Configuration reference

Dense table of every v0.1.0 option.

Version
Latest
v0.1.0 · latest 3 min read

See Configuration for grouped guidance. This page is the dense lookup.

Name Flag Env Type Default Required Valid Behavior
host --host/-H PLOMID_HOST addr 127.0.0.1 no IP/hostname bind; remote needs TLS or opt-in
port --port/-p PLOMID_PORT u16 5432 no 1–65535
data dir --data/-D PLOMID_DATA_DIR path ./data no dir created
username --username/-U PLOMID_USER string plomid no non-empty auth identity
password --password/-W PLOMID_PASSWORD string plomid server / secret CLI+start.sh yes in practice secret, non-default on remote never logged; the built-in plomid is refused on non-loopback binds
log level --log-level PLOMID_LOG_LEVEL/RUST_LOG enum info no trace/debug/info/warn/error
auth --auth PLOMID_AUTH enum scram no scram/password/md5/none none loopback only (trusts identity, still authorizes); md5 = deprecated, needs --allow-legacy-md5
legacy md5 --allow-legacy-md5 PLOMID_ALLOW_LEGACY_MD5 flag off no — explicit downgrade switch; startup warns, logins counted
tls --tls — flag off no — rustls handshake; needs cert+key (validated at startup)
tls cert/key --tls-cert/--tls-key PLOMID_TLS_CERT/KEY path "" with --tls PEM files empty/unparsable/expired/mismatched → exit 1
tls required --tls-required — flag off no — w/o --tls → error; plaintext refused with FATAL 28000
WAL bytes --checkpoint-wal-bytes PLOMID_CHECKPOINT_WAL_BYTES bytes 67108864 no u64, 0=off
segments --checkpoint-segments PLOMID_CHECKPOINT_SEGMENTS u64 16 no 0=off
interval --checkpoint-interval-secs PLOMID_CHECKPOINT_INTERVAL_SECS secs 900 no 0=off
remote plaintext --allow-remote-plaintext PLOMID_ALLOW_INSECURE_REMOTE flag off remote w/o TLS — explicit opt-in
no auto ckpt --no-auto-checkpoint — flag off no — disables policy
max connections --max-connections PLOMID_MAX_CONNECTIONS usize 100 no ≥1 semaphore permits; rejections fail closed, listener kept
statement timeout --statement-timeout PLOMID_STATEMENT_TIMEOUT duration 30s no 0=off, else ms/s/min/h stamped as the session default; per-session SET overrides
idle-in-txn timeout --idle-in-transaction-timeout PLOMID_IDLE_IN_TRANSACTION_TIMEOUT duration 60s no 0=off abandoned transactions release locks
COPY buffer --max-copy-buffer-bytes PLOMID_MAX_COPY_BUFFER_BYTES bytes 67108864 no ≥1 per-connection accumulation; past the cap the COPY aborts
prepared cap --max-prepared-statements PLOMID_MAX_PREPARED_STATEMENTS usize 1024 no ≥1 per connection; past the cap Parse fails
portal cap --max-portals PLOMID_MAX_PORTALS usize 1024 no ≥1 per connection; past the cap Bind fails
COPY dir --copy-dir PLOMID_COPY_DIR path <data-dir>/copy no dir file COPY paths are jailed here; symlinks / non-dirs fail closed
audit log --audit-log PLOMID_AUDIT_LOG path unset (disabled) no file append-only JSONL, fsynced with size rotation; open failure exits 1
profile --profile PLOMID_PROFILE enum auto no edge/laptop/server/auto hardware tier override; typos fail closed; edge tightens checkpoint thresholds

lock_timeout is accepted via SET for protocol compatibility but is not separately enforced: gate waits have no cheap hook, and the statement deadline bounds total time for queries that reach a poll point. Frontend message framing stays capped at 64 MiB. Spill is bounded per operation (1 GiB temp limit, not user-configurable); sort/aggregate spills past 250,000 rows fall back to disk automatically. The maintenance queue is a bounded channel (one queued-or-running item per table, full degrades to inline); workers and memory follow the hardware tier (cgroup-aware).

Engine: pool_capacity by tier (edge 256 / laptop 1024 / server 4096 pages), segment 8 MiB, WAL segment 8 MiB.

Was this page helpful?