Simple bash script to rebalance pool data between all mirrors when adding vdevs to a pool.
This script recursively traverses all the files in a given directory. Each file is copied with a .balance
suffix, retaining all file attributes. The original is then deleted and the copy is renamed back to the name of the original file. When copying a file ZFS will spread the data blocks across all vdevs, effectively distributing/rebalancing the data of the original file (more or less) evenly. This allows the pool data to be rebalanced without the need for a separate backup pool/drive.
The way ZFS distributes writes is not trivial, which makes it hard to predict how effective the redistribution will be. See:
- https://jrs-s.net/2018/04/11/zfs-allocates-writes-according-to-free-space-per-vdev-not-latency-per-vdev/
- https://jrs-s.net/2018/08/24/zfs-write-allocation-in-0-7-x/
Note that this process is not entirely "in-place", since a file has to be fully copied before the original is deleted. The term is used to make it clear that no additional pool (and therefore hardware) is necessary to use this script. However, this also means that you have to have enough space to create a copy of the biggest file in your target directory for it to work.
At no point in time are both versions of the original file deleted.
To make sure file attributes, permissions and file content are maintained when copying the original file, all attributes and the file checksum is compared before removing the original file (if not disabled using --checksum false
).
Since file attributes are fully retained, it is not possible to verify if an individual file has been rebalanced. However, this script keeps track of rebalanced files by maintaining a "database" file in its working directory called rebalance_db.txt
(if not disabled using --passes 0
). This file contains two lines of text for each processed file:
- One line for the file path
- and the next line for the current count of rebalance passes
/my/example/pool/file1.mkv
1
/my/example/pool/file2.mkv
1
To check the current balance of a pool use:
> zpool list -v
NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
bpool 1.88G 113M 1.76G - - 2% 5% 1.00x ONLINE -
mirror 1.88G 113M 1.76G - - 2% 5.88% - ONLINE
ata-Samsung_SSD_860_EVO_500GB_J0NBL-part2 - - - - - - - - ONLINE
ata-Samsung_SSD_860_EVO_500GB_S4XB-part2 - - - - - - - - ONLINE
rpool 460G 3.66G 456G - - 0% 0% 1.00x ONLINE -
mirror 460G 3.66G 456G - - 0% 0.79% - ONLINE
ata-Samsung_SSD_860_EVO_500GB_S4BB-part3 - - - - - - - - ONLINE
ata-Samsung_SSD_860_EVO_500GB_S4XB-part3 - - - - - - - - ONLINE
vol1 9.06T 3.77T 5.29T - - 13% 41% 1.00x ONLINE -
mirror 3.62T 1.93T 1.70T - - 25% 53.1% - ONLINE
ata-WDC_WD40EFRX-68N32N0_WD-WCC - - - - - - - - ONLINE
ata-ST4000VN008-2DR166_ZM4-part2 - - - - - - - - ONLINE
mirror 3.62T 1.84T 1.78T - - 8% 50.9% - ONLINE
ata-ST4000VN008-2DR166_ZM4-part2 - - - - - - - - ONLINE
ata-WDC_WD40EFRX-68N32N0_WD-WCC-part2 - - - - - - - - ONLINE
mirror 1.81T 484K 1.81T - - 0% 0.00% - ONLINE
ata-WDC_WD20EARX-00PASB0_WD-WMA-part2 - - - - - - - - ONLINE
ata-ST2000DM001-1CH164_Z1E-part2 - - - - - - - - ONLINE
and have a look at difference of the CAP
value (SIZE
/FREE
vs ALLOC
ratio) between vdevs.
Due to the working principle of this script, which essentially creates a duplicate file on purpose, deduplication will most definitely prevent it from working as intended. If you use deduplication you probably have to resort to a more expensive rebalancing method that involves additional drives.
Due to the working principle of this script, it is crucial that you only run it on data that is not actively accessed, since the original file will be deleted.
If you do a snapshot of the data you want to balance before starting the rebalancing script, keep in mind that ZFS now has to keep track of all of the data in the target directory twice. Once in the snapshot you made, and once for the new copy. This means that you will effectively use double the file size of all files within the target directory. Therefore it is a good idea to process the pool data in batches and remove old snapshots along the way, since you probably will be hitting the capacity limits of your pool at some point during the rebalancing process.
Since this is a simple bash script, there is no package. Simply download the script and make it executable:
curl -O https://raw.githubusercontent.com/markusressel/zfs-inplace-rebalancing/master/zfs-inplace-rebalancing.sh
chmod +x ./zfs-inplace-rebalancing.sh
Dependencies:
perl
- it should be available on most systems by default
ALWAYS HAVE A BACKUP OF YOUR DATA!
You can print a help message by running the script without any parameters:
./zfs-inplace-rebalancing.sh
Name | Description | Default |
---|---|---|
-c --checksum |
Whether to compare attributes and content of the copied file using an MD5 checksum. Technically this is a redundent check and consumes a lot of resources, so think twice. | true |
-p --passes |
The maximum number of rebalance passes per file. Setting this to infinity by using a value <= 0 might improve performance when rebalancing a lot of small files. |
1 |
--skip-hardlinks |
Skip rebalancing hardlinked files, since it will only create duplicate data. | false |
Make sure to run this script with a user that has rw permission to all of the files in the target directory. The easiest way to achieve this is by running the script as root.
sudo su
./zfs-inplace-rebalancing.sh --checksum true --passes 1 /pool/path/to/rebalance
To keep track of the balancing progress, you can open another terminal and run:
watch zpool list -v
To write the output to a file, simply redirect stdout and stderr to a file (or separate files). Since this redirects all output, you will have to follow the contents of the log files to get realtime info:
# one shell window:
tail -F ./stdout.log
# another shell window:
./zfs-inplace-rebalancing.sh /pool/path/to/rebalance >> ./stdout.log 2>> ./stderr.log
Although this script does have a progress output (files as well as percentage) it might be a good idea to try a small subfolder first, or process your pool folder layout in manually selected badges. This can also limit the damage done, if anything bad happens.
When aborting the script midway through, be sure to check the last lines of its output. When cancelling before or during the renaming process a ".balance" file might be left and you have to rename (or delete) it manually.
Although the --passes
parameter can be used to limit the maximum amount of rebalance passes per file, it is only meant to speedup aborted runs. Individual files will not be process multiple times automatically. To reach multiple passes you have to run the script on the same target directory multiple times.
To increase portability, this script can also be run using docker:
sudo docker run --rm -it -v /your/data:/data ghcr.io/markusressel/zfs-inplace-rebalancing:latest ./data
GitHub is for social coding: if you want to write code, I encourage contributions through pull requests from forks of this repository. Create GitHub tickets for bugs and new features and comment on the ones that you are interested in.
This script was inspired by zfs-balancer.
This software is provided "as is" and "as available", without any warranty.
ALWAYS HAVE A BACKUP OF YOUR DATA!