architecture, data organization and topology structure of CURVE. CurveBS uses the central node Metadata Server (MDS) to manage virtual disk mapping and data replicas distribution. Decentralization strong in most cases. The CurveBS core consists of three parts: 1. Metadata Server (MDS) l Manages and stores metadata information and persists the data in ETCD l Collect cluster status and interfaces for upper-layer applications l Interacts with MDS to add, delete, modify, and query metadata l Interacts with the chunkServer to read and write data 4. Snapshotcloneserver l Independent

file access to the filesystem owner, so that all users (including root) can access the files. allow_root This option is similar to allow_other but file access is limited filetype needs_recovery extent 64bit flex_bg sparse_super large_file huge_file dir_nlink extra_isize metadata_csum Filesystem flags: signed_directory_hash Default mount options: user_xattr acl . effective rights -E, --no-effective print no effective rights -s, --skip-base skip files that only have the base entries -R, --recursive recurse into subdirectories -L, --logical

server Chunkserver is a service instance on a physical serverCurve metadata organization Curve maps virtual block devices to files Each file contains chunks scattered across storage nodes in the cluster

opencurve.io/Agenda • CurveBS Architecture • CurveBS Topology • CurveBS Data Organization • MetaData Server (MDS) • ChunkServer • Client • CurveBS IO processing flow • CurveBS Performance considerations status to operatorCurveFS ArchitectureCurveFS Data OrganizationCurveFS File OrganizationCurveFS Metadata ServerCurveFS ClientCurveFS Mknod FlowCurveFS Write S3 FlowCurveFS Roadmap • CurveFS based on CurveBS

through helm chart • upgrade automation • Supporting Curve Cluster provisioning in helm chart • metadata backup and recovery • MDS / ChunkServer should respect failure domains of Kubernetes • Support drives • Support dynamic volume resizing • chunkserver on PVC support for different data and metadata(HDD, SSD, pmem) • Allow chunkserver on PVCs to automatically grow when the cluster is nearly fullFeature

object时, 执行读的客户端的缓存可能还是原有的chunkinfolist, 可能会去读已经删除的object, 这种时候读会失败 可以使用双重保证 读失败的时候retry, 或许可以重拉metadata 整理后, mds在一个时间间隔内主动告知client这个inode元数据缓存失效, 重拉 写: 只是对chunkinfolist做新增, 不影响整理对原有部分的变更 删除: 已标记为删除的inode不进行整理

）中指向这个打开的文件。打开的文件描述符记录中保存着文件的offset 和 文件status。 每个进程都有个 task_struct 描述符用来描述进程相关的信息，其中有个 files_struct 类型的 files 字段，里面有个保存了当前进程所有已打开文件 描述符的数组，而通过 fd 就可以找到具体的文件描述符:© XXX Page 3 of 23 open & openat 系统调用 by the filesystem. When writeback caching is enabled, the kernel may send read requests even for files opened with O_WRONLY. The filesystem should be prepared to handle this. When writeback caching is