Table of Contents
- Overview
- Installation and Usage
- Checks
- Dependencies
- Comparison to Other Software
- Contributing and Getting Help
- License
This readme documents the current (development) version of distributive.
Distributive is a tool for running distributed health checks in datacenters. It was designed with Consul in mind, but is stack agnostic. It is simple to configure (with JSON checklists) and easy to deploy and run. It has no dependencies, and can be shipped as a speedy 1.3MB (yes, megabytes!) binary.
Usually, some external server will ask the host to execute this program, reading a checklist from a JSON file, and will record this program's exit code and standard out. Distributive's output includes information about which checks in a checklist failed, and how so.
The exit code meanings are defined as Consul, Kubernetes, Sensu, and Nagios recognize them.
- Exit code 0 - Checklist is passing
- Exit code 1 - Checklist is warning
- Any other code - Checklist is failing
As of right now, only exit codes 0 and 1 are used, even if a checklist fails.
To install the development version (potentially unstable):
- Clone this repo:
git clone https://github.com/CiscoCloud/distributive - Get Godep godep.
- (Optional) Test with
drone exec - Build a binary with
go build . - Follow the "Usage" instructions below
We also provide premade RPM packages on Bintray for versioned releases. You can view the RPM source and build RPM snapshots at distributive-rpm.
Distributive currently only supports Linux.
The default behavior is to run any checks specified via --file, --url,
--stdin, or --directory options, or all checks in /etc/distributive.d/ if no
location is specified.
$ distributive --help
[...]
VERSION:
v0.2.4
GLOBAL OPTIONS:
--verbosity info | debug | fatal | error | panic | warn
--file, -f Read a checklist from a file
--url, -u Read a checklist from a URL
--directory, -d Read all of the checklists in this directory
--stdin, -s Read data piped from stdin as a checklist
--no-cache Don't use a cached version of a remote check, fetch it.
--help, -h show help
--version, -v print the version
Examples:
$ /path/to/distributive --verbosity="warn" -f ./samples/filesystem.json
$ distributive --f="/etc/distributive/samples/network.json" --verbosity=debug
$ ./distributive -u "http://pastebin.com/raw.php?i=5c1BAxcX"
$ /distributive --verbosity="info"
$ /path/to/distributive -d "/etc/distributive.d/" # same as default behavior
$ cat samples/filesystem.json | ./distributive -d "" -s=true --verbosity=fatal
Distributive attempts to be as framework-agnostic as possible. It is known to work well with Consul, Kubernetes, Sensu, and Nagios, which have similar design in how they detect passing and failing checks. There is documentation on how to use Distributive with Consul on our Github wiki.
For the impatient, examples of every single implemented check are available in
the samples/ directory, sorted by category. There is extensive documentation
for each check available on our Github wiki.
If you'd like to see how Distributive is used in production environments, take a look at the RPM source, which includes checks used in Mantl.
Distributive itself has no dependencies; it is compiled as a statically linked standalone Go binary. Some checks, however, rely on output from specific packages. These dependencies are outlined for each check on our Github wiki.
Distributive was created with the idea of pushing responsibility to the nodes, It was also designed around the idea of constantly changing infrastructure, with servers being added and destroyed constantly, changing IP addresses, and even changing roles. Integration with Consul provides even greater flexibility.
Serverspec runs on a single control server, and requires each check to be in a directory matching the hostname of the machine to run it on. Distributive was designed for dynamic systems with changing IPs, which can report into Consul, Sensu, or another framework as soon as they are ready, and require little or no centralized configuration. Additionally, Distributive attempts to rely as little as possible on external tools/commands, using mostly just the Go standard library.
Nagios is an end-to-end monitoring, security, and notification framework. It provides many services not included in Distributive, and solves a very different problem. Distributive is simple, lightweight, and easy to configure, and doesn't provide its own scheduling, dashboard, etc. It is designed to be used within frameworks such as Sensu and Consul. Luckily, Distributive conforms to [Nagios exit code specifications] nagios, and can be used just like any other plugin. Its advantage over other plugins is that it is small, fast, and has no dependencies.
Thank you for your interest in contributing! To get started, please check out our wiki.
All comments, questions, and contributions are always welcome. We strive to provide expedient and detailed support for anyone using our software. Please submit any requests via our Github Issues Page, where someone will see it and get to work promptly.
Copyright © 2015 Cisco Systems, Inc.
Licensed under the Apache License, Version 2.0 (the "License").
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.