Everest v0.10.0¶

Changelog¶

✨ New features and improvements

• Fast mounts on large repositories. Ranges and metaranges are now fetched lazily from the local cache on first access instead of being downloaded up front. Mounts complete in seconds regardless of repository size or network latency, and no longer hit the client's 30-second timeout.
• Apply --log-level, --log-format, and --log-output to every command. The logging flags are now global, so diff, commit, and the other subcommands honor them the same way mount does.
• Richer request logging via the user-agent string. HTTP requests now identify themselves with the Everest version, host OS, and mount protocol (e.g. everest/1.2.3 linux/6.17.0-23-generic FUSE), making server-side logs far easier to trace.

🐛 Bug fixes

• NFS mounts survive handler panics. A panic in any NFS handler or filesystem operation is now recovered and logged with its stack trace instead of crashing the mount-server process.
• Clear error for unsupported mount sources. Mounting a non-lakefs:// source (such as local:// or file://) previously crashed the mount server with a nil-pointer panic. The mount command now rejects these sources up front with a clear "use lakefs://" message.
• Idle-read timeout for lakeFS HTTP clients. Hung connections no longer stall mounts indefinitely — an idle-read timeout closes stuck reads so the mount can recover.

Assets¶