[Gluster-users] intended behavior filter translator

cnr cnr.roxx at gmail.com
Mon Jun 22 11:04:08 UTC 2009


Hi,

Ate Poorthuis pisze:
> Hi all,
> 
> Can someone enlighten me about the intended behavior of the filter 
> translator? From the documentation, I thought it would behave the same 
> as NFS mapping/squashing. However, this is not what I see in my setup.
> 
> Let's say I map everything to UID 1500 - using either the fixed-uid or 
> the translate-uid and gid option. Now, on the client side, every file 
> and directory appears to be owned by 1500. If I try to create new files 
> or directories as uid 1001 this fails because of a lack of permission.  
> If I chmod 777 a directory then user 1001 can create new 
> files/directories but cannot change them afterwards as they appear to be 
> owned by 1500. On the server side, those files are owned by 1001. This 
> is exactly opposite of NFS. There mapping everything to 1500 has the 
> result that every file created by 1001 is owned by uid 1500, but 1001 
> can change these files since his uid is mapped to 1500.
> 
> Am I doing something wrong or is this intended behavior? I have tried 
> loading the filter translator on both the client and the server side. 
> They both give the same result. The end goal is to have every user in 
> the network write and read each other's files. I thought uid mapping 
> would be the best way to do this.
> 

I confirm described behaviour of filter translator. Is there any
workaround to map client-side uid & gid to server-side uid & gid? i'm
using glusterfs 2.0.1 and debian-stable fuse module.

regards, konrad szeromski






More information about the Gluster-users mailing list