ZhangRui comments

Results 49 comments of


                                            ZhangRui

1.10 版本消费kafka 一段时间后不继续消费报错

https://github.com/childe/gohangout/issues/82 有点像这个问题 ``` grep join|grep 20: E1213 20:14:14.485455 1 group_consumer.go:461] failed to send heartbeat, restart: parse response of 12(0) from canal-kafka05-kieldsdet:9092 error: The group is rebalancing, so a rejoin is...

1.10 版本消费kafka 一段时间后不继续消费报错

1.8.2 没这个问题，我编译一个c go 凑合用着先

[Bug]: Reading the file failed, and the cfs-client log reported "super.go:405: Flush: ino(318538) err(bad file descriptor) and Get node cache failed: ino(16927248)"

[cfs-client.err.zip](https://github.com/cubefs/cubefs/files/13507968/cfs-client.err.zip)

[Bug]: Reading the file failed, and the cfs-client log reported "super.go:405: Flush: ino(318538) err(bad file descriptor) and Get node cache failed: ino(16927248)"

``` find . -inum 318538 ./.ldm/x86_64-conda-linux-gnu/sysroot/usr/include/bits/wordsize.h ^C less ./.local/miniconda3/envs/trtllm/x86_64-conda-linux-gnu/sysroot/usr/include/bits/wordsize.h ``` There is no problem in checking this file directly according to the log error report.

[Bug]: input/io error

dn log There is definitely free disk space. I don’t know why I am getting this error. ``` 2024/01/18 04:16:19.340211 [ERROR] wrap_operator.go:103: action[OperatePacket] id[Req(2335818448)_Partition(9311)_Extent(57590)_ExtentOffset(0)_KernelOffset(0)_Size(8)_Opcode(OpCreateExtent)_CRC(0)_ResultMesg(DiskNoSpaceErr)] isPrimaryBackReplLeader[true] remote[10.90.140.71:52224], err[op(OpCreateExtent) error(ActionCreateExtent:_no space left...

[Bug]: input/io error

Configured "minWriteAbleDataPartitionCnt": 0 and used https://github.com/cubefs/cubefs/commit/e94b9bf093451e5cb7e7c6f36fed6ace77201415 After optimization, there is still a certain probability of occurrence, but the probability of occurrence is reduced. ``` 2024/01/25 01:15:54.335788 [WARN ] client.go:174: serveRequest:...

[Bug]: After the double data partition is decommissioned and the damaged disk becomes a single data partition, how to repair it?

> @zcola The suitable way is update the metadata of datanode partition then restart. This a desgin issue known as automic incomplete, the release-3.4.0 will be available in the next...

[Bug]: cp a file stuck, can be reproduced

> @zcolaDid it happen only on the specified file? Could you trigger it again and show the file name and the time. Then upload the client log here. The file...

[Bug]: cp a file stuck, can be reproduced

问题依然存在，有办法怎么定位吗不如搞一个debug 版本的cfs-client，我们做一个 D stat 进程的告警，第一时间发现和提取现场，不好复现，业务看起来都用了git ，并通过vscode-server 远程调试 2.5 客户端好像是不会，但不能100% 确定，因为不好复现，而且只有3.3.0 才会做tinyfile 聚合我们很需要这个功能。业务进程会 D state 掉，kill cfs-client 进程是正常退出的，说明cfs-client 没有死掉，里面一些逻辑没有处理好，让业务一直等，杀掉 cfs-client ，进程 D state 立刻解除了，是不是可以理解为业务立刻收到了 io error The problem still...