hypercache/configs/hypercache-controller.toml at main · looelabs/hypercache · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
# HyperCache Central Cluster Controller Configuration
# Deployed on Head / Coordinator node running: ./hypercache-admin -c configs/hypercache-controller.toml controller

# ==============================================================================
# 1. CONTROLLER NETWORK BINDING
# ==============================================================================

# IP address to bind the Central Controller service on
# Set to "0.0.0.0" to listen on all interfaces, or bind to a specific private VPC IP (e.g. "10.0.0.10")
controller_bind_address = "0.0.0.0"

# Port for incoming worker node registrations, heartbeats, and admin API calls
controller_port = 9120

# Optional secondary admin port
controller_admin_port = 9121

# ==============================================================================
# 2. CLUSTER SECURITY & AUTHENTICATION
# ==============================================================================

# Shared secret Bearer token for cluster security
# When configured, worker nodes must provide this token to register, and remote CLI calls require it.
# Local connections from 127.0.0.1 on the controller host bypass authentication automatically.
# Can also be set via: export HYPERCACHE_AUTH_TOKEN="your-cluster-token"
# cluster_auth_token = "your-cluster-token"

# Zero-config local admin: loopback callers skip the token check (default true).
# Set to false whenever traffic can reach the controller through a local proxy (Kubernetes
# sidecars such as Istio/Envoy, nginx, kubectl port-forward): those connections also arrive
# from 127.0.0.1 and would otherwise bypass authentication.
allow_local_admin_bypass = true

# Nodes that have not sent a heartbeat for this many seconds are removed from the registry
# (they show as OFFLINE after 15 s and are never sent fleet operations).
node_gc_secs = 600

# ==============================================================================
# 3. P2P FLEET PRE-WARMING & CONSISTENT SHARDING
# ==============================================================================

# When enabled, fleet-wide preload pulls once from S3 onto a seed node and distributes
# chunks peer-to-peer across remaining worker nodes to eliminate S3 egress amplification.
p2p_preload_enabled = false