Settings are grouped by the section they belong to in the YAML file. Most
also have a command-line flag, listed by --help. See
Configuration for how the file, environment
and flags combine.
general
| Setting |
Default |
Meaning |
default_cloud |
|
Provider for bucket names written without a scheme: aws, gcp, azure, oracle, minio, cloudflare, coreweave. Defaults to the first bucket cloud provider if not set. |
workers |
10 |
Worker threads per bucket |
cloud
| Setting |
Default |
Meaning |
pubsub_enabled |
true |
Default for change notifications; a bucket’s own pubsub/nopubsub wins |
write_probe |
true |
Check at startup whether the credentials can write, and serve the bucket read-only if not |
strict_bucket_access |
false |
Fail startup, instead of serving read-only, when a bucket can’t be written |
event_log.enabled |
false |
Turn on the event log |
event_log.retention_minutes |
60 |
How long event-log entries are kept |
<provider>.credentials |
|
Path to the provider’s credentials file |
<provider>.project |
|
Project, account or organization the buckets belong to |
<provider>.buckets |
|
List of buckets: see Buckets |
<provider>.all_buckets |
false |
Serve every bucket the credentials can list |
disk and cache
| Setting |
Default |
Meaning |
disk.persistent_prefix |
/data |
Where persistent state lives |
disk.ephemeral_prefix |
/data |
Where rebuildable state (cache) lives |
cache.enable_compression |
true |
Compress data stored in read cache |
cache.stop_write_percent |
95 |
Refuse new writes above this disk use |
cleaner.target_percent |
90 |
Read cache disk cleaner’s target usage |
file
| Setting |
Default |
Meaning |
write_session_idle_seconds |
30 |
How long an NFSv3 write session may sit idle before BlixtFS closes and uploads it. NFSv3 clients never send a close of their own. |
write_session_commit_idle_seconds |
5 |
The same, for a session the client has already committed. |
write_session_abandoned_seconds |
0 |
Close any write session idle this long, on every protocol. 0 turns it off. |
indexing
| Setting |
Default |
Meaning |
workers |
32 |
Parallel listing workers |
lazy |
false |
Make every bucket lazy |
fsck_sleep_seconds |
3600 |
Interval between periodic consistency checks |
fix_database_issues |
true |
Allow checks to repair the index |
fix_grace_minutes |
60 |
Leave recently changed files alone |
fix_issues |
false |
Allow checks to write to or delete from the bucket |
user_fsck |
true |
Let users request checks through bfs.fsck |
manual_fsck_cooldown_seconds |
300 |
Minimum time between requested checks of one path |
filesystem
| Setting |
Default |
Meaning |
default_uid |
the server’s user |
Owner of files with no recorded owner |
default_gid |
the server’s group |
Group of files with no recorded group |
nfs, smb, sftp
See Protocols.
monitoring
| Setting |
Default |
Meaning |
allowed_cidrs |
|
Networks allowed to reach the metrics endpoints |
local
Settings for one particular process.
| Setting |
Default |
Meaning |
hostname |
localhost |
Hostname. |