curve

This release is from fs-release0.1.0 branch. And it's mainly for curvefs.

curve-ops-tools: fix chunkserver ip:port error

The first version of CurveFS.

• Change auto map service to systemd standard

• Change auto map service to systemd standard

new features

• curve-nbd: support simulate 512-byte logical/physical block device
• curve-nbd: prevent mounted device to unmap
• curve-nbd: add auto mount options
• nebd: support multi-instances for different user
• client: retry allocate segment until success
• mds: schedule pending online chunkserver

bug fix

• ansible: backup and recover curvetab
• debian package: backup and recover curvetab
• scripts: fix potential overflow when cnovert ip to value
• client: fix segment in SourceReader
• client: explicit stop LeaseExecutor in FileInstance::UnInitialize
• curve-nbd: fix concurrent nbd map
• snapshotclone: fix cancel task lost
• tools: remove copyset data after remove peer
• mds: modify stale heartbeat update error with warning
• chunkserver: fix config change epoch update error

see CHANGELOG-1.3

new feature

• Support 512 bytes aligned IO requests.

see CHANGELOG-1.3

bug fix

• Fix duplicate delete segment in mds on discard feature.
• Fix trash metirc statistic bug.
• Fix copyset status judgment after set pendding on online statue.
• Add mount options in auto map nbd volume script.
• Fix add qos for clone volume.
• Fix crc issues in chunkserver.
• Fix crc issues in mds and add scan-status support in curve_ops_tool.
• Fix core dump when chunkserver SIGINT exit sometimes.

new features

1. Support discard to garbage collection.
2. QoS of client.
3. QoS of SnapshotCloneServer.
4. Recover file in recycleBin.
5. 5. Fill chunks in chunkfile pool with 0 in the background.
6. Silent data CRC verification in the background.
7. Periodic automatic cleaning of recyclebin.

optimization

1. Optimize optimize build script, optimize log printing.
2. Translate some document and code comment from Chinese to English.
3. Mark a online chunkserver to pendding status to migrate it's data.
4. Clean up the remaining copyset after remove peer.
5. Nbd auto map at reboot.
6. ansible script improve:
   • Improve bool type var and some issues.
   • Improve daemon restart.
   • Incorporate some scripts in debain package and ansible template.

bug fix

1. Fix recyclebin space statistics error bug.
2. Fix a metric bug when not use walpool.
3. Fix a nbd map concurrently bug.
4. Fix discard and flatten concurrency issues.
5. Fix max chunk id calculation error when use chunkfile poo as wal pool.
6. Fix client retry when allocate segment failed as cluster space is pool.
7. Add protection for unmap operation when io is not stopped.
8. Fix a segment fault when client read from clone source.

new features

• support nbd auto map at boot

fix some bug from v1.2.0-beta2

• fix some metirc bug
• fix stripe bug that too small stripeUnit cause io error
• fix bug CreateCloneFile missing param
• add option for read/write stripe file as normal
• move some file param from hard code to config file

changes from v1.0.2-rc0

1、change max file length from 4TB to 20TB.
2、move param file max length , file min length, file segment size from hard code to config file

fix some bug from v1.2.0-beta-1

1. fix some ansible script bug
2. fix some log error
3. fix some tool bug

new features

• When add a new pool, default new logic pool is DENY, need tool to enable it.
• When io error occurs, whether nebd-server discards the corresponding rpc request depends on configuration setting(discard default).
• When client read a unallocated space, it does not allocate the segment to improve the space utilization.
• Add data stripe feature.

optimization

• Optimize ansible script, optimize build script, optimize log printing.
• Translate some document and code comment from chinese to english.
• Curve_ops_tool statistics chunkserver capacity by pool.
• Add a script for k8s to attach curve volume.
• curve_ops_tool improve:
  • Check copyset consistency between chunkserver and copyset.
  • Support to take a snapshot of copyset.
  • Support to list the volume on copyset where majority copy is failed.
• Clean up the unit test temporary folder.

bug fix

• Fix clone delete bug.
• Nbd unmap need wait thread exit.
• Mds need check file attach status.
• Fixed when disk fails, copyset reports an error, but sometimes the chunkserver does not exit.
• Fixed the direct_fd leak problem when wal write disk.
• Fixed the atomicity problem of GetFile when not using the file pool.
• Fixed in cluster_basic_test mds start error on InitEtcdClient() which produced coredump

1. 修复了对云主机创建快照失败卡住的bug
2. 部分修复了切换mds后马上 创建云主机镜像，概率性分配chunk失败导致镜像创建失败的bug（修复重试逻辑部分，使得连接失败可以正常重试）
3. 修复没有写满的云主机创建自定义镜像再创建云主机，云主机启动失败的一个bug
4. 修改当前mds，chunkserver，client的日志打印，以避免线上过多无效报警的情况
5. 修复nebd-daemon stop nebd的一个bug
6. 修复同时close大量文件导致nebd死锁的一个bug

1. 修复了对云主机创建快照失败卡住的bug
2. 部分修复了切换mds后马上 创建云主机镜像，概率性分配chunk失败导致镜像创建失败的bug（修复重试逻辑部分，使得连接失败可以正常重试）
3. 修复没有写满的云主机创建自定义镜像再创建云主机，云主机启动失败的一个bug
4. 修改当前mds，chunkserver，client的日志打印，以避免线上过多无效报警的情况
5. 修复nebd-daemon stop nebd的一个bug

1.0.0版本存在bug，无法支持多pool，升级到1.0.1版本及以上需要使用工具mysql2etcd修复mds对copyset的元数据编码，步骤如下：

使用命令：
./mysql2etcd -op=fix_copyset
修复copyset元数据，并观察工具输出有无异常
使用命令
./mysql2etcd -op=check_copyset
校验copyset修复是否一致，观察工具输出有无异常。
然后正常升级mds