[OpenAFS] Add new fileserver
Alexander Boström
abo@kth.se
Sun, 08 Apr 2007 11:05:15 +0200
sön 2007-04-08 klockan 15:18 +0800 skrev Melvin Wong:
> But the client still see only the 1st afs's vol space when I "fs
> diskfree /afs". Must I perform another step or the storage space
> should automatically combine once it sync? Is there any guide out
> there to help me? Appreciate if someone can guide me.
What you see in /afs is the contents of the volume root.afs in your
cell. Traditionally, the contents of that volume is a set of mount
points to the volume root.cell in each of the different AFS cells you
would like to be able to access, including your own cell. These are not
Unix mount points, but rather internal to AFS. They can be created like
this (assuming the name of your cell is example.com):
cd /afs
fs mkm -dir example.com -vol root.cell -cell example.com
fs mkm -dir .example.com -vol root.cell -cell example.com -rw
fs mkm -dir stacken.kth.se -vol root.cell -cell stacken.kth.se
fs mkm -dir openafs.org -vol root.cell -cell openafs.org
and so on. You probably need to actually create the root.cell volume
first, using the vos command. You should also use the vos command to
replicate your root.afs and root.cell volumes (and as many other volumes
as possible) so they look something like this:
$ /usr/sbin/vos exa root.afs -c stacken.kth.se -noauth
root.afs 536870912 RW 55 K On-line
kexchoklad.stacken.kth.se /vicepa
RWrite 536870912 ROnly 0 Backup 536870914
MaxQuota 5000 K
Creation Mon Sep 15 22:58:37 1997
Copy Thu Oct 26 21:25:20 2006
Backup Sun Apr 8 00:58:43 2007
Last Update Mon Jun 27 19:08:13 2005
0 accesses in the past day (i.e., vnode references)
RWrite: 536870912 ROnly: 536870913 Backup: 536870914
number of sites -> 3
server kexchoklad.stacken.kth.se partition /vicepa RW Site
server kexchoklad.stacken.kth.se partition /vicepa RO Site
server fishburger.stacken.kth.se partition /vicepb RO Site
Notice how kexchoklad:/vicepa contains both a the RW volume and a RO
replica. This is recommended since the extra "copy" isn't really a copy,
so it comes essentially without cost.
The .example.com directory is there to help you use read-only
replication (for redundancy and load balancing), which is highly
recommended.
After you've created these, you can start adding more volumes and mount
points to build a useful directory structure. A typical one looks like
this:
/afs - Unix mount point of root.afs in stacken.kth.se.
/afs/stacken.kth.se - AFS mount point of root.cell in stacken.kth.se
/afs/stacken.kth.se/home - AFS mount point of the volume home in
stacken.kth.se
/afs/stacken.kth.se/home/abo - The volume home.abo in stacken.kth.se
/afs/stacken.kth.se/projects - A directory
/afs/stacken.kth.se/projects/arla - The volume projects.arla in
stacken.kth.se
And so on...
Now, to answer your question: Each volume can reside on one partition on
one server, with multiple read-only replicas of it on all or a subset of
all partitions of every server. So to use more servers, you need to
create more volumes.
/abo