We have given special attention to all aspects of Netdata, ensuring that everything throughout its operation is as secure as possible. Netdata has been designed with security in mind.
Table of Contents
- Your data are safe with Netdata
- Your systems are safe with Netdata
- Netdata is read-only
- Netdata viewers authentication
- Registry or how to not send any information to a third party server
Netdata collects raw data from many sources. For each source, Netdata uses a plugin that connects to the source (or reads the relative files produced by the source), receives raw data and processes them to calculate the metrics shown on Netdata dashboards.
Even if Netdata plugins connect to your database server, or read your application log file to collect raw data, the product of this data collection process is always a number of chart metadata and metric values (summarized data for dashboard visualization). All Netdata plugins (internal to the Netdata daemon, and external ones written in any computer language), convert raw data collected into metrics, and only these metrics are stored in Netdata databases, sent to upstream Netdata servers, or archived to backend time-series databases.
The raw data collected by Netdata, do not leave the host they are collected. The only data Netdata exposes are chart metadata and metric values.
This means that Netdata can safely be used in environments that require the highest level of data isolation (like PCI Level 1).
We are very proud that the Netdata daemon runs as a normal system user, without any special privileges. This is quite an achievement for a monitoring system that collects all kinds of system and application metrics.
There are a few cases however that raw source data are only exposed to processes with escalated privileges. To support these cases, Netdata attempts to minimize and completely isolate the code that runs with escalated privileges.
So, Netdata plugins, even those running with escalated capabilities or privileges, perform a hard coded data collection job. They do not accept commands from Netdata. The communication is strictly unidirectional: from the plugin towards the Netdata daemon. The original application data collected by each plugin do not leave the process they are collected, are not saved and are not transferred to the Netdata daemon. The communication from the plugins to the Netdata daemon includes only chart metadata and processed metric values.
Netdata slaves streaming metrics to upstream Netdata servers, use exactly the same protocol local plugins use. The raw data collected by the plugins of slave Netdata servers are never leaving the host they are collected. The only data appearing on the wire are chart metadata and metric values. This communication is also unidirectional: slave Netdata servers never accept commands from master Netdata servers.
Netdata dashboards are read-only. Dashboard users can view and examine metrics collected by Netdata, but cannot instruct Netdata to do something other than present the already collected metrics.
Netdata dashboards do not expose sensitive information. Business data of any kind, the kernel version, O/S version, application versions, host IPs, etc are not stored and are not exposed by Netdata on its dashboards.
Netdata is a monitoring system. It should be protected, the same way you protect all your admin apps. We assume Netdata will be installed privately, for your eyes only.
Viewers will be able to get some information about the system Netdata is running. This information is everything the dashboard provides. The dashboard includes a list of the services each system runs (the legends of the charts under the Systemd Services
section), the applications running (the legends of the charts under the Applications
section), the disks of the system and their names, the user accounts of the system that are running processes (the Users
and User Groups
section of the dashboard), the network interfaces and their names (not the IPs) and detailed information about the performance of the system and its applications.
This information is not sensitive (meaning that it is not your business data), but it is important for possible attackers. It will give them clues on what to check, what to try and in the case of DDoS against your applications, they will know if they are doing it right or not.
Also, viewers could use Netdata itself to stress your servers. Although the Netdata daemon runs unprivileged, with the minimum process priority (scheduling priority idle
- lower than nice 19) and adjusts its OutOfMemory (OOM) score to 1000 (so that it will be first to be killed by the kernel if the system starves for memory), some pressure can be applied on your systems if someone attempts a DDoS against Netdata.
Netdata is a distributed application. Most likely you will have many installations of it. Since it is distributed and you are expected to jump from server to server, there is very little usability to add authentication local on each Netdata.
Until we add a distributed authentication method to Netdata, you have the following options:
If your organisation has a private administration and management LAN, you can bind Netdata on this network interface on all your servers. This is done in Netdata.conf
with these settings:
[web]
bind to = 10.1.1.1:19999 localhost:19999
You can bind Netdata to multiple IPs and ports. If you use hostnames, Netdata will resolve them and use all the IPs (in the above example localhost
usually resolves to both 127.0.0.1
and ::1
).
This is the best and the suggested way to protect Netdata. Your systems should have a private administration and management LAN, so that all management tasks are performed without any possibility of them being exposed on the internet.
For cloud based installations, if your cloud provider does not provide such a private LAN (or if you use multiple providers), you can create a virtual management and administration LAN with tools like tincd
or gvpe
. These tools create a mesh VPN allowing all servers to communicate securely and privately. Your administration stations join this mesh VPN to get access to management and administration tasks on all your cloud servers.
For gvpe
we have developed a simple provisioning tool you may find handy (it includes statically compiled gvpe
binaries for Linux and FreeBSD, and also a script to compile gvpe
on your Mac). We use this to create a management and administration LAN for all Netdata demo sites (spread all over the internet using multiple hosting providers).
In Netdata v1.9+ there is also access list support, like this:
[web]
bind to = *
allow connections from = localhost 10.* 192.168.*
Use one nginx (or one apache) server to provide authentication in front of all your Netdata servers. So, you will be accessing all your Netdata with URLs like http://nginx.host/netdata/{NETDATA_HOSTNAME}/
and authentication will be shared among all of them (you will sign-in once for all your servers). Check this wiki page for more information on configuring nginx for such a setup.
To use this method, you should firewall protect all your Netdata servers, so that only the nginx IP will allowed to directly access Netdata. To do this, run this on each of your servers (or use your firewall manager):
NGINX_IP="1.2.3.4"
iptables -t filter -I INPUT -p tcp --dport 19999 \! -s ${NGINX_IP} -m conntrack --ctstate NEW -j DROP
commands to allow direct access to Netdata from an nginx proxy
The above will prevent anyone except your nginx server to access a Netdata dashboard running on the host.
For Netdata v1.9+ you can also use netdata.conf
:
[web]
allow connections from = localhost 1.2.3.4
Of course you can add more IPs.
For Netdata prior to v1.9, if you want to allow multiple IPs, use this:
# space separated list of IPs to allow access Netdata
NETDATA_ALLOWED="1.2.3.4 5.6.7.8 9.10.11.12"
NETDATA_PORT=19999
# create a new filtering chain || or empty an existing one named netdata
iptables -t filter -N netdata 2>/dev/null || iptables -t filter -F netdata
for x in ${NETDATA_ALLOWED}
do
# allow this IP
iptables -t filter -A netdata -s ${x} -j ACCEPT
done
# drop all other IPs
iptables -t filter -A netdata -j DROP
# delete the input chain hook (if it exists)
iptables -t filter -D INPUT -p tcp --dport ${NETDATA_PORT} -m conntrack --ctstate NEW -j netdata 2>/dev/null
# add the input chain hook (again)
# to send all new netdata connections to our filtering chain
iptables -t filter -I INPUT -p tcp --dport ${NETDATA_PORT} -m conntrack --ctstate NEW -j netdata
script to allow access to Netdata only from a number of hosts
You can run the above any number of times. Each time it runs it refreshes the list of allowed hosts.
Of course, there are many more methods you could use to protect Netdata:
-
bind Netdata to localhost and use
ssh -L 19998:127.0.0.1:19999 remote.netdata.ip
to forward connections of local port 19998 to remote port 19999. This way you can ssh to a Netdata server and then usehttp://127.0.0.1:19998/
on your computer to access the remote Netdata dashboard. -
If you are always under a static IP, you can use the script given above to allow direct access to your Netdata servers without authentication, from all your static IPs.
-
install all your Netdata in headless data collector mode, forwarding all metrics in real-time to a master Netdata server, which will be protected with authentication using an nginx server running locally at the master Netdata server. This requires more resources (you will need a bigger master Netdata server), but does not require any firewall changes, since all the slave Netdata servers will not be listening for incoming connections.
The default configuration uses a public registry under registry.my-netdata.io (more information about the registry here: mynetdata-menu-item ). Please be aware that if you use that public registry, you submit the following information to a third party server:
- The url where you open the web-ui in the browser (via http request referer)
- The hostnames of the Netdata servers
If sending this information to the central Netdata registry violates your security policies, you can configure Netdat to run your own registry.
path | owner | permissions | netdata | comments |
---|---|---|---|---|
/etc/netdata |
user root group netdata |
dirs 0755 files 0640 |
reads | netdata config files may contain sensitive information, so group netdata is allowed to read them. |
/usr/libexec/netdata |
user root group root |
executable by anyone dirs 0755 files 0644 or 0755 |
executes | netdata plugins permissions depend on the file - not all of them should have the executable flag. there are a few plugins that run with escalated privileges (Linux capabilities or setuid ) - these plugins should be executable only by group netdata . |
/usr/share/netdata |
user root group netdata |
readable by anyone dirs 0755 files 0644 |
reads and sends over the network | Netdata web static files these files are sent over the network to anyone that has access to the netdata web server. Netdata checks the ownership of these files (using settings at the [web] section of netdata.conf ) and refuses to serve them if they are not properly owned. Symbolic links are not supported. Netdata also refuses to serve URLs with .. in their name. |
/var/cache/netdata |
user netdata group netdata |
dirs 0750 files 0660 |
reads, writes, creates, deletes | Netdata ephemeral database files Netdata stores its ephemeral real-time database here. |
/var/lib/netdata |
user netdata group netdata |
dirs 0750 files 0660 |
reads, writes, creates, deletes | Netdata permanent database files Netdata stores here the registry data, health alarm log db, etc. |
/var/log/netdata |
user netdata group root |
dirs 0755 files 0644 |
writes, creates | Netdata log files all the Netdata applications, logs their errors or other informational messages to files in this directory. These files should be log rotated. |