[Gluster-users] has anyone seen the impact of "running out of space" on a gluster volume?

[Joe Landman]

It looks like there are several possible modes, and I am wondering what 
the response to these modes are with gluster.

1) a brick runs out of space during a write.  While there is more space 
available elsewhere in the unit, this one brick is full.

2) the aggregate file system runs out of space.  This is a slight 
variation on the above, in that an allocation should occur on the least 
full brick at some point (when is that crossover?)

So, what should a user that is trying to do a write, where the write 
exceeds the size of the brick, observe in terms of an error return?

ENOSPC? something like this?

A customer just reported a hang for a 100TB file system after filling 
it.  Basically df and related hung.  They are using the native client 
due some unresolved issues on the NFS server side.  This may be part of 
the issue as well, as I'd expect the NFS server to be somewhat more 
likely to report errors people expect.

Anyone run into this?  We are going to do some experimentation here 
before figuring out if this is something that warrants a bug report, but 
I wanted to see if someone else had seen something like this.

Joe


-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics Inc.
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
        http://scalableinformatics.com/sicluster
phone: +1 734 786 8423 x121
fax  : +1 866 888 3112
cell : +1 734 612 4615