Repository navigation
Conversation
The stateless KEDA scale-to-zero apps that can run multiple replicas now stay always-on at 2 (survives a pod/node failure) instead of scaling to zero: whoami (stateless) and fleetdm (app tier is stateless; state lives in MySQL/Redis). HTTPScaledObject replicas.min 0->2; initial replicas 1->2. The other KEDA apps genuinely can't run >1 and stay single-pod: - hubble-ui: keeps per-pod connection state (streaming sessions bounce); - headlamp: OIDC login state is in-memory per pod; - actual-budget: SQLite on an RWO PVC (single writer); - longhorn-ui: max-1 admin dashboard. These stay exempt from require-three-replicas because KEDA owns their replica count (the floor governs statically-declared replicas, not autoscaled ones). Comments updated to reflect HA-where-possible. Trade-off: this abandons scale-to-zero for whoami/fleetdm (always 2 pods). Revert per-app by setting the HTTPScaledObject min back to 0 if the idle cost isn't worth it. Validated: ksail workload validate passes for the changed files. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
Author
|
Closing — KEDA scale-to-zero apps should keep min 0 (they're already drain-safe). Dropped from the stack. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stacked on #1901. The KEDA scale-to-zero apps — made HA where the app actually supports multiple replicas.
Always-on HA (min 2)
HTTPScaledObjectreplicas.min: 0 → 2, initial replicas1 → 2. KEDA still owns the count; these stay exempt from the floor (the floor governs statically-declared replicas, not autoscaled ones).Genuinely single-pod (stay as-is, documented)
This abandons scale-to-zero for whoami/fleetdm (always 2 pods, ~idle cost). The maintainer deliberately built scale-to-zero for these; revert per-app by setting the
HTTPScaledObjectminback to0if the idle cost outweighs the availability.Validation:
ksail workload validatepasses for the changed files.🤖 Generated with Claude Code