Skip to main content

Hard Fork Upgrades

One of the significant limitations of the normal upgrade procedure via governance is that it requires waiting for the entire duration of the voting period. This duration makes it unsuitable for automated upgrades that involve patches for security vulnerabilities or other critical components.

A faster alternative to using governance is to create a Hard Fork procedure. This procedure automatically applies the changes from an upgrade plan, allowing them to be executed at a given block height without the need of having to create a governance proposal.

The high-level strategy for coordinating an upgrade is as follows:

  1. The vulnerability is fixed on a private branch that contains breaking changes.
  2. A new patch release (e.g. v8.0.0 -> v8.0.1) needs to be created that contains a hard fork logic and performs an upgrade to the next breaking version (e.g. v9.0.0) at a predefined block height.
  3. Validators upgrade their nodes to the patch release (e.g. v8.0.1). In order to perform the hard fork successfully, it’s important that enough validators upgrade to the patch release so that they make up at least 2/3 of the total validator voting power.
  4. One hour before the upgrade time (corresponding to the upgrade block height), the new major release (e.g. v9.0.0) including the vulnerability fix is published.

Important: The release needs to be created with 1hr anticipation because the release binaries take ~30min to be created and validators need a buffer time to download them and update their cosmovisor settings.