[Gluster-users] [Gluster-devel] Need suggestion to change out put of volume status command

[JF Le Fillâtre]

Hello all,

On 27/11/14 07:15, Jiffin Tony Thottan wrote:

>> Status of volume: xcube
>> Gluster process                                 Port     Online   Pid
>> ------------------------------------------------------------------------
>> Brick 192.168.44.108:/home/brick                49157    Y        6030
>> Brick 192.168.44.108:/home/brick.rdma           49157    Y        6030
>> NFS Server on localhost                         2049     Y        6532
>> Self-heal Daemon on localhost                   N/A      N        6234
> 
> May be ,it looks like weird one showing two brick process with same
> pid.But this change
> is only for tcp,rdma volumes, in other cases it will continue with
> existing volume status
> command. So it won't effect any current test scripts

It will at some point, not just right now if you don't use RDMA. Which may be soon if the packages start pulling in RDMA dependencies and your adapters support it.

All of those modifications will require rewriting scripts to take care of corner cases, even if you're not hitting them right now. The question is, how much will you have to modify them? In that case above, the fact that you can't just grep on a brick name or on ^Brick and get one unique result per brick will require quite a lot more logic to handle than adding one field on the line.

I believe that brick entries should stay unique, and that the RDMA port should be encoded some other way (and not in the name -- see my previous email).

Thanks!
JF