Summary: Online Disksuite 1.0 and 4.1.3

From: Doug_Tassan (tassand@tpd.eds.com)
Date: Sat Aug 12 1995 - 02:58:54 CDT


My initial question:

> Hi managers,
>
>
> If anyone has experience with ODS 1.0 and 4.1.3 please help!
>
> I'm trying to build a file system on a 20 GB disk array. Newfs
> is telling me that I've either maxed the inodes or cylinder
> groups out. Increasing and decreasing the parameters does
> not change anything.
>
> The disk array has a single controller controlling multiple
> 9GB drives. The controller looks like one big disk.
>
> It's my understanding that ODS 1.0 increases the 2GB file system
> limit to 1 TB.
>

 
We were receiving errors from newfs that basically told us the
cylinders per group were reduced from 16 to 6. So we ran newfs with
option -Nv to see what was being passed to mkfs. Then we ran the
mkfs command reducing the cylinders per group to 6 using the ncpg
option in mkfs. After that, the file system built normally.

Also, ODS did let us have file systems greater than 2GB without having
to do anything special. We now have 2 150GB arrays with on large
file system on each.

Doug

Thanks to all those who responded. Their responses are as follows:

-----------------------------------------------------------------------
 From Philip.Kao@artecon.com Mon Jul 31 18:01:41 1995

 Doug,
 
 I just did this so...
 ODS has a ridiculous upper limit, BUT SunOS still has two major
 limitations: 2.1 Gbyte per partition and 7 partitions per disk.
 
 With those two limitations, you can only get 2.1 x 7 = 14 Gbytes
 per RAID disk formatted on SunOS.
 
 So if you want a 20Gbyte partition, then you will need to use
 ODS to concatenate two different RAID disks together.
 
 Your /etc/md.conf will look something like this:
 /dev/md1a 7 1 /dev/sd1a ... /dev/sd1h
 /dev/md2a 7 1 /dev/sd2a ... /dev/sd2h
 /dev/md3a 2 1 /dev/md1a /dev/md2a
 
 where /dev/rsd1 is the first RAID disk
   and /dev/rsd2 is the second RAID disk
 
 -phil
-----------------------------------------------------------------------

>From jeffw@triple-i.com Tue Aug 1 08:23:47 1995.

One thing to watch out for is that the stock dump & restore will
puke if you have more than 4m inodes (actually dump will appear
to work, and then restore may puke when you try to restore).
There is a patch for Online: Backup that supposedly fixes this
(you just use the dump & restore from the patch).

Have you tried changing the -i option:

          -i bytes/inode
              This specifies the density of inodes in the file
              system. The default is to create an inode for
              each 2048 bytes of data space. If fewer inodes
              are desired, a larger number should be used; to
              create more inodes a smaller number should be
              given.

I always use -i16384 or 32768 on big filesystems that will
contain mostly larger files.

Jeff

-----------------------------------------------------------------------

>From /DD.ID=INSDRS01/G=Ross/S=Stocks/@nt.com Tue Aug 1 09:05:49 1995

True, but you must partition the disks into no larger than 2GB due to
SunOS4.1.3 limitations. I believe you newfs those partitions. Then you
create metapartitions using ODS.

Also you want all your true partitions within a metapartition to have
like geometry. Other wise ODS will utilize the least efficient of the
geometries to build its metapartitions and some of your disk capacity
will be wasted. If you have mixed geometries it is best if you can use
multiple metapartitions each of which only uses like geometries. That
will maximize the efficiency of your disk utilization.

I haven't dealt with it in over a year. It was one of the smoother
installs I've been through. Only a couple snags that weren't especially
difficult to get help on and the rest went smooth as silk. And so far
there have been no problems with the ODS. (But we are only using it for
concatenation, one of the simpler applications).

Hope this helps. Feel free to reply with follow up questions if need
be.

Regards,
Ross
ross.stocks@nt.com

-----------------------------------------------------------------------

>From kevin@uniq.com.au Tue Aug 1 18:45:36 1995

      
Thanks for the response...

-----------------------------------------------------------------------

--
Doug Tassan   			EDS - Technical Products Div.
E-Mail: tassand@tpd.eds.com	Voice:  (214) 612-6613



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:31 CDT