs=10 pin_cpus=2,3,4,5,6,7,8,9,10,11
You must have a PF_RING+DNA license in order to do this. You can sniff each packet only once.
Load the DNA NIC driver (i.e. ixgbe) on each worker host.
Run "ethtool -L dna0 combined 10" (this will establish 10 RSS queues on your NIC) on each worker host. You must make sure that you set the number of RSS queues to the same as the number you specify for the lb_procs option in the node.cfg file.
On the manager, configure your worker(s) in node.cfg:
[worker-1] type=worker host=10.0.0.50 interface=dna0 lb_method=pf_ring lb_procs=10
You must have a PF_RING+DNA license and a libzero license in order to do this. You can load balance between multiple applications and sniff the same packets multiple times with different tools.
Load the DNA NIC driver (i.e. ixgbe) on each worker host.
Run "ethtool -L dna0 1" (this will establish 1 RSS queues on your NIC) on each worker host.
Run the pfdnacluster_master command on each worker host. For example:
pfdnacluster_master -c 21 -i dna0 -n 10
Make sure that your cluster ID (21 in this example) matches the interface name you specify in the node.cfg file. Also make sure that the number of processes you're balancing across (10 in this example) matches the lb_procs option in the node.cfg file.
If you are load balancing to other processes, you can use the pfringfirstappinstance variable in zeekctl.cfg to set the first application instance that Zeek should use. For example, if you are running pfdnacluster_master with "-n 10,4" you would set pfringfirstappinstance=4. Unfortunately that's still a global setting in zeekctl.cfg at the moment but we may change that to something you can set in node.cfg eventually.
On the manager, configure your worker(s) in node.cfg:
[worker-1] type=worker host=10.0.0.50 interface=dnacluster:21 lb_method=pf_ring lb_procs=10
| [1] | Some Linux kernel versions between 3.10 and 4.7 might exhibit
a bug that prevents the required symmetric hashing. The script available
in the GitHub project can-i-use-afpacket-fanout
can be used to verify whether This issue has been fixed in all stable kernels for at least 5 years. You're unlikely to be affected. |