[feature]: Add KEY_PREFIX to Redis cache and make clear_cache optional on startup

[cxhello]

Summary

Plane's Docker entrypoint (docker-entrypoint-api.sh) unconditionally runs python manage.py clear_cache on every container startup. This command calls Django's cache.clear(), which in django-redis translates to a Redis FLUSHDB — wiping all keys in the entire Redis database, not just Plane's own cache entries.

When REDIS_URL points to a shared Redis instance (common in organizations that deploy Plane into existing infrastructure), this silently destroys data belonging to other services on the same database.

Why should this be worked on?

In a real production incident, Plane was deployed into a Kubernetes cluster where REDIS_URL was misconfigured to point at a shared Redis db0 used by other services. Every container restart (6 rounds of iterative deployment) triggered FLUSHDB, wiping all keys — including task queues, result caches, and runtime state of other applications — causing cascading failures across all dependent services.

Proposed improvements:

1. Make clear_cache skippable: Add an environment variable (e.g. PLANE_SKIP_CACHE_CLEAR=true) so operators can disable the startup cache flush when sharing a Redis instance.
2. Add KEY_PREFIX to Redis cache config: Configure a prefix like plane: in Django's CACHES setting, so even if clear_cache runs, it only affects Plane's own keys. This requires changing the clear_cache management command to use cache.delete_pattern("plane:*") instead of cache.clear(), since django-redis's clear() always calls FLUSHDB regardless of KEY_PREFIX (upstream issue).
3. Document the risk: Note in the self-hosting docs that REDIS_URL should point to a dedicated Redis database, and that startup will flush it.

Reference: django-redis clear() implementation calls client.flushdb() unconditionally. Django core has the same open issue: #35039.

[pradeep-sgh]

Hi @cxhello! Thanks for highlighting this — running FLUSHDB on startup poses a serious risk when sharing Redis instances.

I'd like to work on this issue. My proposed plan:

1. In docker-entrypoint-api.sh (and local entrypoints), wrap clear_cache with a check for PLANE_SKIP_CACHE_CLEAR (defaulting to false to maintain backward compatibility).
2. Add KEY_PREFIX support to CACHES in plane/settings/common.py (e.g., defaulting to "plane" or configurable via REDIS_KEY_PREFIX).
3. Update the clear_cache management command to utilize cache.delete_pattern() scoped to the prefix rather than executing an unconditional cache.clear() / FLUSHDB.

Could you please assign this to me, or let me know if you'd like to see any adjustments to this approach? Thanks!