zfs discuss - Feb 2010 - mimic size=1024m setting in /etc/vfstab when using rpool/swap (zfs root)

Hi,

While using ufs root, we had an option for limiting the /tmp size using mount -o
size manual option or setting size=1024m in the vfstab.

Do we have any comparable option available when we use zfs root. If we execute 
zfs set size=1024m rpool/swap
it resizes the whole of the swap area which results in reducing the VM size. 

AFAIK in the ufs option for limiting swap size, we are just limiting the
filesystem behaviour of the tmpfs filesystem while the kernel will still be able
to use the swap space for paging etc. Is this correct understanding.

Thanks in advance. 

Rgds
PK
-- 
This message posted from opensolaris.org

On Jan 31, 2010, at 10:15 PM, Prakash Kochummen wrote:
> Hi,
> 
> While using ufs root, we had an option for limiting the /tmp size using
mount -o size manual option or setting size=1024m in the vfstab.
This is no different when using ZFS root.  /tmp is, by default, a tmpfs file
system.
> Do we have any comparable option available when we use zfs root. If we
execute
> zfs set size=1024m rpool/swap
> it resizes the whole of the swap area which results in reducing the VM
size.
Yes, but the equivalent for a UFS root would be to format the swap device and
change the partition size.  Clearly, the ZFS solution is simpler and safer.
> AFAIK in the ufs option for limiting swap size, we are just limiting the
filesystem behaviour of the tmpfs filesystem while the kernel will still be able
to use the swap space for paging etc. Is this correct understanding.
By default, the Solaris 10 installer does not use UFS for swap.  So there is no
"ufs option for limiting swap size."
 -- richard

Thanks for the reply.

Sorry i confused you too. when I mentioned ufs , i just meant ufs root scenario
(pre u6).

Suppose I have a 136G Hdd which as my boot disk,which has been sliced it like
s0-80gb (root slice)
s1-55Gb (swap slice) 
s7- (SVM metadb) 

My understanding was that if I had 16Gb of memory, my "df -h /tmp"
output should be something like
tmp=16+55 - (memory used by solaris)

And if I specify size=512m in the /etc/vfstab file for swap, my /tmp would still
show $tmp size instead of 512m but I would not be able to copy an 700mb iso into
/tmp.

Hope I am clear. 

Thanks
-- 
This message posted from opensolaris.org

On Jan 31, 2010, at 11:24 PM, Prakash Kochummen wrote:
> Thanks for the reply.
> 
> Sorry i confused you too. when I mentioned ufs , i just meant ufs root
scenario (pre u6).
> 
> Suppose I have a 136G Hdd which as my boot disk,which has been sliced it
like s0-80gb (root slice)
> s1-55Gb (swap slice) 
> s7- (SVM metadb) 
> 
> My understanding was that if I had 16Gb of memory, my "df -h
/tmp" output should be something like
> tmp=16+55 - (memory used by solaris)
> 
> And if I specify size=512m in the /etc/vfstab file for swap, my /tmp would
still show $tmp size instead of 512m but I would not be able to copy an 700mb
iso into /tmp.
ZFS does not change tmpfs.
 -- richard

my bad, since zfs/zfs root, never really bothered to look at the /etc/vfstab
file, looked at it now and everything is same as before .

Thanks a lot for your reply.

Rgds
PK
-- 
This message posted from opensolaris.org