Jumpstart vs. Normal Installation

[kozlow]

I have been fighting through installing and mirroring a new Solaris 9 system running on a Sunfire V100. I decided to install using Jumpstart since this is my first go round. In using this forum, several users have told me not to use Jumstart with replies like the following:
Aha - Jumpstart strikes again!

0 root wm 258 - 320 125.51MB
::::
4 swap wu 0 - 257 513.98MB

A normal installation would give you

0 root wm 0 - 63 125.51MB
1 swap wu 64 - 320 513.98MB

Where can I get a "normal" layout of the whole disk and instruction on how to complete it? The manuals I recieved from Sun don't seem to help and I would like to do this right.... the second time...... Her is the disk size:
2 backup wm 0 - 19155 37.27GB (19156/0/0) 78156480

Thanks.....

 

[mattrumik]

partitioning explicit

filesys rootdisk.s0 16384 /
filesys rootdisk.s1 1024 swap
filesys rootdisk.s4 4096 /var
filesys rootdisk.s5 8192 /usr/local
filesys rootdisk.s6 free /export/home

there are docs on the web which detail everything
otherwise ask
matt

 

[kozlow]

It was also suggested to me to have s7 with 30Meg for the SDS database. Not sure how you define that as unused space until you run the metadb. Another was /usr on s3 for 8Gig and / on s0 for 4Gig. From another thread, it was suggested (like you did) to keep /var seperate from root because a lot of temp files can fill up /var and you can then go into single user mode and clean up if it halts your system. What is the advantage of combining / and /usr (if any)?

I am assuming that free means all space that in not allocated.

(Someday, we'll all look back on this and it will all seem funny)

 

[julianbarnett]

There are two main reasons for splitting up the root disk:

1] Isolation - ensure one filesystem filling up doesn't impact on others, especially root.
2] Space - with smaller disks certain filesystems do not have to be on the root disk.

Considering item 1, only filesystems which are fairly dynamic need to be considered. This means /var due to log files which might grow uncontrollably and /export/home which users may accidentally fill up.

Item 2 is more historical when systems only had 1Gb disks - then /etc, /usr had to be put on other disks. With a 37Gb disk this is not an issue!

Another reason for minimising the number of paritions is that you are limited by the OS to 6 slices per disk - unless you use soft partitions in SDS which is messy.

There is no problem having /usr contained within the root partition - in fact you gain some resilience since all your executables and libraries are available. Previous posts about not changing root default shell to say /usr/bin/ksh no longer apply.

I am always tempted to have /usr/local as a separate filesystem since packages you install here - GNU software, database executables and application software can take up a large amount of space and shoudn't really be mixed with core OS software. This is your get out of jail partition. If the root disk starts to fill up you can always relocate /usr/local somewhere else - if it's in with root things start to get messy. Some software gets installed into /opt - cater for this by making /opt a soft link to /usr/local/opt

As to sizes - thats a whole new debate. Look at what you are already using and add 50% is my rule of thumb when re-laying out a disk. There's no point in allocating 12Gb to root if it is never going to grow above 1.8Gb.

Webstart will always layout the disks in an odd fashion because it installs itself on your boot disk to start off. I always use the old fashioned way of installing Solaris - boot from Software CD 1 of 2 and use the Interactive Installation Program - details in the Advanced Installation Guide. You will probably not be able to be specific about the last 2 partitions - just leave them free and format them once Solaris is up and running.

Set up: Slice 0 = / 4Gb+ ?
Slice 1 = swap 1Gb+ ?
Slice 3 = /var 1Gb+ ?
Slice 4 = /usr/local 4Gb+ ?
Slice 5 = /export/home ??Gb Up to you how much
Slice 6 = free Possible use later
Slice 7 = free (last couple cylinders for metadb)

 

[kozlow]

Thanks for the info... This forum has been a great help.

I have been unable to re-initialize my filesystems (booting off of the cd), getting a PANIC [cpu0] bad trap = 31 after setting up the network. It then does a normal boot off of the disk.

I am going to try somethings and see if I can fix it. Was going to look at booting of the OS disk and see if I can start everything from scratch... Nothing to lose....

First, I may remove all the meta setups and see if I can clean everything up. Not sure why that would cause a panic, but it is trial by error right now....