Can not extend a JFS file system
pui ming Wong
Now when i tried to increase one of its filesystem size, it failed.
with message:
chfs: 0506-932 Can not extend a JFS file system with nbpi = 4096
and frag = 4096 past 134217728 (512 byte blocks)
What should have been done when the LV was created in first place.
Maarten Kreuger
In your standard JFS, there are limits to the total number of inodes
that are available. If you look at your filesystem with df -v you can
see the total number of inodes available. In normal situations this is
overkill. If you know what the filesystem is going to be used for, you
can probably choose a higher value for nbpi. (number of blocks per inode).
The default values ensure that for each block of (standard fragment size
of 4096) there is a inode to hold a file for it. This way you can
always fully use the available space for files.
For Oracle data filesystems for instance, where there are but a few
files, you can choose a nbpi value of 131072, which means that for each
32 fragments there is an inode available. for a large filesystem, that
is still lots of inodes. (e.g. 32 GB fs => 32GB / 128 kb ) = 500K inodes )
See the chfs man page for more info:
http://publibn.boulder.ibm.com/doc_link/en_US/a_doc_lib/cmds/aixcmds1/chfs.htm
So if you create a filesystem with default settings, 64GB is the maximum
size of your filesystem. The only way to enalrge it is to backup,
remove, recreate with a larger nbpi value, and restore...
Have fun,
Maarten Kreuger
Norman Levin
working
on the assumption that the number of inodes is constant and that all a chfs does
is.... I'm not sure. After all, the size can be changed independently of using
a chfs command.
When you make a fs larger with chfs, the number of inodes must increase to handle
the files that could be created in the new larger space. You can test this by:
a: create a filesystem that will hold 1 512 block, nbpi=512 and frag=512.
Since an inode is 128 bytes, you have approximately 20% of the disk being used
for
inodes since you are setting up to handle 512 byte files. A df will show this
high percentage of space used in an empty filesystem (actually, it will be closer
to 25%)
Next, do a "chfs -a size=+1 /testfs"
This will increase your filesystem by 1 PP - effectively doubling the space.
If the inodes did not change, you should get ALL of that additional space
available
for file and see the utilization drop to 13%. Instead, it may go from 25 to 23%
Try it again (effectively adding a 50% bump in disk space as you go to 3
partitions)
A du will probably show 21% utilization.
Do it a few more times. See if you can get it below 20% utilization.
So, there is a relationship between frag size and nbpi as to maximum file system
size (you get a message from mkfs when you create), but I don't think it is
hinged on fixed inodes.
norm (shoot me down) levin