?
Solved

New Solaris 9 install - disk layout?

Posted on 2006-05-16
29
Medium Priority
?
3,003 Views
Last Modified: 2013-12-21
I'm installing Solaris 9 on a server that has four drives of 70 gig each.  I installed the OS on first disk c0t0d0 and it looks like the following:

/dev/dsk/c0t0d0s0    15127687  105041 14871370     1%    /
/dev/dsk/c0t0d0s5    5042282 1482560 3509300    30%    /usr
/proc                      0       0       0     0%    /proc
mnttab                     0       0       0     0%    /etc/mnttab
fd                         0       0       0     0%    /dev/fd
/dev/dsk/c0t0d0s3    10085388  105684 9878851     2%    /var
swap                 13536776      48 13536728     1%    /var/run
swap                 13537728    1000 13536728     1%    /tmp
/dev/dsk/c0t0d0s2    15127687 1837856 13138555    13%    /opt
/dev/dsk/c0t0d0s4    5042282  514862 4476998    11%    /swap
/dev/dsk/c0t0d0s1    15127687  200292 14776119     2%    /export/home

If I decide to install Veritas VXFS, I should probably leave the layout as is and encapsulate all the disks?  If I was to create these filesystems on different disks out of the four disks, would that be a problem after installing Veritas VXFS in order to create that mirror?
0
Comment
Question by:nsome
  • 12
  • 6
  • 6
  • +2
29 Comments
 
LVL 14

Expert Comment

by:arthurjb
ID: 16696118
My personal preference would be to not use Veritas, since Solaris 9 has disksuite built in which will do a fine job.

The layout as you have it wastes alot of disk space. With /usr and /var broken out as you have it (Which is great in my opinion) the / partition does not need to be veery big, 1 gig is even a lot more than necessary.  

Anyway, what I would do at this point would be to take a second drive and mirror this one using disksuite.

Then take the other 2 and make a mirrored pair with those.  No reason to add the complexity of a third party product when Solaris has it built in, and works well...
0
 
LVL 38

Assisted Solution

by:yuzh
yuzh earned 400 total points
ID: 16696329
Why not use Solaris 9 Volume Manager to setup the mirror or RAID, it is installed
by default (the procedure is exactly the same as Solostice disk suite), please remeber to use a small partion/slice (10-15MB) for metastate database.

Here are some good docs with examples:
http://www.adminschoice.com/docs/solstice_disksuite.htm
http://www.idevelopment.info/data/Unix/Solaris/SOLARIS_CreateVolumes_VolumeManager9Commands.shtml
http://www.itworld.com/Comp/1375/nls_unixraid050519/index.html (part 1)
http://www.itworld.com/Comp/1375/nls_unixraid2050526/index.html (part 2)


0
 
LVL 21

Assisted Solution

by:tfewster
tfewster earned 800 total points
ID: 16696907
I notice you've used slice 2 for a filesystem, wheras normal convention is leave it as the whole disk. Personally, I'd do:
s0 / 1Gb
s1 swap  4Gb (Or = memory?)
s2 (whole disk)
s3 /var
s4 /usr
s5
s6 Remaining space - Create soft partitions for /export/home
s7 Disksuite metadb  100Mb


I wouldn't encapsulate the OS disk(s) into Veritas volume manager; Altough it makes resizing filesystems easier, recovery is much more difficult if you have a problem. Disksuite works fine for the OS disks
0
VIDEO: THE CONCERTO CLOUD FOR HEALTHCARE

Modern healthcare requires a modern cloud. View this brief video to understand how the Concerto Cloud for Healthcare can help your organization.

 
LVL 21

Expert Comment

by:tfewster
ID: 16696917
Oops, hit "submit" too soon...
s5 = opt

s3/4/5 to be sized 50% more than you think you'll need, but I wouldn't over-size /var - A regular housekeeping policy should keep it clean ;-)
0
 

Author Comment

by:nsome
ID: 16698444
I take it that a few of you use or like Veritas vxfs / volume manager?

How come?
0
 
LVL 14

Expert Comment

by:arthurjb
ID: 16699972
>I take it that a few of you use or like Veritas vxfs / volume manager?

At one time it was the best way to provide a safe file system, but now the included Solaris tools do just as good a job, for everyday useage.  

Plus the Solaris tools are free and much easier to use.  

During the last 10 years, I have not had a disk failure or process where I said, I really wish I had Veritas. Maybe in a huge environment with tens of thousands of users, and clustered machines, vxfs might be a good investment in time and money.

Suns Disksuite and Metatools have provided great service and no loss of data.
0
 

Author Comment

by:nsome
ID: 16701294
I am planing on using veritas cluster in this configuration.  How does disk suite play in this configuration or is it negligible?
0
 
LVL 11

Assisted Solution

by:jekl2000
jekl2000 earned 400 total points
ID: 16702506
We use both. Volume manager for local disk and VXFS for SAN disk. We have also gotten away from hard partitioning on local drives. We have a slice 2 for the backup and one large slice 7 the size of the drive and use soft partitions. It has made life much easier. Volume manager is a nice product for free.
0
 

Author Comment

by:nsome
ID: 16702542
Why are you using VXFS for SAN disk only and not  solaris volume manager?
0
 
LVL 21

Expert Comment

by:tfewster
ID: 16702775
> I take it that a few of you use or like Veritas vxfs / volume manager?

Oops, are my prejudices showing? ;-)  I like the AIX & HP-UX integrated version of Veritas tools, but the add-on package (on any OS) feels just that - An add-on.

I try to stick to the policy of reserving "internal" disks for the OS & standard software stack (e.g backup, scheduling & monitoring software), leaving a good chunk in reserve. The real apps + data go on separate disks - SAN if it's available and justifiable. As apps & data tend to grow, Veritas is a good tool for managing those filesystems.  But keep the OS setup simple, as the extra layer of complication with Veritas can make recovery from OS problems more difficult.

If you're setting up a cluster, I'd say vxvm is ideal for filesystems that will be switched to another server as it can import disk groups that were not exported cleanly (e.g from a crashed server); Other volume managers always seem to have a issue over which server "owns" the disk (I could be wrong, Veritas Cluster might make all that straightforward)
0
 
LVL 11

Expert Comment

by:jekl2000
ID: 16703750
>Why are you using VXFS for SAN disk only and not  solaris volume manager?

Due to the large volume of disk we have basically. Plus we use cluster and dmp for failover.
Its easier to manage a large number of disk with Veritas.

We jumpstart most of our servers and have the Volume manager stuff preconfigured for the local disk.
0
 
LVL 14

Expert Comment

by:arthurjb
ID: 16703793
>We have also gotten away from hard partitioning on local drives.

This is like a chevy/ford debate, since I can give you alot of reasons why the hard partitioning is the best thing for a server.

First, soft partitioning adds cpu cycles to the mix, not many but still some.

I think that the boot disk should be partitioned into / /usr /var /opt  then the rest used for whatever  then that disk should be mirrored.

I have used soft partitiong for large RAIDs or mirrors for the data, but not for the os (boot) disk...

vxfs for a large san that may need to be moved to another host is possibly a good call, but in the case of a single server...

Not to be nosy, but what is your reason for adding the overhead of vxfs?  As you can see most of the experts here find it unnecessary.
0
 

Author Comment

by:nsome
ID: 16705696
First of all - fantastic input from all of the experts!

This will be a 2-3 node cluster, which per a few of the experts, may be a good call to use VXFS for the shared data and disksuite for local systems and disk group importing / deporting may be "smoother" under Veritas as one expert indicated.  Also there won't be many disks in the shared storage - half a dozen or so...does it still necessitate VXFS to alleviate the potential disk group import problem after a failover as mentioned above or will disksuite handle it like a champ?  

jekl2000 indicated "We have also gotten away from hard partitioning on local drives. " whereas arthurjb believes that hard partitioning is better for a system disk.

Which one is best call and why?

0
 
LVL 14

Accepted Solution

by:
arthurjb earned 400 total points
ID: 16705889
>Which one is best call and why?

There is really no "best", but here is why I set up a file system in a partitioned  way;

On a properly set up solaris system, much of the file system doesn't grow beyond its original size.   Having the sections that may grow on seperate partitions allows you to monitor the health of the system.

If your root partition fills up, your machine will crash.  The var filesystem is the one most likely to grow for various reasons, mostly because the logs start filling up.  the usr partition can also grow becasuse of /usr/tmp and /usr/spool.

If var fills up, it may still crash the system, but it usually does not, and when you delete the ofending files the system quickly returns to normal.

If /usr fills up, same thing, it may crash, but it usually doesn't.

By seperating out the /var and /usr you can quickly see which part of the system is causing the disk to fill.  If you do what seems to have become more common lately, and just take 8-10gigs for the os and 2-4 gigs for swap, then you may miss a problem that is filling up the logs, and you will have to use du or some other useage tool to figure out where all the disk space has gone.  Running these tools takes a lot of time on a large files system, while a df -lh takes almost no time.

I also like to design the system for worst case operations, so I try to use the lowest overhead and such.  Using a mirrored pair for the boot disk gives you two independant disks that you can use in a pinch if you have some sort of a disaster operation.  

If you keep good nots in the vfstab, as to the identity of the raw partitions, a quick change in the vfstab and a reboot can have you booting off either of the mirrors, without needing to use any disk management tools...  Not a common thing, I've only had to do it once in the last 15 years, but it sure made things a lot easier.  

0
 
LVL 11

Expert Comment

by:jekl2000
ID: 16708940
The main reason we went to soft partitions was the ability to expand /usr /var /opt and other partitions easily. We also have alot of request for space for this or that on boxes that are not san attached, plus certain products that require their own partition. I like it because of the capability to change things easily. Granted, there is nothing wrong with hard partitioning, have done it for years. I am anxiously awaiting implementation of vfs which could make Veritas and Volume manager obsolete.

 
0
 

Author Comment

by:nsome
ID: 16751316
Upon initial installation of Solaris, hard partitions are defined.  In my case I have two disks, c0t0d0 & c1t0d0.  On disk 1 c0t0d0, the / , /usr, /opt and /export/home partitions were created.  I want to be able to mirror this configuration onto the second disk c1t0d0.

I need some clarification / contrasting when doing this with Solaris Volume Manager with respect to how different would these steps be when doing the same task with AIX LVM where a volume group consisting of physical volumes is configured with logical volumes and corresponding filesystems...?

0
 
LVL 21

Expert Comment

by:tfewster
ID: 16755066
The principle is the same:
- The disk is partitioned ("hard" slices" in SVM, "soft" volumes in LVM);
- A special device file is created that addresses the "metadevice"  (the Volume Manager interprets calls to this device file and handles reads/writes to the physical device[s], e.g. mirrors, stripes, volumes spanning disks)
- The filesystem is created on the metadevice. Both ufs & jfs support journalling, growing filesystems etc. so the facilities are similar.

Consider that LVM (HP-UX and AIX) are based on Veritas Volume Manager,  but the actual commands used for those 3 volume managers appear very different as each manufacturer has rewritten the tools to fit with their own "style"; SVM does the same job, it just feels like an add-on product.
0
 

Author Comment

by:nsome
ID: 16786098
I'm still in need of connecting the dots from initiall installation and definition of slices /, /opt, /var, /export/home  as portions of the system disk to bringing the system under Solaris VM control in order to mirror this disk with another disk.  What happens in this case?  Do system partitions become volumes under volume control?  Can someone explain the magic here?

0
 

Author Comment

by:nsome
ID: 16848224
Comment from tfewster
Date: 05/17/2006 12:13AM PDT
      Comment       Accept

I notice you've used slice 2 for a filesystem, wheras normal convention is leave it as the whole disk. Personally, I'd do:
s0 / 1Gb
s1 swap  4Gb (Or = memory?)
s2 (whole disk)
s3 /var
s4 /usr
s5
s6 Remaining space - Create soft partitions for /export/home
s7 Disksuite metadb  100Mb


For s6, do I need to leave that slice alone and not hard partition it?  

Also for s7 do you mean to create a 100 mb hard partition for the metadb itself?
0
 
LVL 21

Expert Comment

by:tfewster
ID: 16849690
Based on the actual filesystem usage in your original post, the following should work for you:

s0 / 1Gb
s1 swap  4Gb (Or = memory?)
s2 (whole disk)
s3 /var  2Gb
s4 /usr  3Gb
s5 /opt  3GB
s6 Remaining space (Approx 52GB  ) - Create soft partitions for /export/home, /opt/hugepackage, /backup_software
    etc. as needed,  giving you maximum flexibility
s7 Disksuite metadb  100Mb

s6 needs to be a hard partition at low level, but when brought under SVM control you can create soft partitions within that metadevice

Yes, s7 should be a hard partition, reserved for the metadbs
0
 

Author Comment

by:nsome
ID: 16851702
For s1 swap, you indicate 4GB or = memory;  the box has 16 GB memory - should I make it 16 GB?  When do you draw the line?
0
 
LVL 14

Expert Comment

by:arthurjb
ID: 16855333
That is a good question, I have not dealt with a system with that much memory, but the rule of thumb in the past was 2xmemory.
0
 

Author Comment

by:nsome
ID: 16855522
32 GB???? Does that rule still apply?
0
 
LVL 14

Expert Comment

by:arthurjb
ID: 16855580
Probably not, but I don't know the real answer.

Because this question is old, you may have to post it as a seperate question, "How much swap space for a sun with16g of memory?"  to catch the eyes of other experts...
0
 

Author Comment

by:nsome
ID: 16855628
yeah...
0
 
LVL 21

Assisted Solution

by:tfewster
tfewster earned 800 total points
ID: 16856740
If I recall correctly, the 2 x memory guideline was to allow you to make up for lack of real memory by using swap (disk) space, and the system wouldn't start processes unless it could allocate swap space for it to be paged out to.

Even though the paging algorithms have been made much smarter over time, as memory is much cheaper now and paging is several orders of magnitude slower than using real memory, it's to be avoided if possible. Though you would still need to allocate 16Gb of swap space if you wanted to be able to capture a full memory dump in case of a crash...
0
 
LVL 38

Expert Comment

by:yuzh
ID: 16858018
The size of the swap space is depands what appliaction software is running on the box (the
requirements of the system's software applications).

But you should allocate at least as the SAME size of the RAM for your system, to enable to save a worst-case crash dump.

When the system is using swap, it will slow down, if you are positive that you have enough memory to run your application, you can make swap smaller but
(>= memory size)

Have a look at the following page to learn more about Solaris swap space:
http://www.itworld.com/Comp/2378/swol-0496-perf/

http://www.alise.lv/ALISE/technolog.nsf/0/59136f9072dc58d8422569fa0057b095?OpenDocument
0
 

Author Comment

by:nsome
ID: 16858197
I have configured it with 16gb to reflect maximum memory amount.  
0
 

Author Comment

by:nsome
ID: 16864486
I've asked another related question and hoping for the experts here to look at it as well.

http://www.experts-exchange.com/Operating_Systems/Solaris/Q_21879432.html
0

Featured Post

Upgrade your Question Security!

Add Premium security features to your question to ensure its privacy or anonymity. Learn more about your ability to control Question Security today.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Hello fellow BSD lovers, I've created a patch process for patching openjdk6 for BSD (FreeBSD specifically), although I tried to keep all BSD versions in mind when creating my patch. Welcome to OpenJDK6 on BSD First let me start with a little …
This tech tip describes how to install the Solaris Operating System from a tape backup that was created using the Solaris flash archive utility. I have used this procedure on the Solaris 8 and 9 OS, and it shoudl also work well on the Solaris 10 rel…
Learn how to navigate the file tree with the shell. Use pwd to print the current working directory: Use ls to list a directory's contents: Use cd to change to a new directory: Use wildcards instead of typing out long directory names: Use ../ to move…
This video shows how to set up a shell script to accept a positional parameter when called, pass that to a SQL script, accept the output from the statement back and then manipulate it in the Shell.
Suggested Courses
Course of the Month15 days, 12 hours left to enroll

850 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question