ZFS: Snapshots and clones on zfs filesystems
Article Number: 185 | Rating: Unrated | Last Updated: Sun, Jun 3, 2018 9:13 AM
ZFS: Snapshots and clones on zfs filesystems
|
# Tested on RHEL 6 & 7
# A snapshot is an only-read photograph of a filesystem. When taking a snapshot, it is # stored in a way so further transactions on filesystem will only be carried out on origin # filesystem and not on snapshot itself. This way it will be possible to get back to # previous status by doing a "rollback" # A clone is equivalent to a read-write copy of the snapshot # Clones and snapshots are not data copies but state ones, so they don't use any space when # created. It is when origin filesystem is modified when differences are being stored, thus # consuming disk space. If a rollback is done, these differences are overwritten and space # is freed-up again. # Note: Clones may be created only from existing snapshots. First we take the "photo" of # the origin filesystem and then we create the clone. # Snapshots are very useful, for instance to carry out tests without being afraid of losing # important data. # Given following zfs zfs list NAME USED AVAIL REFER MOUNTPOINT c_pool 2.15M 3.84G 19K /c_pool c_pool/zfs01 2.02M 3.84G 2.02M /zfs01 <--- # Create a snapshot from a ZFS # ------------------------------------------------------------------------------------------ zfs snapshot c_pool/zfs01@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 2.50M 3.84G 19K /c_pool c_pool/zfs01 2.02M 3.84G 2.02M /zfs01 c_pool/zfs01@snapshot01 0 - 2.02M - <--- # Rollback a ZFS to a previous state # ------------------------------------------------------------------------------------------ # First we do some modifications on filesystem cd /zfs01 dd if=/dev/urandom of=temp.file.01 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 24.7394 s, 84.8 kB/s dd if=/dev/urandom of=temp.file.02 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 25.3346 s, 82.8 kB/s ls -lrt total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 17:10 temp.file.01 -rw-r--r-- 1 root root 2097152 Feb 3 17:11 temp.file.02 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.18M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 19K - <---- note the differences # and then, rollback zfs rollback c_pool/zfs01@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 176K 3.84G 19K /c_pool c_pool/zfs01 20K 3.84G 19K /zfs01 c_pool/zfs01@snapshot01 1K - 19K - <--- # files have disappeared: ls -lrt total 0 # Remove a snapshot # ------------------------------------------------------------------------------------------ zfs destroy c_pool/zfs01@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 174K 3.84G 19K /c_pool c_pool/zfs01 19K 3.84G 19K /zfs01 # To have different points of restoration, several snapshots may be taken at different times # ------------------------------------------------------------------------------------------ dd if=/dev/urandom of=temp.file.00 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 25.2053 s, 83.2 kB/s ls -lrt total 2051 -rw-r--r-- 1 root root 2097152 Feb 3 16:57 temp.file.00 zfs snapshot c_pool/zfs01@snapshot01 dd if=/dev/urandom of=temp.file.01 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 25.9611 s, 80.8 kB/s ls -lrt total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 16:57 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 16:58 temp.file.01 zfs snapshot c_pool/zfs01@snapshot02 dd if=/dev/urandom of=temp.file.02 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 25.5691 s, 82.0 kB/s ls -lrt total 6152 -rw-r--r-- 1 root root 2097152 Feb 3 16:57 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 16:58 temp.file.01 -rw-r--r-- 1 root root 2097152 Feb 3 16:59 temp.file.02 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 6.17M 3.84G 19K /c_pool c_pool/zfs01 6.04M 3.84G 6.03M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - c_pool/zfs01@snapshot02 9K - 4.02M - # If we try to rollback to oldest snapshot: zfs rollback c_pool/zfs01@snapshot01 cannot rollback to 'c_pool/zfs01@snapshot01': more recent snapshots or bookmarks exist use '-r' to force deletion of the following snapshots and bookmarks: c_pool/zfs01@snapshot02 # If we need to rollback to first snapshot, first we have to rollback to the newer one, # destroy it and, then, rollback to the oldest snapshot zfs rollback c_pool/zfs01@snapshot02 ll total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 16:57 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 16:58 temp.file.01 zfs rollback c_pool/zfs01@snapshot01 cannot rollback to 'c_pool/zfs01@snapshot01': more recent snapshots or bookmarks exist use '-r' to force deletion of the following snapshots and bookmarks: c_pool/zfs01@snapshot02 zfs destroy c_pool/zfs01@snapshot02 zfs rollback c_pool/zfs01@snapshot01 ls -lrt total 2051 -rw-r--r-- 1 root root 2097152 Feb 3 16:57 temp.file.00 # Otherwise, we could have used '-r' option to recursively rollback to desired snapshot. # This will destroy all intermediate snapshots. zfs rollback -r c_pool/zfs01@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 2.15M 3.84G 19K /c_pool c_pool/zfs01 2.02M 3.84G 2.02M /zfs01 c_pool/zfs01@snapshot01 1K - 2.02M - # Displaying snapshots # ------------------------------------------------------------------------------------------ zfs list -t snapshot NAME USED AVAIL REFER MOUNTPOINT c_pool/zfs01@snapshot01 1K - 2.02M - # Accessing snapshot contents # ------------------------------------------------------------------------------------------ # As long as zfs "snapdir" property is set to "visible", snapshot's contents should be # accessible by entering ".zfs" directory under zfs's mount point. # There should be one directory per snapshot containing directory/file structures existing # at the moment snapshot was taken zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 8.15M 9.62G 19K /c_pool c_pool/zfs01 8.05M 9.62G 8.03M /zfs01 c_pool/zfs01@snapshot01 10K - 4.02M - c_pool/zfs01@snapshot02 11K - 6.03M - zfs get all c_pool/zfs01 | grep snapdir c_pool/zfs01 snapdir visible local cd /zfs01/.zfs/snapshot ls -l total 1 drwxr-xr-x. 2 root root 3 Feb 3 21:55 snapshot01 drwxr-xr-x. 2 root root 4 Feb 3 21:55 snapshot02 # Each of the directories contains directory/file structures existing at the moment when # snapshot was taken: ls -lR .: total 1 drwxr-xr-x. 2 root root 3 Feb 3 21:55 snapshot01 drwxr-xr-x. 2 root root 4 Feb 3 21:55 snapshot02 ./snapshot01: total 2051 -rw-r--r--. 1 root root 2097152 Feb 3 21:51 temp.file.01 ./snapshot02: total 4101 -rw-r--r--. 1 root root 2097152 Feb 3 21:51 temp.file.01 -rw-r--r--. 1 root root 2097152 Feb 3 21:55 temp.file.02 # I experienced some troubles while trying to access snapshot contents on virtual systems # (both VMWare and Oracle VM VirtualBox virtual systems). For the moment, I'll let it drop # because virtual servers are not the main target of this procedure.
# Cloning a snapshot # ------------------------------------------------------------------------------------------ # Let's create one zpool, with one zfs and one snapshot zpool create c_pool sdb sdc zfs create -o mountpoint=/zfs01 c_pool/zfs01 cd /zfs01 dd if=/dev/urandom of=temp.file.00 bs=1M count=2 zfs snapshot c_pool/zfs01@snapshot01 dd if=/dev/urandom of=temp.file.01 bs=1M count=2 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.21M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - zfs clone c_pool/zfs01@snapshot01 c_pool/zfs02 # Snapshot c_pool/zfs01@snapshot01 has been copied and will be writeable on # c_pool/zfs02 clone zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.22M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - c_pool/zfs02 1K 3.84G 2.02M /c_pool/zfs02 ls -lrt /zfs01 total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.01 ls -lrt /c_pool/zfs02 total 2051 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 dd if=/dev/urandom of=/c_pool/zfs02/temp.file.02 bs=1M count=2 2+0 records in 2+0 records out 2097152 bytes (2.1 MB) copied, 25.0426 s, 83.7 kB/s ls -lrt /c_pool/zfs02 total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 17:17 temp.file.02 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 6.23M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - c_pool/zfs02 2.01M 3.84G 4.02M /c_pool/zfs02 # Removing a clone/snapshot # ------------------------------------------------------------------------------------------ zfs destroy c_pool/zfs02 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.22M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - # Note: If a snapshot has one or more clones we won't be able to destroy it unless clones # are destroyed first: zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.23M 3.84G 19K /c_pool c_pool/zfs01 4.03M 3.84G 4.02M /zfs01 c_pool/zfs01@snapshot01 9K - 2.02M - c_pool/zfs02 1K 3.84G 2.02M /c_pool/zfs02 zfs destroy c_pool/zfs01@snapshot01 cannot destroy 'c_pool/zfs01@snapshot01': snapshot has dependent clones use '-R' to destroy the following datasets: c_pool/zfs02 zfs destroy c_pool/zfs02 zfs destroy c_pool/zfs01@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.15M 3.84G 19K /c_pool c_pool/zfs01 4.02M 3.84G 4.02M /zfs01 # Promoting a clone # ------------------------------------------------------------------------------------------ # Once a clone in place, we can use to replace original dataset. We will make clone # independent of the snapshot it was created from and, then, remove snapshot(s) and # origin filesystem so our clone will replace it zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.16M 3.84G 19K /c_pool c_pool/product 4.02M 3.84G 4.02M /product <---- ll /product total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 17:20 temp.file.01 zfs snapshot c_pool/product@snapshot01 zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.16M 3.84G 19K /c_pool c_pool/product 4.02M 3.84G 4.02M /product c_pool/product@snapshot01 0 - 4.02M - <---- zfs clone -o mountpoint=/clone c_pool/product@snapshot01 c_pool/clone zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.19M 3.84G 19K /c_pool c_pool/clone 1K 3.84G 4.02M /clone <---- c_pool/product 4.02M 3.84G 4.02M /product c_pool/product@snapshot01 0 - 4.02M - # Make some modifications to clone (this is clone's purpose indeed) vi /clone/mynewfile [...] ll /product /clone /product: total 4101 -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 17:20 temp.file.01 /clone: total 4102 -rw-r--r-- 1 root root 20 Feb 3 17:23 mynewfile -rw-r--r-- 1 root root 2097152 Feb 3 17:15 temp.file.00 -rw-r--r-- 1 root root 2097152 Feb 3 17:20 temp.file.01 # Promote the clone zfs promote c_pool/clone # Among other things, the existing snapshot becomes dependent of the clone that has been # promoted. # Take a look to new "USED" space value for the clone too (It's not a copy of the clone # anymore but an independent dataset) zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.55M 3.84G 19K /c_pool c_pool/clone 4.03M 3.84G 4.02M /clone <---- c_pool/clone@snapshot01 9K - 4.02M - <---- c_pool/product 0 3.84G 4.02M /product # If we try, for instance, to remove the promoted clone we won't be able because now # it has a dependent snapshot: zfs destroy c_pool/clone cannot destroy 'c_pool/clone': filesystem has children use '-r' to destroy the following datasets: c_pool/clone@snapshot01 # Should we need a current snapshot of promoted clone, we have to create a new one because # the existing one is a snapshot from the original contents # Now we are ready to replace the original dataset with the new one (promoted clone). # Take into account that to be able to rename mountpoints (if needed) we'll have to # remount the datasets zfs rename c_pool/product c_pool/product.orig zfs get all c_pool/product.orig | grep mountpoint c_pool/product.orig mountpoint /product local zfs set mountpoint=/product.orig c_pool/product.orig # On RHEL 7 F.S. has been already mounted so following two lines are not necessary: mkdir /product.orig zfs mount c_pool/product.orig zfs list NAME USED AVAIL REFER MOUNTPOINT c_pool 4.62M 3.84G 19K /c_pool c_pool/clone 4.03M 3.84G 4.02M /clone c_pool/product.orig 9K 3.84G 4.02M /product.orig zfs rename c_pool/clone c_pool/product zfs set mountpoint=/product c_pool/product # On RHEL 7 F.S. has been already mounted so following lines is not necessary: zfs mount c_pool/product zfs list -t all NAME USED AVAIL REFER MOUNTPOINT c_pool 4.24M 3.84G 19K /c_pool c_pool/product 4.03M 3.84G 4.02M /product c_pool/product@snapshot01 9K - 4.02M - c_pool/product.orig 9K 3.84G 4.02M /product.orig |
Subscribe to Article
Print Article
Email Article to Friend
Export to PDF
Export to MS Word