SUMMARY: UFS within UFS

From: Sun Manager <sunfired_at_gmail.com>
Date: Thu Mar 22 2007 - 23:19:54 EDT
Thanks for those quick responses. The concensus was to use the loop back fs
(lofs) works great. My boss is smiling again. Thanks Graeme and Francisco
who pointed me to the right direction.

mount -F lofs <absolute path to directory to be mounted> <mount point>

Doing a df gives me displays the mounted directory with the full long
absolute pathname , kind of clutters the display but what the heck - it
works great !!

--SF



---------- Original message ----------
From: Sun Manager <sunfired@gmail.com>
Date: Mar 22, 2007 3:50 PM
Subject: UFS within UFS
To: sunmanagers@sunmanagers.org


Hi Gurus,

I have a system here that has mounted partions for some users to upload
their data. Each user has a fixed space in which they can upload their data.
Users sftp into a chrooted environment to upload their data. One user (my
boss) wants extra space in his area to upload his data. I tried to link a
directory which had enough space from within his work directory, but it will
not work because it is a chrooted environment and he cannot traverse down
the link. I had no other partitions  free to mount it onto his work area. So
I thought I had a brilliant idea and proceeded to create a filesystem within
a filesystem and have it mounted to a subdirectory in his work area. Here is
what I did:

There is about 50g free in /usr1

# mkfile 50g /usr1/newpart.ufs
# lofiadm -a /usr1/newpart.ufs
# newfs /dev/rlofi/1
# mounst /dev/lofi/1 /<my boss's work area>/temp

All this worked fine. However there is a big performance hit. Using Bonnie I
tried to test IO on a space of 1Gb and get te following result

              -------Sequential Output-------- ---Sequential Input--
--Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block---
--Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec
%CPU
         1024   326  0.7   327  0.2   437  0.6 11973 36.7 148240 85.0 432.8
1.6

Just a few hundred K/sec? That's not acceptable. Doing the same test on a
normal partition yields

              -------Sequential Output-------- ---Sequential Input--
--Random--
              -Per Char- --Block--- -Rewrite-- -Per Char- --Block---
--Seeks---
Machine    MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU  /sec
%CPU
         1024 35563 77.2 40732 25.4 30640 33.2 33897 99.9 213579 96.5
31582.6 165.8

What explains this poor performance? How can I improve performance on the
new mount point? Am I missing something?

Thanks in advance

--SF
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Thu Mar 22 22:20:45 2007

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:44:05 EST