NAME¶

nbdkit-cache-filter - nbdkit caching filter

SYNOPSIS¶

 nbdkit --filter=cache plugin [plugin-args...]
                              [cache=writeback|writethrough|unsafe]
                              [cache-min-block-size=SIZE]
                              [cache-max-size=SIZE]
                              [cache-high-threshold=N]
                              [cache-low-threshold=N]
                              [cache-on-read=true|false|/PATH]

DESCRIPTION¶

"nbdkit-cache-filter" is a filter that adds caching on top of a plugin.

This is useful if a plugin is slow or expensive to use, because nbdkit will try to minimize requests to the plugin by caching previous requests. You should use "cache-on-read=true" (not the default) to cache reads from slow plugins.

Note that many NBD clients are able to do caching, and because the caching happens on the client side it will usually be more effective than caching inside the server. This filter can be used if the client does not have effective caching, or (with "cache=unsafe") to defeat flush requests from the client (which is unsafe and can cause data loss, as the name suggests).

This filter only caches image contents. To cache image metadata, use nbdkit-cacheextents-filter(1) between this filter and the plugin. To accelerate sequential reads, use nbdkit-readahead-filter(1) or nbdkit-scan-filter(1) on top of this filter.

PARAMETERS¶

cache=writeback

Store writes in the cache. They are not written to the plugin unless an explicit flush is done by the client.

This is the default caching mode, and is safe if your client issues flush requests correctly (which is true for modern Linux and other well-written NBD clients).

cache=writethrough

Always force writes through to the plugin.

This makes the cache less effective, but is necessary if your client does not issue correct flush requests.

cache=unsafe

Ignore flush requests. Never write to the plugin unless the cache grows too large.

This is dangerous and can cause data loss, but this may be acceptable if you only use it for testing or with data that you don't care about or can cheaply reconstruct.

cache-min-block-size=SIZE

Set the minimum block size used by the cache. This must be a power of 2 and ≥ 4096.

The default is 64K, or the block size of the filesystem which contains the temporary file storing the cache (whichever is larger).

cache-max-size=SIZE

cache-high-threshold=N

cache-low-threshold=N

(nbdkit ≥ 1.10)

Limit the size of the cache to "SIZE". See "CACHE MAXIMUM SIZE" below.

cache-on-read=true

(nbdkit ≥ 1.10)

Cache read requests as well as write and cache requests. Any time a block is read from the plugin, it is saved in the cache (if there is sufficient space) so the same data can be served more quickly later.

Note that if the underlying data served by the plugin can be modified by some other means (eg. something else can write to a file which is being served by nbdkit-file-plugin(1)), this option will cause nbdkit to serve stale data because reads won't always go through to the plugin.

cache-on-read=false

Do not cache read requests (this is the default).

cache-on-read=/PATH

(nbdkit ≥ 1.28)

When /PATH (which must be an absolute path) exists, this behaves like "cache-on-read=true", and when it does not exist like "cache-on-read=false". This allows you to control the cache-on-read behaviour while nbdkit is running.

CACHE MAXIMUM SIZE¶

By default the cache can grow to any size (although not larger than the virtual size of the underlying plugin) and you have to ensure there is sufficient space in $TMPDIR for it.

Using the parameters "cache-max-size", "cache-high-threshold" and "cache-low-threshold" you can limit the maximum size of the cache.

This requires kernel and filesystem support (for fallocate(2) "FALLOC_FL_PUNCH_HOLE"), so it may not work on all platforms.

Some examples:

"cache-max-size=1G"

The cache is limited to around 1 gigabyte.

"cache-max-size=1G cache-high-threshold=95 cache-low-threshold=80"

Once the cache size reaches 972M (95% of 1G), blocks are reclaimed from the cache until the size is reduced to 819M (80% of 1G).

The way this works is once the size of the cache exceeds "SIZE" ✕ the high threshold, the filter works to reduce the size of the cache until it is less than "SIZE" ✕ the low threshold. Once the size is below the low threshold, no more reclaim work is done until the size exceeds the high threshold again.

The default thresholds are high 95% and low 80%. You must set 0 < low < high. The thresholds are expressed as integer percentages of "cache-max-size".

Least recently used blocks are discarded first.

ENVIRONMENT VARIABLES¶

"TMPDIR"

The cache is stored in a temporary file located in /var/tmp by default. You can override this location by setting the "TMPDIR" environment variable before starting nbdkit.

FILES¶

$filterdir/nbdkit-cache-filter.so

The filter.

Use "nbdkit --dump-config" to find the location of $filterdir.

VERSION¶

"nbdkit-cache-filter" first appeared in nbdkit 1.2.

SEE ALSO¶

nbdkit(1), nbdkit-file-plugin(1), nbdkit-cacheextents-filter(1), nbdkit-cow-filter(1), nbdkit-readahead-filter(1), nbdkit-filter(3), qemu-img(1).

AUTHORS¶

Eric Blake

Richard W.M. Jones

COPYRIGHT¶

Copyright Red Hat

LICENSE¶

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
• Neither the name of Red Hat nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY RED HAT AND CONTRIBUTORS ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL RED HAT OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.