fleet.example.toml

# uptime-bench fleet configuration
#
# Copy to fleet.toml and edit for your environment.
# fleet.toml is NOT committed — it contains hostnames, IPs, and credentials.
# This example file IS committed and serves as the reference format.

# ---------------------------------------------------------------------------
# Control plane
# ---------------------------------------------------------------------------

[control]
# Timeout for control plane requests to any fleet member.
timeout = "10s"

# Shared token for authenticating harness requests to fleet member control APIs.
# All fleet members must be configured with this same token.
# In production, point this at a file with restricted permissions (0600).
# Locally, set CONTROL_TOKEN in .env instead.
auth_token_file = "/etc/uptime-bench/control-token"

# ---------------------------------------------------------------------------
# Adapter call budgets
# ---------------------------------------------------------------------------
# Per-adapter API call limits. The harness aborts a run with
# resolution_reason = "budget_exceeded" if any adapter reaches its limit.
# If an adapter is not listed here, the adapter's own default (MaxCallsPerRun
# in Capabilities) applies. Set to 0 for unlimited (self-hosted services).
#
# How to estimate: (run_duration + grace_period) / check_frequency * 2
# gives a rough upper bound for Retrieve polling calls per run.

[adapters.jetmon-v1]
max_calls_per_run = 0          # self-hosted, no API cost

[adapters.uptimerobot]
max_calls_per_run = 100

[adapters.pingdom]
max_calls_per_run = 100

[adapters.datadog-synthetics]
max_calls_per_run = 200

[adapters.better-uptime]
max_calls_per_run = 100

# ---------------------------------------------------------------------------
# Nameservers
# ---------------------------------------------------------------------------
# The fleet's authoritative DNS nameservers. At least two entries are required
# to use dns_ns_unavailable scenarios. All nameservers in this list serve as
# NS records for every domain listed in their [[nameservers.domains]] array.
#
# The nameserver binary listens on dns_port for DNS traffic and on
# control_port for harness control commands. Both are required.

[[nameservers]]
id           = "ns-01"
address      = "203.0.113.10"
control_port = 9100
dns_port     = 53
domains      = ["bench.example.com", "probe.example.net"]
# Public hostnames the registrar's NS + glue records advertise for
# this member. The DNS server emits authoritative A records for these
# so the child zone matches the parent's glue (RFC 1034 §4.2.2);
# without them, in-bailiwick nameservers return NXDOMAIN for their
# own hostnames and DNSSEC / monitoring-service preflight checks may
# flag the zone as inconsistent.
hosts        = ["ns1.bench.example.com", "ns1.probe.example.net"]

[[nameservers]]
id           = "ns-02"
address      = "203.0.113.11"
control_port = 9100
dns_port     = 53
domains      = ["bench.example.com", "probe.example.net"]
hosts        = ["ns2.bench.example.com", "ns2.probe.example.net"]

# ---------------------------------------------------------------------------
# Target servers
# ---------------------------------------------------------------------------
# Each [[targets]] entry is one server running the uptime-bench target binary.
# A single server hosts multiple virtual hosts (sites). Sites are distinguished
# by Host header (HTTP) and SNI (TLS). Failure state is tracked per
# (host, path) pair — multiple failures can be active simultaneously
# across sites and paths on the same server.
#
# The target binary listens on ports 80 and 443 for monitor traffic and on
# control_port for harness control commands.

[[targets]]
id           = "target-01"
address      = "203.0.113.20"
control_port = 9000

  [[targets.sites]]
  id    = "bench-a"
  host  = "bench-a.bench.example.com"
  paths = ["/", "/api/health", "/shop"]

  [[targets.sites]]
  id    = "bench-b"
  host  = "bench-b.bench.example.com"
  paths = ["/", "/blog"]

  # Optional generated host range for capacity tests. The DNS server resolves
  # matching names without expanding each hostname into fleet.toml, and the
  # target HTTP server serves arbitrary Host headers from the same binary.
  # This example covers site-0000001.load.bench.example.com through
  # site-1000000.load.bench.example.com.
  [[targets.generated_sites]]
  id           = "load"
  host_pattern = "site-%07d.load.bench.example.com"
  start        = 1
  count        = 1000000
  paths        = ["/"]

[[targets]]
id           = "target-02"
address      = "203.0.113.21"
control_port = 9000

  [[targets.sites]]
  id    = "probe-a"
  host  = "probe-a.probe.example.net"
  paths = ["/", "/api"]

  [[targets.sites]]
  id    = "probe-b"
  host  = "probe-b.probe.example.net"
  paths = ["/"]

# ---------------------------------------------------------------------------
# Domains
# ---------------------------------------------------------------------------
# Domain-level configuration. Links each domain to its registered nameservers
# and sets the TTL for all records under the domain. Keep TTL low (≤30s) so
# failure injection and recovery propagate quickly to resolvers.
#
# nameservers must match [[nameservers]] ids defined above.
# nameserver_hosts is optional. Use it when the parent zone delegates to
# out-of-zone NS hostnames, or when one nameserver member has multiple public
# hostnames and a domain should publish only a specific pair.

[[domains]]
name        = "bench.example.com"
registrar   = "Namecheap"
nameservers = ["ns-01", "ns-02"]
nameserver_hosts = ["ns1.bench.example.com", "ns2.bench.example.com"]
ttl         = 30

[[domains]]
name        = "probe.example.net"
registrar   = "Namecheap"
nameservers = ["ns-01", "ns-02"]
nameserver_hosts = ["ns1.probe.example.net", "ns2.probe.example.net"]
ttl         = 30

# ---------------------------------------------------------------------------
# Certmint
# ---------------------------------------------------------------------------
# The certificate-library producer. Optional: a fleet without certmint can
# still run scenarios that don't need real-CA TLS expiration coverage,
# falling back to fleet-CA / generated certs.
#
# When configured, the harness reads this section and forwards the
# library URL to targets at provision time so each target's TLS listener
# polls certmint for new manifest entries. The same shared CONTROL_TOKEN
# used elsewhere in the fleet authenticates the polling.

[certmint]
id           = "certmint-01"
address      = "203.0.113.30"
library_port = 9200