SeaweedFS is a distributed storage system for object storage (S3), file systems, Iceberg table buckets, and Lance vector buckets, designed to handle billions of files with O(1) disk access and effortless horizontal scaling.
Go blob-storagecloud-drivedistributed-file-systemdistributed-storageerasure-codingfusehadoop-hdfshdfskuberneteslancelancedbobject-storageposixs3s3-storageseaweedfstiered-file-systemvector-search
# Put this file in one of the locations, with descending priority
# ./admin.toml
# $HOME/.seaweedfs/admin.toml
# /etc/seaweedfs/admin.toml
# this file is read by "weed admin"
# Maintenance task settings declared here are applied every time the admin
# server starts, overriding values saved from the admin UI under
# <dataDir>/conf. Commented-out keys keep their saved or default values.
# Requires -dataDir.
# Each value can also be set via environment variable, e.g.
# export WEED_MAINTENANCE_VACUUM_GARBAGE_THRESHOLD=0.3
[s3]
# The address clients reach the S3 gateway at, including the scheme.
# The file browser uses it to offer copyable object URLs; without it,
# the copy action is hidden.
# public_endpoint = "https://s3.example.com"
[maintenance]
# toggle the entire maintenance system (task detection and execution)
# enabled = true
[maintenance.vacuum]
# enabled = true
# vacuum volumes with more deleted content than this ratio
# garbage_threshold = 0.3
# how often to scan for volumes needing vacuum
# scan_interval_seconds = 7200
# max_concurrent = 2
# only vacuum volumes older than this (rounded up to whole hours)
# min_volume_age_seconds = 86400
# max retry attempts for a failed vacuum job (admin runtime default)
# retry_limit = 1
# seconds to wait between retry attempts
# retry_backoff_seconds = 10
[maintenance.balance]
# enabled = true
# trigger balancing when volume distribution exceeds this ratio
# imbalance_threshold = 0.2
# scan_interval_seconds = 1800
# max_concurrent = 1
# minimum number of volume servers before balancing
# min_server_count = 2
# max retry attempts for a failed balance job
# retry_limit = 1
# seconds to wait between retry attempts
# retry_backoff_seconds = 15
[maintenance.erasure_coding]
# enabled = true
# only erasure code volumes this full
# fullness_ratio = 0.95
# only erasure code volumes not modified for this long
# quiet_for_seconds = 3600
# scan_interval_seconds = 3600
# max_concurrent = 1
# min_size_mb = 30
# only process volumes from these collections, comma separated, "*" and "?" wildcards
# and regex patterns allowed ("" = all collections)
# collection_filter = ""
# disk tags preferred for shard placement
# preferred_tags = ["fast", "ssd"]
# EC shard placement constraint, e.g. "020"; empty uses the master default replication
# replica_placement = ""
# fail planning when the placement constraints cannot be met instead of relaxing them
# strict_placement = false
# volume-slot deadband for shard placement: nodes whose free volume slots differ by at
# most this much tie and fewer live shards wins; applied times data shards (0 = strict)
# free_slots_deadband = 0
# shard copy concurrency. Leave unset for one in-flight copy per destination disk.
# 0 sends shards sequentially. A positive N is a global cap, still one per disk.
# max_shard_transfers = 4
# max retry attempts for a failed erasure coding job
# retry_limit = 1
# seconds to wait between retry attempts
# retry_backoff_seconds = 30