Advanced Gravwell Configuration#
This document describes some more advanced configuration options for Gravwell installations, including information on configuring storage wells, data ageout, and multi-node clusters.
Gravwell is optionally a distributed system, with multiple indexers comprising a Gravwell cluster. The default installation will install both the webserver and indexer on the same machine, but with an appropriate license configuration, running a cluster is just as simple as running a single instance.
Configuration Files & Overlay Directories#
Almost all Gravwell components provide both a monolithic config file (e.g. /opt/gravwell/etc/gravwell.conf
, /opt/gravwell/etc/simple_relay.conf
) and a config overlay directory (e.g. /opt/gravwell/etc/gravwell.conf.d/
) into which you can drop snippets of additional configuration.
Any file in the config overlay directory ending in .conf
will be merged into the base config loaded from the config file, with files processed in alphabetical order. This follows the standard convention used in Linux systems, as in /etc/sudoers
and /etc/sudoers.d
.
For instance, rather than adding a new well by editing /opt/gravwell/etc/gravwell.conf
directly, you could create a file named /opt/gravwell/etc/gravwell.conf.d/syslog-well.conf
and include only the well configuration in it:
[Storage-Well "syslog"]
Location=/opt/gravwell/storage/syslog/
Tags=syslog
Similarly, one can drop Simple Relay Listener definitions into files in /opt/gravwell/etc/simple_relay.conf.d/
and so on.
General Configuration#
The core configuration file is designed to be shared by both the webserver and indexer, and is located by default at /opt/gravwell/etc/gravwell.conf
For a detailed listing of configuration options see this page.
The most important items in the configuration file are the Ingest-Auth
, Control-Auth
, and Search-Agent-Auth
configuration parameters. The Control-Auth
parameter is the shared secret that the webserver and indexers use to authenticate each other. If an attacker can communicate with your indexers and has the Control-Auth
token, he has total access to the data they store. The Ingest-Auth
token is used to validate ingesters, and restricts the ability to create tags and push data into Gravwell. Gravwell prides itself on speed, which means an attacker with access to your Ingest-Auth
token can push a tremendous amount of data into Gravwell in a very short amount of time. The Search-Agent-Auth
token allows Gravwell’s Search Agent utility to automatically connect to the webserver and issue searches on the behalf of users. These tokens are important and you should protect them carefully.
Attention
In clustered Gravwell installations, it is essential that all nodes are configured with the same Ingest-Auth
and Control-Auth
values to enable proper intercommunication.
Webserver Configuration#
The webserver acts as the focusing point for all searches, and provides an interactive interface into Gravwell. While the webserver does not require significant storage, it can benefit from small pools of very fast storage so that even when a search hands back large amounts of data, users can quickly navigate their results. The webserver also participates in the search pipeline and often performs some of the filtering, metadata extraction, and rendering of data. When deploying a webserver, we recommend a reasonably sized solid state disk (NVME if possible), a memory pool of 16GB of RAM or more, and at least 4 physical cores. Gravwell is built to be extremely concurrent, so more CPU cores and additional memory can yield significant performance benefits. An Intel E5 or AMD Epyc chip with 32GB of memory or more is a good choice, and more is always better.
Two configuration options tell the webserver how to communicate with indexers. The Remote-Indexers
option specifies the IPs or hostnames of the indexers, and the Control-Auth
option gives a shared key used by the webserver to authenticate to the indexers. A webserver connecting to three indexers might contain the following in its gravwell.conf
:
Control-Auth=MySuperSecureControlToken
Remote-Indexers=net:10.0.1.1:9404
Remote-Indexers=net:10.0.1.2:9404
Remote-Indexers=net:10.0.1.3:9404
Note
The indexers listed above are listening for control connections on port 9404, the default. This port is set by the Control-Port
option in the indexer’s gravwell.conf file.
Webserver TLS#
By default, Gravwell does not generate TLS certificates. For instructions on setting up properly-signed TLS certificates or self-signed certificates on the webserver, refer to the TLS/HTTP instructions.
Webserver Configuration Pitfalls#
Missing or misconfigured Remote-Indexers
Missing or mismatched Control-Auth tokens
Mismatched licenses on webserver and backend
Both the webserver and indexer must have compatible licenses
Poor network connectivity between the webserver and indexers
High latency, low bandwidth, or misconfigured MTU sizes.
Firewalls blocking access to indexer or webserver ports
The default is 9404
Indexer Configuration#
Indexers are the storage centers of Gravwell and are responsible for storing, retrieving, and processing data. Indexers perform the first heavy lifting when executing a query, first finding the data then pushing it into the search pipeline. The search pipeline will perform as much work as possible in parallel on the indexers for efficiency. Indexers benefit from high-speed low-latency storage and as much RAM as possible. Gravwell can take advantage of filesystem caches, which means that as you are running multiple queries over the same data it won’t even have to go to the disks. We have seen Gravwell operate at over 5GB/s per node on well-cached data. The more memory, the more data can be cached. When searching over large pools that exceed the memory capacity of even the largest machines, high speed RAID arrays can help increase throughput.
We recommend indexers have at least 32GB of memory with 8 CPU cores. If possible, Gravwell also recommends a very high speed NVME solid state disk that can act as a hot well, holding just a few days of of the most recent data and aging out to the slower spinning disk pools. The hot well enables very fast access to the most recent data, while enabling Gravwell to organize and consolidate older data so that he can be searched as efficiently as possible.
There are a few key configuration options in an indexer’s gravwell.conf which affect its general behavior:
Control-Port
sets the port on which the indexer will listen for incoming connections from a webserver. Default 9404.Control-Auth
sets a shared secret which webservers use to authenticate. Defaults to a randomly-generated string.Ingest-Port
andTLS-Ingest-Port
specify which ports to listen on for unencrypted and encrypted ingest traffic, respectively.Ingest-Auth
sets a shared secret used by data ingesters to authenticate to the indexer. Defaults to a randomly-generated string.
Indexers store their data in wells. Each well stores some number of tags. If a well contains 100GB of data tagged “pcap” and 10MB of data tagged “syslog”, searching for syslog data means the indexer also has to read the pcap data from the disk, slowing down the search. For this reason we strongly suggest creating separate wells for tags you anticipate will contain a lot of data. See the ‘Tags and Wells’ section for more information.
Data Ageout#
Gravwell supports an ageout system whereby data management policies can be applied to individual wells. The ageout policies control data retention, storage well utilization, and compression. For more information about configuration data ageout see the Data Ageout. section
Well Replication#
A Gravwell cluster with multiple indexer nodes can be configured so that nodes replicate their data to one another in case of disk failure or accidental deletion. See the replication documentation for information on configuring replication.
Query Acceleration#
Gravwell supports the notion of “accelerators” for individual wells, which allow you apply parsers to data at ingest to generate optimization blocks. Accelerators are just as flexible as query modules and are transparently engaged when performing queries. Accelerators are extremely useful for needle-in-haystack style queries, where you need to zero in on data that has specific field values very quickly. See the Accelerators section for more information and configuration techniques.
Password Complexity#
Gravwell supports the option to enforce password complexity on users when not in single sign on mode. Enabling password complexity requirements is performed by adding the following structure to the gravwell.conf
file:
[Password-Controls]
Min-Length=<integer>
Require-Uppercase=<bool>
Require-Lowercase=<bool>
Require-Number=<bool>
Require-Special=<bool>
The default Gravwell deployment does not enforce any rules on password complexity, and because Gravwell uses a secure bcrypt password hashing system, we have no way to enforce these rules after the fact. Once you enable password complexity requirements all future password changes will be required to abide by the requirements.
Here is an example configuration block that requires complex passwords that are at least 10 characters in length:
[Password-Controls]
Min-Length=10
Require-Uppercase=true
Require-Lowercase=true
Require-Number=true
Require-Special=true
Note that Gravwell fully supports UTF-8 character sets and that many languages do not have the concept of case. So while the password パスワードを推測することはできません!#$@42
may look very complex, it doesn’t meet the requirements above due to the lack of upper and lower case values.
Version Compatibility#
Certain versions of the indexer and webserver are only compatible with specific versions of other indexers and webservers. The table below details version compatibility restrictions. Mismatched webservers and indexers will not run.
API Version |
Indexer/Webserver Version Compatibility |
---|---|
1 |
Any version between 1.0-3.3 |
2 |
4.0 and 4.1 |
3 |
4.2 |
Ingesters are always backwards compatible with older versions of indexers as they negotiate the ingest protocol version when they connect. However, some new features may be disabled if there is a significant version mismatch. We recommend using the ingester version that matches your indexer version.
Shell Installer Options#
The Gravwell shell installer supports several flags to make automated installation or deployment easier. The following flags are supported:
Flag |
Description |
---|---|
|
Display installer help menu |
|
Installer will not generate self-signed certificates |
|
Assume all defaults and automatically accept EULA |
|
Do not generate random ingest and control secrets |
|
Do not install Gravwell indexer component |
|
Do not install the Gravwell webserver component |
|
Do not install the Gravwell search agent (typically used with |
|
Do not start the components after installation |
|
Do not install the automated debug report component |
|
Use a specific config file |
Common use-cases for advanced installation requirements#
If you are deploying Gravwell to a cluster with multiple indexers, you would not want to install the webserver component on your indexer nodes.
If you are using an automated deployment tool you don’t want the installer stopping and asking a questions.
If you already have your list of indexers with ingest and control shared secrets, specifying a configuration file at install time can greatly speed up the process.
For example, to install the indexer component without installing the webserver or randomizing passwords, run:
root@gravserver# bash gravwell_installer.sh --no-questions --no-random-passwords --no-webserver
If you choose to randomize passwords, you will need to go back through your indexers and webserver and ensure the Control-Auth
parameter in the gravwell.conf
file matches for the webserver and each indexer. You’ll also want to set the same Ingest-Auth
value on all the indexers.