博客公告:
(1)本博客所有博客文章搬迁至《博客虫》http://blogchong.com/
(4)该博客内容还会继续更新,不过会慢一些。
http://blog.sina.com.cn/s/blog_8c243ea30101kkiy.html
类型 | 详细 |
备注 | 该文档为在使用GlusterFS_3.4.1中发现的bug,详细内容为向红帽bugzilla报告的主要内容。 |
相关描述 | ² ² ² ² |
目录
2.1 AFR: cannot get volume status when one node down. 1
2.2 AFR: change one file in one brick,prompt "[READ ERRORS]" whenopen it in the client. 2
2.3 AFR: lose files in one node, "ls" failed in the client, butopen normally. 2
2.4 AFR: “volume heal newvolume full” recover file -- deleted filenot copy from carbon node. 3
1 文档说明
该文档为在进行GlusterFS_3.4.1版本实践中所遇到的Bug,至此,贴出来,是为了寻求大家的帮助,看看大家在使用的过程中有没有遇到相同的问题。
个人也想明确是个人使用有问题,还是系统真的有这些Bug。
目前以下4个Bug已经向红帽的bugzilla提交,有兴趣的也可以去看看。
以下给出4个bug的链接:
https://bugzilla.redhat.com/show_bug.cgi?id=1029482
https://bugzilla.redhat.com/show_bug.cgi?id=1029492
https://bugzilla.redhat.com/show_bug.cgi?id=1029496
https://bugzilla.redhat.com/show_bug.cgi?id=1029506
2 Buglist
2.1 AFR: cannot get volume status whenone node down
Component:replicate
Version:3.4.1 & 3.3.2
Bug Number:1029482
Description of problem:
It is OK when all nodes are up,but break one(cut off net or shutdown node) or more cannot get volume status(command:volumestatus).
Version-Release number of selected component (ifapplicable):
3.4.1 & 3.3.2
How reproducible:
always
Steps to Reproduce:
1.create a AFR volume and start
2.gluster volume status(normal)
3.break one node (cut off net:make ethX down--ifconfig ethXdown)
4.gluster volume status(abnormal)
Actual results:
cannot get anything or "Anthor transactior is in progress,Pleasetry again after sometime"
Expected results:
get something about volume status
Additional info:
I get something from bug_807428,but it appear both in 3.4.1 and3.3.2
2.2 AFR: change one file in onebrick,prompt "[READ ERRORS]" when open it in the client
Component:replicate
Version:3.4.1 & 3.3.2
Bug Number:1029492
Description of problem:
AFR volume: change a file in one brick,execute "gluster volumeheal afr_vol full"(actuclly,it doesnot work).open this file in theclient,prompt "[READ ERRORS]".
Version-Release number of selected component (ifapplicable):
3.4.1 & 3.3.2
How reproducible:
always and i have test many times.
Steps to Reproduce:
1.create a afr volume: gluster volume create afr_vol replica 3192.168.8.{80,81,82}:/mnt/sdb1
2.started it: gluster volume start afr_vol
3.change one file in 192.168.8.80:/mnt/sdb1 (just add a row ordelete a row)
4.execute heal:gluster volume heal afr_vol full(it doesnot workand you can see from 8.80:/mnt/sdb1)
5.open this file (used 'vim')
Actual results:
prompt "[READ ERRORS]"
Expected results:
open file normally
Additional info:
"volume heal Volume_name full" doesnot work and I have test manytimes.
2.3 AFR: lose files in one node, "ls"failed in the client, but open normally
Component:replicate
Version:3.4.1 & 3.3.2
Bug Number:1029496
Description of problem:
AFR volume: create a afr volume. delete a file inbrick(default node:system get metadata from this node),"ls" failedin client,but open normally.
Version-Release number of selected component (ifapplicable):
3.4.1 & 3.3.2
How reproducible:
only delete file in special node (default node)
Steps to Reproduce:
1.create a afr volume
2.delete a file in default node(system get metadatafrom this node)
3."ls" in client
Actual results:
cannot show this file,but can use it normally. Andit can trigger heal.
Expected results:
show this file
Additional info:
2.4 AFR: “volume heal newvolume full”recover file -- deleted file not copy from carbon node
Component:replicate
Version:3.4.1 & 3.3.2
Bug Number:1029506
Description of problem:
AFR volume: create a AFR volume, and then change afile in one brick (just add a row, so it is a wrong file).I deletethis wrong file in this brick, and then execute "gluster volumeheal afr_full". Guess what? yes,it recover a file in this brick,but ... this file is a wrong file(changed by me),not a copy fromcarbon node. why ?
Version-Release number of selected component (ifapplicable):
3.4.1 & 3.3.2
How reproducible:
always and I have test many times
Steps to Reproduce:
1.create a AFR volume : afr_vol
2.change a file in one brick (just add a row)
3.gluster volume heal afr_vol full
4.recover a file (changed by me,not a copy fromcarbon node)
Actual results:
this file is the file which changed by me
Expected results:
recover a normal file (copy from other node)
Additional info:
so why?