You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Add PodDisruptionBudget support to OpenFGA Helm chart #328
The OpenFGA Helm chart currently does not expose PodDisruptionBudget (PDB) configuration for the main OpenFGA workload. This makes it harder to enforce minimum availability during voluntary disruptions such as node drains and cluster upgrades.
Problem
In production Kubernetes environments, PDB is commonly required to:
Prevent full disruption during maintenance operations
Align with platform reliability policies
Keep at least one replica available during evictions
Today, users cannot configure this directly in chart values for the OpenFGA deployment.
Proposal
Add first-class PDB support to the OpenFGA chart with configurable values.
Suggested values:
pdb:
create: falseminAvailable: 1maxUnavailable:
Expected behavior
Render a policy/v1 PodDisruptionBudget when pdb.create is true
Target the OpenFGA workload labels used by the deployment
Support either minAvailable or maxUnavailable
Validate that both are not set at the same time
Acceptance criteria
New PDB template added for OpenFGA chart
Values documented in chart values and README
Template renders only when enabled
Helm template output is valid on supported Kubernetes versions
Reacted by David Leonardo Gutiérrez, snavarron, Dolan Antenucci and Roland Wolf
#327 is complementary but should remain separate: it controls Deployment rollouts, while a PDB protects availability during voluntary evictions such as node drains and cluster upgrades.
For this issue, I suggest supporting both integer and percentage values for minAvailable/maxUnavailable, enforcing that exactly one is configured, and adding tests for disabled, each valid mode, and invalid dual configuration. Keeping this as a focused follow-up PR avoids delaying #327.
Activity
rwo-trackunit commentedon Aug 19, 2026
I've added a PR to handle deployment strategy, that could be expanded to handle this too. Assuming it gets approved.
Siddhant-K-code commentedon Aug 24, 2026
#327 is complementary but should remain separate: it controls Deployment rollouts, while a PDB protects availability during voluntary evictions such as node drains and cluster upgrades.
For this issue, I suggest supporting both integer and percentage values for
minAvailable/maxUnavailable, enforcing that exactly one is configured, and adding tests for disabled, each valid mode, and invalid dual configuration. Keeping this as a focused follow-up PR avoids delaying #327.