Releasing features with LaunchDarkly
Read time: 3 minutes
Last edited: Nov 20, 2024
Overview
This topic compares several options LaunchDarkly provides for releasing features to production safely and gradually.
Release options comparison
The following table compares functionality that LaunchDarkly provides for releasing features:
| Release option | Description | Example use case | Availability |
|---|---|---|---|
| Percentage rollouts | This option on a flag's targeting rule serves a given flag variation to a specified percentage of contexts. The percentage of customers receiving a variation does not change over time. To automatically increase the percentage, use progressive rollouts, described below. Affects one targeting rule in one environment. | Use percentage rollouts if you want to randomly allocate traffic by context kind and attribute, for example, if you want to test a new feature on a subset of end users in production. If you later change the percentage in the percentage rollout, the flag variation that any particular customer receives may change each time you change the percentage. To learn more, read Percentage rollout logic. | Percentage rollouts are available to all plans and work with all flag types. |
| Progressive rollouts | This option on a flag's targeting rule serves a given flag variation to a specified percentage of contexts, and gradually increases that percentage over a specified time. With this option, the percentage of customers receiving a variation automatically increases over time. Affects one targeting rule in one environment, over time. | Use progressive rollouts if you want to randomly allocate traffic by context kind, and automatically increase the amount of traffic to a specific flag variation over time. As the rollout progresses, the flag variation that any particular customer receives changes only once. | Progressive rollouts are available to all plans and work with all flag types. |
| Guarded rollouts | This option on a flag's targeting rule serves a given flag variation to a specified percentage of contexts, and gradually increases that percentage over a specified time. This option monitors the rollout for regressions based on metrics you select, and can automatically notify you and roll back if a regression is detected. Guarded rollouts are one of the rollout strategies available in each release pipeline phase. Affects one targeting rule in one environment, over time. | Use guarded rollouts if you want to randomly allocate traffic by context kind, and automatically increase the amount of traffic to a specific flag variation over time, while monitoring selected metrics. For example, this option can monitor latency and error rates specific to the traffic receiving the selected flag variation. LaunchDarkly notifies you or automatically reverts the rollout if regressions are detected. | Guarded rollouts is only available on a Guardian Edition plan. Guarded rollouts are available for all flag types. |
| Release pipelines | Release pipelines let you move flags through a series of phases, rolling out flags to selected environments and audiences following automated steps. Other release options work for a single flag targeting rule, whereas a release pipeline standardizes your release process across flags and environments. For each phase, you select the environment, audience, and rollout strategy, as well as whether approvals are required. Affects one targeting rule across multiple environments over time. | Use release pipelines in combination with guarded rollouts to standardize and automate the release process for your flags across multiple environments. | Release pipelines are only available on an Enterprise plan. Release pipelines are available for boolean flags only. |
Additional release management tools
In addition to release pipelines and several rollout options, LaunchDarkly provides tools for managing the steps of your release. Choose the tools that work best for your release and change management processes.
The topics in the Release management tools category describe these tools in detail: