v1.1.3

juicedata/juicefsv1.1.3Jun 21, 2024by SandyXSD

AI Summary

This is the third patch release for JuiceFS v1.1, focusing on stability improvements and bug fixes across gateway, sync, metadata engines, and VFS components.

Key Highlights

  • Upgrade golang.org/x/net to 0.23
  • Fix memory release issues in the gateway
  • Fix metadata engine bugs for PostgreSQL, Redis, and SQL
  • Fix VFS race conditions and lock handling

Full Release Notes

This is the third patch release for JuiceFS v1.1. It has 21 commits from 6 contributors, thanks to @SandyXSD @tangyoupeng @zhijian-pro @davies @polyrabbit @prOOrc !

# Changed
- deps: upgrade golang.org/x/net to 0.23 (#4911)

# Bugfix

- cmd/gateway: fix the issue that memory is not released immediately after object downloaded (#4507)
- cmd/gateway: fix the size of JfsObjects.CompleteMultipartUpload (#4893)
- cmd/sync: fix the issue that manager's IP address is never set in cluster mode (#4939)
- cmd/clone: set the attr should be done after checking permissions (#4745)
- cmd/clone: fix source mode check (#4856)
- meta: fix the issue that `df` may see negative values (#4591)
- meta: fix the protection check for compaction (#4901)
- meta/redis: fix some transactions that are not wrapped by multi/exec (#4801)
- meta/sql: fix the issue that the nanosecond parts of mtime/ctime may be incorrect (#4618)
- meta/postgreSQL: fix the address parse (#4731)
- vfs: fix the issue that file length may not be updated immediately after fallocate & copyfilerange (#4427)
- vfs: fix the issue that file entry may still be returned if it's deleted during readdir (#4453)
- vfs: fix the issue that the source file may not be up-to-date in copy_file_range (#4436)
- vfs: fix panic when reading and writing s.writer concurrently (#4710)
- vfs: fix the race of getting file length (#4876)
- vfs: fix the issue that plock may be unlocked by mistake (#4910)
- hadoop: make libjfs singleton to avoid possible jnr weakhashmap infinity loop (#4407)
- hadoop: use auto-increment fs handle to avoid possible conflict (#4411)
- hadoop: fix guid lookupGroup (#4532)
- hadoop: rmr check perm only when the directory is not empty (#4539)