what is the typical Size for slices of /,/opt,/var??? 3

[StraightL]

What is typical recommendation for .
/
/usr
/usr/openwin ( is it necessary to be a separate slice)?
/opt
/var
/tmp
swap

because my recent install of VM failed for limited size of "/"

I want to know the rule of thumb

thanks!

 

[crasho2001]

Hi,

You can see one example from running system below.It is solaris 8.

tests> df -k
Filesystem kbytes used avail capacity Mounted
/dev/dsk/c0t0d0s0 191611 77878 111817 42% /
/dev/dsk/c0t0d0s3 1489367 817009 657465 56% /usr
/dev/dsk/c0t0d0s4 8729385 158731 8483361 2% /var
/dev/dsk/c0t0d0s5 14882455 4588068 10145563 32% /opt

You have nothing to do with /tmp. As a recommendation you should use /dev/dsk/c0t0d0s1 as a swap area and size should be 2x "physical memory" or more.

Actually , you can define only / . You do not need to define usr,var,opt seperately. You can do very big partition of /.
/opt and /var depends on your software. For my software /opt and /var requires very big disk partition.
I do not think also you need to define /usr/openwin seperately.

Regards

 

[daFranze]

there are two different partition "religions", the "one slice for all OS data" and the "separate slice for each filesystem"

I prefere the "one slice"; my "/" Filesystem is 5-6Gigs swap depends on the applications but rule-of-thumb is 2xRAM (on some hosts we have 5xRAM); mountoption should be "logging", this will speed up fsck in case of a crash

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[StraightL]

one slice for all OS data" is very nice for Solaris ,since future extend is impossible.
But what is the disadvantage?

 

[daFranze]

you just have one Filesystem, if your logfiles grow fast and fills up your filesystem....
if your system crashes and fsck can't fix the FS you have to restore the whole FS, not just a few pieces; large filesystems take some time to run fsck (if there is no "clean" flag set - the system was halted or crashed)

Best Regards, Franz
--
Solaris System Manager from Munich, Germany
I used to work for Sun Microsystems Support (EMEA) for 5 years

 

[spamly]

The "two times physical memory" concept for swap space is no longer really needed. It used to be the rule of thumb because prior to putting a process into memory, Solaris would go to swap space and reserve enough space to swap it out in the future. Things have changed and Solaris (8 for sure, possibly before...) doesn't do this. It will now go out and reserve space in swap when it needs to, not before. Makes things more efficient.

Anyway, the general rule of thumb now is to have enough to perform a full system dump (in the event of a panic), plus a little extra for legitimate. Yeah, not exactly a perfect formula, but if you're really using swap space heavily you've got issues.

Hope this helps.