Today is notable in the history of OpenSolaris/Solaris development. Today the source code for ZFS, Sun’s next generation filesystem, was published. Today, binaries of Nevada build 27 became available so you can try ZFS yourself. And some night soon, we hope, you’ll sleep better knowing that your data is safe and sound because ZFS is protecting it.

Anyway, I want to talk about the marvelous integration work Eric did between ZFS and Zones. In ZFS, storage is pooled, and you can create one or many filesystems from the pool. Each filesystem acts as a container for those below it: filesystem properties set on a parent are inherited to children. Even better, you can allocate one or many of these containers to a zone– and then the zone can manage that container for itself, creating new child filesystems. This is an incredibly powerful form of delegated administration and I think it’s a great example of how our suite of technologies can be converged.

Here’s an example I did on my test system using a very small slice on a disk (the test box
happened to have only one disk):

trolls # zpool create mypool c0d0s3
trolls # zpool list
NAME                    SIZE    USED   AVAIL    CAP  HEALTH     ALTROOT
mypool                 95.5M   32.5K   95.5M     0%  ONLINE     -
trolls # zfs create mypool/myzone-data
trolls # zfs set compression=on mypool/myzone-data
trolls # zfs set quota=30m mypool/myzone-data

So at this point, we’ve created a storage pool, enabled compression, and made sure that
myzone-data will never grow larger than 30MB in size. Now, we’ll add that dataset
to a zone on the system:

trolls # zonecfg -z myzone
zonecfg:myzone> add dataset
zonecfg:myzone:dataset> set name=mypool/myzone-data
zonecfg:myzone:dataset> end
zonecfg:myzone> \^D
trolls #
trolls # zoneadm -z myzone boot

Ok! Let’s now look inside the zone:

trolls # zlogin myzone
[Connected to zone 'myzone' pts/3]
Last login: Fri Nov  4 17:49:18 on console
Sun Microsystems Inc.   SunOS 5.11      onnv-gate:2005-11-07    October 2007
myzone # zfs list
NAME                   USED  AVAIL  REFER  MOUNTPOINT
mypool                45.0K  79.5M     8K  /mypool
mypool/myzone-data       8K  30.0M     8K  /mypool/myzone-data
myzone #
myzone # zfs create mypool/myzone-data/project1
myzone # zfs set mountpoint=/project1 mypool/myzone-data/project1
myzone # zfs set quota=10m mypool/myzone-data/project1

Cool: the zone adminstrator can further quota and subdivide the storage. Did
that work?

myzone # cat /dev/urandom > /project1/test
cat: output error (0/1040 characters written)
Disc quota exceeded
myzone # ls -lh /project1/
total 20492
-rw-------   1 root     root       10.0M Nov 15 13:28 test

As you can see, you can use ZFS to give a great deal of administrative freedom
to the administrator inside of zone without providing access to raw devices.
Even better, the tools offer one stop shopping for creating and managing filesystems and their properties.
Technorati Tag: Solaris