ZFS Basics Demo (12 minutes)

Description

This demo showcases some of the cool features of ZFS - easy administration, snapshots, backups, rollback, working with pools and mirrors.

OpenSolaris Versions Supported

OpenSolaris 2008.05 or higher.

Points to Hit

ZFS is very cool
ZFS is easy to administer
Snapshots are cheap, fast and easy to create
You can backup snapshots including incremental snapshots
Rollbacks also take less than a second independent on data size
Working with storage pools is easy
You can create a mirror in few seconds
You can grown storage pools easily
More reasons for ZFS: http://opensolaris.org/os/community/zfs/docs/zfs_last.pdf

Demo Prep

Using VirtualBox's Virtual Disk Manager, create 4 dynamically expanding 1GB disks (ZFSDemo1.vdi, ZFSDemo2.vdi, ZFSDemo3.vdi and ZFSDemo4.vdi). They will be used for the RAID. The type should be SATA.

Optionally, since this demo uses ls, configure ls colors.

Note, this demo could also be done using USB sticks, however, I prefer to use a virtual image as it presents "cleaner" output. For example, as I have Time Slider enabled, the output from 'zfs list -t all' is harder to weed though.

Gotchas

None known.

Demo

Open a terminal and enlarge it (Ctrl Shift +)
pfexec su to avoid having to using pfexec.

Working with Pools

We start by looking at our existing pool and cover the steps required to create a new pool. The new pool will be a mirrored pool. First, look at the existing pool:

Run zpool list to see the existing pool (on a default install, this will be rpool)
Run zpool status to see the health status of the pool

Let's create a new mirrored pool:

Run format to get the device names. Note, c = controller number, t = target number, d = logical unit number (or ask the audience if they remember what command to run to get the device names).
Run zpool create -f demo mirror device1 device2.
Run zpool list and note the 1GB available.
Run zpool status demo to view the disks in the new pool.
Run zpool add -f demo mirror device3 device4.
Run zool list and show the available size has been doubled.
Run zpool status demo to view the disks added to the pool.
Mention other features: clones, export/import, compression, encryption, quotas and raidz. For details about these features see the ZFS preso: http://opensolaris.org/os/community/zfs/docs/zfs_last.pdf

Working with File Systems and Snapshots

Note, as part of the pool creation process, a file system was also created and mounted. Not bad - one command to do all that.

Run zfs list to show the file systems, including the new demo file system that was just created. Explain that the first column is the dataset name, comprised of the <pool name>/<file system name>. Explain that the last column is the mount point. Explain why all active file systems have the same amount of GB available (because they use the same storage pool). This could also be posed as a question to the audience, to make sure they were paying attention during your presentation .
Create a new FS by running zfs create demo/myfs. By default, this will mount the myfs directory at (/demo/myfs). Alternatively, you can show that it is possible to mount the directory anywhere, independent of the pool structure, using the mountpoint option, e.g., zfs create -o mountpoint=/export/myfs demo/myfs.
Note how fast it was to create the new file system - you can create file systems like directories.
Run zfs list to show the new FS.
Run cd /demo/myfs and ls. The directory is empty.
Run touch file1.txt. Edit it's contents to say something.
Run mkfile 100M large-file. Then run ls. Mention no formatting was necessary (as is required for all other new file systems before they can be used). Optionally, edit the file's contents: "Before snapshot".
Run zfs snapshot demo/myfs@today.
Run zfs list -t all | grep demo to see the snapshot that was created.
Run touch file2.txt and rm file1.txt (or edit it's contents: "After snapshot"). Then run ls.
Run zfs list -t all | grep demo. Note the snapshot isn't taking up any space. This is because all of its contents are still intact (file1.txt).
Run zfs rollback demo/myfs@today. Then run ls. Only file1.txt should be there. Mention advantages of instantaneous snapshots and rollbacks.
Run zfs list. Note the snapshot is using 0K again.

Working with Large Snapshots

Tell the audience that they might not believe you that the snapshots are so fast because we are just working with empty files. So let's try something brutal - delete the /opt directory.

As of 2008.11, the opt directory is no longer a separate file system.
Run zfs snapshot rpool/ROOT/opensolaris@today.
Run du -hs /opt to show it's current size.
Run cd /opt and ls.
Delete all files in /opt by running rm -rf *. Run *ls*.
Run zfs list. Note how the snapshot is now using several hundred megabytes.
Run zfs rollback rpool/ROOT/opensolaris@today. Run ls. Show that all files have been recovered. Mention the operation that took the longest time was deleting the files.

Backup and Restore

What if I want to make a backup and move my file system as a whole?

Run zfs send demo/myfs@today > /tmp/zfsdata.
Run zfs destroy -r demo/myfs. Explain the -r option recursively destroys all children, including snapshots.
Run zfs list.
Run ls /demo/myfs.
Run zfs receive -d demo < /tmp/zfsdata. The -d option also causes the snapshot to be created.
Run ls /demo/myfs.
Run zfs list -t all | grep demo

Clones

What if I want to create a copy of my file system?

Run zfs clone demo/myfs@today demo/myfsclone
Run cd /demo/myfsclone
Run ls -lh

RAID-Z

Mention that RAID-Z (Redundant Array of Inexpensive Disks) is similar to RAID-5 but it's a bit better because it checksums on multiple disks, so data security is higher and performance is also better thanks to organizing data in a way that the disks are well harnessed (less going there and back to get data).

Run zpool list.
Run format to get the device names. Note, c = controller number, t = target number, d = logical unit number
Let's show raidz. Run zpool create demo raidz device1 device2 device3.
Run zpool list
Run zpool status
Run zfs create demo/myfs
Run cp -r /opt /demo/myfs
Run zpool status
Run zpool destroy demo.

Demo Cleanup

Delete the pool and filesystems created during the demo:

zpool destroy demo
zfs destroy rpool/ROOT/pensolaris/opt@today

ZFSBasicsDemo