dwarfs/doc/dwarfs.md

dwarfs(1) -- mount highly compressed read-only file system

SYNOPSIS

dwarfs image mountpoint [options...]

DESCRIPTION

dwarfs is the FUSE driver for DwarFS, a highly compressed, read-only file system. As such, it's similar to file systems like SquashFS, cramfs or CromFS, but it has some distinct features.

Other than that, it's pretty straightforward to use. Once you've created a file system image using mkdwarfs(1), you can mount it with:

dwarfs image.dwarfs /path/to/mountpoint

OPTIONS

In addition to the regular FUSE options, dwarfs supports the following options:

• -o cachesize=value: Size of the block cache, in bytes. You can append suffixes (k, m, g) to specify the size in KiB, MiB and GiB, respectively. Note that this is not the upper memory limit of the process, as there may be blocks in flight that are not stored in the cache. Also, each block that hasn't been fully decompressed yet will carry decompressor state along with it, which can use a significant amount of additional memory. For more details, see mkdwarfs(1).

• -o blocksize=value: Size reported for files in st_blksize. You can use this to optimize throughput in certain situations.

• -o readahead=value: How much data to read ahead when receiving a read request. This is experimental and disabled by default. If you perform a lot of large, sequential reads, throughput may benefit from enabling readahead.

• -o workers=value: Number of worker threads to use for decompressing blocks. If you have a lot of CPUs, increasing this number can help speed up access to files in the filesystem.

• -o uid=num: Override the user ID for the whole file system. This option is not supported on Windows.

• -o gid=num: Override the group ID for the whole file system. This option is not supported on Windows.

• -o decratio=value: The ratio over which a block is fully decompressed. Blocks are only decompressed partially, so each block has to carry the decompressor state with it until it is fully decompressed. However, if a certain fraction of the block has already been decompressed, it may be beneficial to just decompress the rest and free the decompressor state. This value determines the ratio at which we fully decompress the block rather than keeping a partially decompressed block. A value of 0.8 means that as long as we've decompressed less than 80% of the block, we keep the partially decompressed block, but if we've decompressed more then 80%, we'll fully decompress it.

• -o offset=value|auto: Specify the byte offset at which the filesystem is located in the image, or use auto to detect the offset automatically. This is only useful for images that have some header located before the actual filesystem data.

• -o imagesize=value: Explicitly set the size of the filesystem image in bytes, starting from the offset. This can be used in cases where the image is embedded in a larger file.

• -o mlock=none|try|must: Set this to try or must instead of the default none to try or require mlock()ing of the file system metadata into memory.

• -o enable_nlink: Set this option if you want correct hardlink counts for regular files. If this is not specified, the hardlink count will be 1. Enabling this will slow down the initialization of the fuse driver as the hardlink counts will be determined by a full file system scan (it only takes about a millisecond to scan through 100,000 files, so this isn't dramatic). The fuse driver will also consume more memory to hold the hardlink count table. This will be 4 bytes for every regular file inode.

• -o readonly: Show all file system entries as read-only. By default, DwarFS will preserve the original writability, which is obviously a lie as it's a read-only file system. However, this is needed for overlays to work correctly, as otherwise directories are seen as read-only by the overlay and it'll be impossible to create new files even in a writeable overlay. If you don't use overlays and want the file system to reflect its read-only state, you can set this option.

• -o case_insensitive: Perform case-insensitive lookups in the mounted file system, i.e. an entry orignally named ReadMe.txt can be accessed as readme.txt, README.TXT, or rEaDmE.tXt. This works across all platforms. When mounting a file system with many files, this may be slightly slower and consume slightly more memory as case- insensitive lookup requires an additional mapping table that is built on-demand. Note that this is not supported if the file system contains directories with entries that only differ in case.

• -o preload_category=category: Preload all blocks from this category when mounting the file system. This is typically used together with the mkdwarfs "hotness" categorizer. If the cache size is too small, only as many blocks as will fit in the cache will be preloaded.

• -o preload_all Preload all blocks from the file system. This is only useful for file systems where all uncompressed blocks fit fully into the configured cache size. To see the uncompressed block size, you can use dwarfsck. If the cache size is too small, only as many blocks as will fit in the cache will be preloaded.

• -o (no_)cache_image: By default, dwarfs tries to ensure that the compressed file system image will not be cached by the kernel (i.e. the default is -o no_cache_image). This will reduce the memory consumption of the FUSE driver to slightly more than the cachesize, plus the size of the metadata block. This usually isn't a problem, especially when the image is stored on an SSD, but if you want to maximize performance it can be beneficial to use -o cache_image to keep the compressed image data in the kernel cache.

• -o (no_)cache_files: By default, files in the mounted file system will be cached by the kernel (i.e. the default is -o cache_files). This will significantly improve performance when accessing the same files over and over again, especially if the data from these files has been (partially) evicted from the block cache. By setting the -o no_cache_files option, you can force the fuse driver to not use the kernel cache for file data. If you're short on memory and only infrequently accessing files, this can be worth trying, even though it's likely that the kernel will already do the right thing even when the cache is enabled.

• -o debuglevel=name: Use this for different levels of verbosity along with either the -f or -d FUSE options. This can give you some insight over what the file system driver is doing internally, but it's mainly meant for debugging and the debug and trace levels in particular will slow down the driver. This defaults to info in foreground mode (-f, -d) and to warn in background mode.

• -o analysis_file=file: Write the paths of all files that were opened while the file system image was mounted to this file. This can be used as a set of "hot" files for the hotness categorizer in mkdwarfs. See the mkdwarfs documentation for details on producing images optimized for fast access times after mounting.

• -o tidy_strategy=none|time|swap: Use one of the following strategies to tidy the block cache. none is the default strategy that never tidies the cache. Blocks will only be evicted from the cache if the cache is full and a more recently used block is added to the cache. time enables a time-based tidying strategy. Every tidy_interval, the block cache is traversed and all blocks that have not been accessed for more than tidy_max_age will be removed. swap enables a swap-based tidying strategy. Every tidy_interval, the block cache is traversed and all blocks that have been fully or partially swapped out by the kernel will be removed.

• -o tidy_interval=time: Used only if tidy_strategy is not none. This is the interval at which the cache tidying thread wakes up to look for blocks that can be removed from the cache. This must be an integer value. Suffixes ms, s, m, h are supported. If no suffix is given, the value will be assumed to be in seconds.

• -o tidy_max_age=time: Used only if tidy_strategy is time. A block will be removed from the cache if it hasn't been used for this time span. This must be an integer value. Suffixes ms, s, m, h are supported. If no suffix is given, the value will be assumed to be in seconds.

• -o block_allocator=malloc|mmap: Select the allocator for decompressed file system blocks. By default, blocks will be allocated using malloc. However, depending on the way that malloc is implemented on your system, you may find that memory used by dwarfs isn't released despite using cache tidying. In this case, using the mmap block allocator is much more likely to release the memory. Note, however, that the mmap allocator can be slower than the malloc allocator. If your use case causes large numbers of blocks to be constantly created/evicted (e.g. you have a hugh image and are randomly accessing a large fraction of the files), this may impact the performance.

• -o seq_detector=num: Threshold, in blocks, for the sequential access detector. If the most recently accessed num blocks are sequential, then the block following the sequence is prefetched. This can significantly increase throughput if data is accessed sequentially. A value of 0 completely disables detection and prefetching.

• -o perfmon=name[+name...]: Enable performance monitoring for the list of +-separated components. This option is only available if the project was built with performance monitoring enabled. Available components include fuse, filesystem_v2, inode_reader_v2 and block_cache.

• -o perfmon_trace=file: Write JSON trace data for all components enabled by --perfmon to this file when the process exits.

• --man: If the project was built with support for built-in manual pages, this option will show the manual page. If supported by the terminal and a suitable pager (e.g. less) is found, the manual page is displayed in the pager.

There's two particular FUSE options that you'll likely need at some point, e.g. when trying to set up an overlayfs mount on top of a DwarFS image:

• -o allow_root and -o allow_other: These will ensure that the mounted file system can be read by either root or any other user in addition to the user that started the fuse driver. So if you're running dwarfs as a non-privileged user, you want to -o allow_root in case root needs access, for example when you're trying to use overlayfs along with dwarfs. If you're running dwarfs as root, you need allow_other.

TIPS & TRICKS

Adding a DwarFS image to /etc/fstab

This should be relatively straightforward if you're already familiar with adding other FUSE file systems to /etc/fstab. An entry looks like this:

dwarfs#/path/to/image.dwarfs /mnt/mountpoint fuse noauto,defaults,user,cachesize=1g 0 0

The first bit before the # tells mount to look for mount.dwarfs, which is installed as a symbolic link to the DwarFS FUSE driver. The part after the # looks pretty much like any other fstab entry. It starts with the path of the file system image to mount, followed by the mount point, followed by the file system type (fuse), and finally followed by a set of options.

If you want to automatically mount a DwarFS file system, you'll also need the allow_other option to make sure non-privileged users will be able to access the data. If you want to work with overlays, you'll need either allow_other or allow_root. For any of these options to work, you will have to set user_allow_other in /etc/fuse.conf.

Setting up a writable file system on top of a DwarFS image

This will show you how to set up a read/write layer on top of a read-only DwarFS image, which can be incredibly handy if you want to be able to partially and/or temporarily modify/amend the contents of your read-only image.

My primary use case for this feature is keeping over 1000 Perl versions in the DwarFS image and then setting up a read/write layer to be able to install additional modules for each of these versions. When I didn't need the modules anymore, I could just completely wipe the read/write layer and get my pristine original set of Perl versions back.

Here's what you need to do:

• Create a set of directories. In my case, these are all located in /tmp/perl as this was the original install location.

  cd /tmp/perl
  mkdir install-ro
  mkdir install-rw
  mkdir install-work
  mkdir install
  

• Mount the DwarFS image. -o allow_root is needed to make sure overlayfs has access to the mounted file system. In order to use -o allow_root, you may have to uncomment or add user_allow_other in /etc/fuse.conf.

  dwarfs perl-install.dwarfs install-ro -o allow_root
  

• Now set up overlayfs.

  sudo mount -t overlay overlay -o lowerdir=install-ro,upperdir=install-rw,workdir=install-work install
  

• That's it. You should now be able to access a writeable version of your DwarFS image in install.

You can go even further than that. Say you have different sets of modules that you regularly want to layer on top of the base DwarFS image. In that case, you can simply build a new DwarFS image from the read-write directory after unmounting the overlayfs, and selectively add this by passing a colon-separated list to the lowerdir option when setting up the overlayfs mount:

sudo mount -t overlay overlay -o lowerdir=install-ro:install-modules install

If you want this merged overlay to be writable, just add in the upperdir and workdir options from before again.

Optimizing Performance and Memory Usage

Depending on your use case, you may want to ensure that dwarfs isn't constantly consuming large amounts of memory. Or you may want to make sure the file system can always be accessed as quickly as possible. There are several options to tune performance based on your use case.

If you don't care much about memory, use the cachesize option to make sure as many decompressed file system blocks as possible can be kept in memory.

If your file system image is relatively small, you can also use the preload_all option to immediately populate the cache after mounting.

The more interesting use case is if you want to be conservative about memory, but still don't want to sacrifice performance too much. Maybe you only need to access a lot of files directly after mounting and then only infrequently need to access other files. If this is the case, you can use the tidy_strategy, tidy_interval and tidy_max_age options. With these options, you can usually keep the cachesize relatively large in order to maintain good throughput when accessing files, but the cache will be tidied up quickly, releasing the memory again if it is no longer accessed. A useful configuration could look like this:

dwarfs image mountpoint -otidy_strategy=time,tidy_interval=5s,tidy_max_age=10s

This will check the cache every 5 seconds and evict any blocks from the cache that haven't been accessed for more than 10 seconds. What sounds good in theory can be tricky in practice: just because dwarfs has freed the memory doesn't necessarily mean that the memory allocator will really return the memory to the system.

If dwarfs is built with jemalloc, the memory allocator can be tuned to return memory to the system quickly by setting the MALLOC_CONF environment variable, for example:

MALLOC_CONF="background_thread:true,dirty_decay_ms:5000,muzzy_decay_ms:5000"

If dwarfs is not build with jemalloc, it is still possible to run it with jemalloc by using LD_PRELOAD:

LD_PRELOAD=/usr/lib/libjemalloc.so dwarfs image mountpoint ...

The exact location of the jemalloc shared object depends on your system. If that is also not an option, you can use the block_allocator option to dwarfs:

dwarfs image mountpoint -oblock_allocator=mmap,tidy_strategy=time,...

This will instruct dwarfs to not use malloc for allocating blocks, but rather use mmap. This should work nicely, albeit with some potential impact on performance, especially with smaller block sizes.

Optimizing Application Startup Time

If you're using DwarFS as storage for an application container, you may want to optimize startup time. There are different ways to do that.

If the application is going to read most of the file system image data during startup, and the image is relatively small, it's worth trying to just use the preload_all option. This will fill the cache with blocks from the file system image as soon as it is mounted and can already have a significant impact on startup time.

If the application is only using a small subset of the data in the image during startup, you can use "hotness" analysis and build an image that is optimized to improve startup speed. It's basically profile guide optimization for file systems.

First you have to build an initial image you can use to perform the analysis. Then, you use the analysis_file option to mount the image:

dwarfs image mountpoint -oanalysis_file=/tmp/image.prof

While the image is mounted, you start the application (or perform whichever task you want to optimize the file system for). Then, when you unmount the image, the file you have specified will contain a list of all paths in the image that have been accessed, in the order in which the access happened.

You can then use the hotness categorizer in mkdwarfs, potentially along with explicit ordering, to build the optimized image:

mkdwarfs -i input -o image --categorize=hotness --hotness-list=/tmp/image.prof

Or, with additional explicit ordering:

mkdwarfs -i input -o image --categorize=hotness --hotness-list=/tmp/image.prof \
         --order hotness::explicit:file=/tmp/image.prof

This will order the files in the hotness category using the same order as in the profile. Otherwise, they will be ordered by similarity.

Once you have built this optimized image, you can mount it using the preload_category option:

dwarfs image mountpoint -opreload_category=hotness

This will preload all hotness blocks into the cache immediately after mounting and hopefully speed up application startup significantly.

There are plenty of other ways you can tune how the image is generated. For example, if the input data already contains compressed files, you may want to add the incompressible categorizer. This will not only speed up the creation of the file system image as mkdwarfs won't waste time trying to compress incompressible data, but also speed up access as the data won't need to be decompressed. Also, you could think about using different compression algorithms for the "hot" and "cold" files, e.g. something fast like zstd for the hot files and lzma for the cold files.

AUTHOR

Written by Marcus Holland-Moritz.

COPYRIGHT

Copyright (C) Marcus Holland-Moritz.

SEE ALSO

mkdwarfs(1), dwarfsextract(1), dwarfsck(1), dwarfs-format(5)