fsdb is a lightweight and secure Unix daemon that provides basic key-value storage functionality over a UNIX domain socket. A local, atomic, multi-threaded, syslog-audited, filesystem-backed object store with consistent performance under load.

• ✅ Fast local IPC via UNIX domain sockets
• ✅ HTTP/1.1 POST support: speak to fsdb from any language or frontend via HTTP, using nginx proxy or curl!
• ✅ Extremely scalable 2-level folder sharding for 100M+ files
• ✅ Safe concurrent access via POSIX file locks
• ✅ Audit logging with timestamped entries
• ✅ Process-safe with flock-based PID file locking
• ✅ Signal handling for graceful shutdown
• ✅ Hardened with -fPIE, -D_FORTIFY_SOURCE=2, stack protector, and secure file sanitization
• ✅ Memory-safe: passes Valgrind and Clang analyzer with zero leaks
• ✅ Can run on tmpfs, xfs, or any Linux filesystem
• 🔒 Secure and concurrent: uses epoll, pthread, and safe file locking (fcntl)
• 🔌 Communication via Unix socket

🧪 Commands Supported

• CREATE <db>: initialize DB structure

• INSERT <db> <id> <data>: insert new record

• UPDATE <db> <id> <data>: update existing

• GET <db> <id>: fetch contents

• DELETE <db> <id>: remove entry

• CHECK <db> <id>: test existence

• TOUCH <db> <id>: create an empty file

• HTTP POST Parameters:

  • ACTION — one of INSERT, UPDATE, GET, DELETE, EXISTS, TOUCH, CREATE
  • db — database name
  • id — key
  • data — value (if needed by action)

🔐 Security & Hardening

Uses flock() and fcntl() locks Sanitizes all id and db inputs Resists TOCTOU vulnerabilities Can run under chroot or as an unprivileged user Validated with valgrind, clang --analyze, and sanitizers

🧱 Architecture

main() initializes socket, threads, and event loop Clients accepted via epoll, queued using a ring buffer Worker threads process requests from the queue Each database is a subdirectory; each key is a file

🧠 Recommended Filesystem Tuning

For best performance:

Use XFS with directory indexing and large inode allocation Mount with noatime, nodiratime, inode64 (XFS-specific) For high-performance volatility: mount tmpfs on specific DBs