SUMMARY: DiskSuite: growing NFS-exported fs okay?

From: Richard W. Sabourin (rws@cs.brown.edu)
Date: Mon Feb 15 1993 - 10:17:49 CST


In article <1993Feb10.150018.14525@cs.brown.edu> I wrote:
>
>We have an NFS server using the DiskSuite software to concatenate disks
>into mondo-big partitions (under 4.1.3). We now wish to use growfs(8) to
>expand one of these metadisks on the fly. My question is: can I do this
>without forcing the clients to unmount the partition, and without ending up
>with stale NFS file handles? (I believe fsirand(8) causes this problem)

I received a couple of e-mail replies saying that there would be no problem;
including this one:

|Date: Wed, 10 Feb 93 12:15:11 MST
|From: Brian.Berliner@Central.Sun.COM (Brian Berliner)
|
|You should be fine. The growfs(8) command should lock the file system
|whicle it is being expanded. This will cause client NFS requests to be
|dropped by the server (for that particular file system), which will make
|the client believe that the server is down. The growfs command will then
|extend the file system by adding more cylinder groups. The "fsirand" pass
|is only applied to the newly created inodes. Your existing inodes will
|remain untouched.
|
|Please try it and report back!
|
| -Brian

This morning we performed this operation, and had no problems, as
advertised. Mount, lookup and read operations over NFS were not blocked
(Didn't try writing, it was over before I could find a writable-by-me
directory).

So the answer is: yes, growfs(8) is NFS-friendly!

        Rick S.



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:28 CDT