Bug 1389910 - RFE : Allow ":" as brick path name
Summary: RFE : Allow ":" as brick path name
Keywords:
Status: CLOSED UPSTREAM
Alias: None
Product: GlusterFS
Classification: Community
Component: cli
Version: mainline
Hardware: All
OS: All
medium
low
Target Milestone: ---
Assignee: Atin Mukherjee
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2016-10-29 12:57 UTC by Alessandro Corbelli
Modified: 2018-11-19 05:29 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2018-11-19 05:20:43 UTC
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Github 578 0 None None None 2018-11-19 05:29:47 UTC

Description Alessandro Corbelli 2016-10-29 12:57:22 UTC
gluster doesn't allow the use of ":" in the brick path name

# gluster volume replace-brick gv0 1.2.3.4:/export/brick1/brick
1.2.3.4:/export/pci-0000:01:00.0-scsi-0:0:2:0/brick commit force
wrong brick type: 1.2.3.4:/export/pci-0000:01:00.0-scsi-0:0:2:0/brick,
use <HOSTNAME>:<export-dir-abs-path>
Usage: volume replace-brick <VOLNAME> <SOURCE-BRICK> <NEW-BRICK> {commit force}


There are 2 possible solutions to this:

1) get the hostname up to the first occurence of ":" and leave the rest as brick path. This would breack IPv6 compatibility and IPv6 users should use hostnames and not IPv6 addresses (that containes ":")

2) get the hostname up to the first occurrence of ":/". This preserve the IPv6 compatibility (as there is no '/' in an IPv6 addresses) and should also fix this bug as the brick path must always be an absolute path (thus always start with a '/')

The use of ":" in brick path is usefull when you want to call each brick as the physical path of the underlying disk, like for ZFS. In my test case, the "pci-0000:01:00.0-scsi-0:0:2:0" is the disk located at device "pci-0000:01:00.0-scsi-0:0:2:0". When you have multiple bricks per servers and multiple servers for each cluster, referring to disk path is better than calling "brick1" or "brick2"
"brick1" is hard to remember, while with "1.2.3.4:/pci-0000:01:00.0-scsi-0:0:2:0" you know immediatly that is disk at slot number2 on server 1.2.3.4

Comment 1 Vijay Bellur 2018-11-19 05:29:47 UTC
Migrated to github:

https://github.com/gluster/glusterfs/issues/578

Please follow the github issue for further updates on this bug.


Note You need to log in before you can comment on or make changes to this bug.