Refinery is designed to sit within your infrastructure where all sources of Honeycomb events can reach it. A standard deployment will have a cluster of servers running Refinery accessible via a load balancer. Refinery instances must be able to communicate with each other to concentrate traces on single servers.
For the quickest way to get started using Refinery:
To begin, your Refinery cluster requires a minimum of:
linux/amd64 or linux/arm64 operating systemIn many cases, Refinery only needs one node. If experiencing a large volume of traffic, you may need to scale out to multiple nodes, and likely need a small Redis cluster to handle scaling.
We recommend increasing the amount of RAM and the number of cores after your initial set-up. Use our scaling and troubleshooting documentation to learn more.
Find our latest release of Refinery for your operating system and architecture on GitHub.
Use the command line script below to download the latest released .rpm asset for x86_64 from GitHub, install, and then run Refinery:
curl -L -O https://github.com/honeycombio/refinery/releases/download/latest/refinery-2.1.0.x86_64.rpm
rpm -ivh refinery-2.1.0.x86_64.rpm
systemctl start refinery.service
In the above example, we use systemctl to run the Refinery service.
Logs can be found at /var/log/journal/refinery.service.log.
The recommended way to run Refinery on Kubernetes is with its Helm chart.
Add the Honeycomb Helm repository.
helm repo add honeycomb https://honeycombio.github.io/helm-charts
Update the Honeycomb repository to ensure helm is aware of the latest versions.
helm repo update refinery honeycomb/refinery
Install the latest version of the Refinery Helm chart with default values.
helm install refinery honeycomb/refinery
To install a specific version or override the default values, refer to Refinery Helm chart documentation to learn more about configuration.
The Refinery executable has the following command line flags:
-c, --config=/etc/refinery/refinery.yaml)-r, --rules_config=/etc/refinery/rules.yaml)-v, --version-d, --debuglocalhost:6060 and :6069 by default).
Can be used with a debug service, which allows you to use pprof to visualize and analyze profiling data.-h, --help Ensure that your Refinery cluster(s) meet the minimum system requirements.
We recommend that your list of Refinery peers be configured through Redis (see Redis-based peer management). The Redis server can be small since it only maintains the list of peers. Should the Redis service become unavailable, Refinery instances will continue to use their last known peer list for inter-peer communication.
For services that are currently sending events directly to Honeycomb’s API, update these services’ Honeycomb API Host property to be the URL for your Refinery cluster’s load balancer to start sampling events with Refinery.
The default configuration at installation will allow you to run Refinery without any other changes. Be sure to run Refinery in Dry Run Mode to verify your configuration before dropping traffic.
To tune Refinery to your needs, Refinery’s configuration requires additional changes in two files:
config.yaml contains general configuration options, such as Network, Access Key, and Peer Management settings.
Read our documentation on all available configuration options.
rules.yaml contains sampling rules configuration.
Read our documentation on all supported sampling methods and their associated configuration.
As you modify, check out our scaling and troubleshooting documentation for recommendations and strategies.