UPGRAGDE NOTICE
UPGRAGDE NOTICE
If your CubeFS version is v2.3.x or before, please refer to the UPGRADE NOTICE in v2.4.0 for upgrading steps.
if you CubeFS version is v3.2.1 or before, and need to upgrade to v3.3.*,you must follow these upgrade steps:
- When upgrading the metanode node, you need to add "raftSyncSnapFormatVersion": 0 to the configuration file.
- After all metanodes are upgraded, remove this configuration item so that raftSyncSnapFormatVersion defaults to 1.
- Restart all metanodes.
- Please upgrade the client at the end, the master has enhanced the check of the client, otherwise it will cause the client to mount abnormally
If your Blobstore version is v1.1.0 or before which built with cubefs-blobstore (https://github.com/cubefs/cubefs-blobstore) , please refer to UPGRADE to v3.2.0 following these steps #1556.
Main Feature
objectnode: ObjectNode support post object. （#2695，@yhjiango）objectnode: S3 sts and signature auth. (#2488, @yhjiango)libsdk/meta: Support dir lock. (#3344, @longerfly)master/data/meta: Support write disable option for volume. (#2537, @NaturalSelect)master: Support freeze volume and delayed deletion of volume. (#2922, @shuqiang-zheng)master: Support compressing client/partitions api response data to save bandwidth。 (#3314, @lily-lee)
Enhance
objectnode: Concurrent multipart upload security mechanism (#2928, @tangdeyi)master: enable rename atomic operation by default (#2956, @Victor1319)master: Optimize the display of capacity when "used" is greater than "total". (#3315, @true1064)libsdk: Support more api function for libsdk. (#3316, @longerfly)data: Little io error only set current datapartition as broken. (#3317, @true1064)client: Write to tiny extent when file size less than 1M. (#3318, @leonrayang)master: Optimize CPU cost when invoking the volStat api (#3319, @true1064)master: Optimize the speed of applying raft snapshot for master module. (#3320, @NaturalSelect)meta: Support rotating and deleting inode files to avoid excessive size. (#3321, @NaturalSelect)data: Speed up the startup speed of the datanode. (#3322, @NaturalSelect)master: Optimize the volume deletion process. (#3323, @Victor1319)master: Add a rdonly flag to the client/partitions api to indicate whether a volume is read-only. (#3324, @true1064)data: Support asynchronous deletion of expired datapartitions. (#3325, @Victor1319)data: Prioritize listening to external service port when starting the datanode. (#3326, @Victor1319)data: Synchronously persist dp meta information when creating dp.(#3335, @bboyCH4)client: Support retrying requests for failed extent writes when the client encounters limit io error. ( #3338, @NaturalSelect)
Bugfix
objectnode: PartNumber of multipart must be greater than 0. (#2885, @yhjiango）objectnode: Listobjectv1 with delimiter may lead to panic. (#2931, @tangdeyi)client: When the recycle bin is enabled, avoid deleting non-empty directories. (#3328, @bboyCH4)libsdk: Fix some issues encountered during the usage of the libsdk. (#3329, @longerfly)data: Optimize the blocking issue during data partition migration in snapshot synchronization. (#3330, @bboyCH4)meta: Fix the issue of possible duplication caused by the unpersisted meta unique id. (#3331, @true1064)master: Master follower replicas only request the leader replica to obtain the partition list.(#3332, @NaturalSelect)client: Fix the issue that interruption causes failure to write data to ec.(#3333, @Victor1319)data: Fix the issue of memory leaks caused by blocking during data transmission between datanode leader and follower. (#3336, @Victor1319)data: During the process of reusing network connections, there is a possibility of using incorrect network response messages. (#3337, @Victor1319)
Contributors
leonrayang, true1064, and 8 other contributors