[Pvfs2-users] Out of space

[Lars Roland]

Hi

I have a 5 node cluster where each node is having 2TB of space, due to
some changes in the data model we are running out of space within a
month or two. According to the FAQ PVFS2 does not handle the addition
of a new I/O server very well (i.e. it will not get used more than the
existing ones, since the data is still striped round-robin over the
old servers). Fortunately I am running LVM on each of the servers so
my idea is to do something like the following:

---
Add a extra node, create 5 partitions and run NBD on these, now add
these network block devices to the existing LVM volumes on the
existing I/O servers (one for each server) and resize the physical
file system (ZFS) on each I/O servers.
----

Will this work (in short if I resize the underlying file system will
PVFS2 then use the extra space automatically ?). I know I will se a
performance degradation (I am trying to get money for five servers
instead of one so I can add one server to each existing one) but in my
case running out of space is worse than a performance degradation.

Is there anyone working on adding the ability to only stripe on a
section on servers (such that I can add extra servers and get instant
space utilization) ?.

--
Lars