# An auth token for admin access to Diom
#
# It's useful for bootstrapping a new Diom cluster, or using it in CI pipelines or other
# automated testing environments where you need a stable, well-known token for testing or
# scripted setup.
#
# If you're using it to bootstrap a new Diom cluster, it’s recommended that you only use
# it during bootstrapping, and remove this configuration once done.
# admin_token =

# How often to run background cleanup/garbage collection jobs
#
# Correctness should never be affected by this, just wasted memory/disk.
background_cleanup_interval_ms = 10000

# YAML bootstrap data. Takes precedence over `bootstrap_cfg_path`
# bootstrap_cfg =

# The path to a YAML bootstrap file
# bootstrap_cfg_path =

# Maximum time to wait for cluster initialization at startup
#
# If this is unset, we will wait indefinitely.
# bootstrap_max_wait_ms =

# The environment (dev, staging, or prod) that the server is running in.
environment = "dev"

# When fsyncing, should we use fsync(2) or fdatasync(2)
fsync_mode = "sync-data"

# The address to listen on
listen_address = "[::]:8624"

# The log format that all output will follow. Supported: default, json
log_format = "default"

# The log level to run the service with. Supported: info, debug, trace
log_level = "info"

# How to persist data to the actual underlying database
#
# This is similar to the `cluster.log_sync` options, but applies to the actual
# primary data as opposed to the log, and is applied at every batch commit from the
# underlying replication system.
sync_mode = "buffer"

# Configuration for the cluster/replication system
[cluster]

# Address that other nodes should use to communicate with this one.
#
# If not passed, we'll attempt to discover it at boot time.
# This cannot currently be changed after cluster initialization.
# advertised_address =

# Automatically initialize the cluster on bootup if we can't discover any
# peers and we don't have any existing state.
#
# If you initialize all peers at exactly the same time, this can potentially cause errors.
auto_initialize = true

# Timeout for new connections.
#
# If you want to be tolerant of dropped packets, this should be set to at least TO + ε,
# where TO is the initial TCP retransmission timer (typically either 1s or 3s,
# depending on your operating system).
connection_timeout_ms = 3100

# Timeout for discovery requests.
#
# This should be set to approximately 2X the RTT of your farthest-apart nodes.
discovery_request_timeout_ms = 10000

discovery_timeout_ms = 30000

# The minimum time to let an election run for.
#
# This should be set to at least 5x the RTT of your farthest-apart nodes
# and must not be less than `cluster_election_timeout_max`.
election_timeout_max_ms = 3500

# The minimum time to let an election run for.
#
# This should be set to at least 4x the RTT of your farthest-apart nodes,
# and must not be less than `heartbeat_interval_ms`.
election_timeout_min_ms = 1500

# How often to send heartbeats.
#
# This controls how fast lost leaders can be detected.
# Must not be less than `replication_request_timeout`.
heartbeat_interval_ms = 500

# The address to listen on for replication.
listen_address = "[::]:8625"

log_index_interval_ms = 600000

# Location to store logs. For high-throughput systems, this should be a separate volume.
#
# Defaults to a subdirectory under the persistent DB path if not passed.
# log_path =

# Automatically attempt to determine the log sync interval from observed fsync timings
log_sync_interval_auto = true

# Interval (in transactions) between fsyncing the commit log.
#
# This can be used to force transactions to fsync logs more often than the
# default `log_sync_interval_ms` timer. If `log_sync_mode` is set to "buffer", it's
# reasonable to set this value to `1` to flush to the OS buffer on every log.
#
# If this is set to 0, only the interval timer will be used
#
# If this is set to a value higher than 1 and the interval timer is long, then
# single-threaded clients (including bootstrap) will be extremely slow.
log_sync_interval_commits = 0

# Interval (in milliseconds) between fsyncing the commit log.
#
# If `log_sync_interval_auto` is set to true, this is just the initial estimate
# and will be auto-scaled
log_sync_interval_ms = 2

# Should a log sync actually trigger an fsync?
#
# If this is set to "buffer" and a node suffers a catastrophic failure where OS buffers
# are not written to disk, that node should be erased and re-snapshotted before being
# re-added to the cluster.
log_sync_mode = "sync"

# Human-facing name for this cluster.
#
# Only used in discovery and debugging.
name = "diom"

# How many commits behind must the current node be to be considered "lagging" and eligible for
# re-snapshotting?
#
# The ideal value here depends both on your data-set size and on your average write-rate. If
# your data is large, then setting this value too small can mean that a snapshot can never
# catch up because it'll take too long to replicate. Typically this should be around twice the
# number of commits that you generate in the time it takes to replicate a full snapshot.
replication_lag_threshold = 50000

# Timeout for replication requests.
#
# This should be set to approximately 2X the RTT of your farthest-apart nodes.
replication_request_timeout_ms = 5000

# Shared secret for intra-cluster communications
#
# This must be the same on all nodes.
# secret =

# Other nodes that we should attempt to join a cluster with at boot time.
seed_nodes = []

# The minimum time to let an election run for.
#
# This should be set to at least 5x the RTT of your farthest-apart nodes
# and must not be less than `cluster_election_timeout_max`.
send_snapshot_ms = 30000

# Trigger a background snapshot after this many milliseconds
snapshot_after_ms = 900000

# Trigger a background snapshot after this many writes
# snapshot_after_writes =

# Location to store snapshots.
#
# This volume must have at least as much space as the persistent DB path
# and ephemeral DB path combined. Defaults to a subdirectory under the
# persistent DB path if not passed.
# snapshot_path =

startup_discovery_delay_ms = 10

# Storage configuration for the ephemeral database, which is used by rate-limiting and some
# other modules, and can be placed on less-durable storage, such as local instance storage
[ephemeral_db]

# Amount of memory to reserve for the database layer's
# caches for this database type.
#
# Can be specified as a bare value of bytes (e.g., 1024000), a unit-ed amount
# (e.g., 1024MiB), or a percentage (e.g., 20%), which will be applied against
# the current cgroup limit if present and the total system memory otherwise
cache_size = "20%"

# Filename under the directory specified in `path`.
filename = "fjall_ephemeral"

# Directory in which this database is stored
path = "./db"

# Configuration for verifying JWT bearer tokens.
#
# When set, bearer tokens in JWT format are verified using this configuration.
# The JWT must contain a `role` claim (string) and may contain a `context` claim
# (object with string values) that is forwarded to internal diom handlers.
[jwt]

# JWT algorithm.
#
# Supported values are HS256, HS384, HS512, RS256, RS384, RS512, ES256, ES384, PS256, PS384, PS512.
# algorithm =

# Secret for JWT algorithm HS256, HS384 or HS512
# secret =

# Public key PEM for JWT algorithm RS256, RS384, RS512, ES256, ES384, PS256, PS384 or PS512
# public_key_pem =

# Expected `aud` values. When set, the token must contain one of these
# values in its `aud` claim. When absent, `aud` is not validated.
# audience =

# Expected `iss` values. When set, the token's `iss` claim must match one
# of these values. When absent, `iss` is not validated.
# issuer =

[opentelemetry]

# The OpenTelemetry address to send events to if given.
#
# Currently only GRPC exports are supported.
# address =

# The OpenTelemetry address to send metrics to if given.
#
# If not specified, the server will attempt to fall back
# to `opentelemetry_address`.
# metrics_address =

metrics_period_ms = 10000

# OpenTelemetry metrics protocol
#
# By default, metrics are sent via GRPC. Some metrics destinations, most
# notably Prometheus, only support receiving metrics via HTTP.
metrics_protocol = "grpc"

# The ratio at which to sample spans when sending to OpenTelemetry.
#
# When not given it defaults to always sending.
# If the OpenTelemetry address is not set, this will do nothing.
# sample_ratio =

# The service name to use for OpenTelemetry. If not provided, it defaults to "diom".
service_name = "diom"

# Storage configuration for the persistent database, which is used by most modules and
# should be placed on durable storage
[persistent_db]

# Amount of memory to reserve for the database layer's
# caches for this database type.
#
# Can be specified as a bare value of bytes (e.g., 1024000), a unit-ed amount
# (e.g., 1024MiB), or a percentage (e.g., 20%), which will be applied against
# the current cgroup limit if present and the total system memory otherwise
cache_size = "20%"

# Filename under the directory specified in `path`.
filename = "fjall_persistent"

# Directory in which this database is stored
path = "./db"