<p>Run <code>rclone config</code> to setup. See <ahref="http://rclone.org/docs/">rclone config docs</a> for more details.</p>
<pre><code>rclone config</code></pre>
<h2id="install-from-source">Install from source</h2>
<p>Make sure you have at least <ahref="https://golang.org/">Go</a> 1.5 installed. Make sure your <code>GOPATH</code> is set, then:</p>
<pre><code>go get -u -v github.com/ncw/rclone</code></pre>
<p>and this will build the binary in <code>$GOPATH/bin</code>. If you have built rclone before then you will want to update its dependencies first with this</p>
<pre><code>go get -u -v github.com/ncw/rclone/...</code></pre>
<p>First you'll need to configure rclone. As the object storage systems have quite complicated authentication these are kept in a config file <code>.rclone.conf</code> in your home directory by default. (You can use the <code>--config</code> option to choose a different config file.)</p>
<p>The easiest way to make the config is to run rclone with the config option:</p>
<pre><code>rclone config</code></pre>
<p>See the following for detailed instructions for</p>
<p>Source and destination paths are specified by the name you gave the storage system in the config file then the sub path, eg "drive:myfolder" to look at "myfolder" in Google drive.</p>
<p>You can define as many storage paths as you like in the config file.</p>
<p>Copy the source to the destination. Doesn't transfer unchanged files, testing by size and modification time or MD5SUM. Doesn't delete files from the destination.</p>
<p>Note that it is always the contents of the directory that is synced, not the directory so when source:path is a directory, it's the contents of source:path that are copied, not the directory name and contents.</p>
<p>If dest:path doesn't exist, it is created and the source:path contents go there.</p>
<p>Let's say there are two files in sourcepath</p>
<pre><code>sourcepath/one.txt
sourcepath/two.txt</code></pre>
<p>This copies them to</p>
<pre><code>destpath/one.txt
destpath/two.txt</code></pre>
<p>Not to</p>
<pre><code>destpath/sourcepath/one.txt
destpath/sourcepath/two.txt</code></pre>
<p>If you are familiar with <code>rsync</code>, rclone always works as if you had written a trailing / - meaning "copy the contents of this directory". This applies to all commands and whether you are talking about the source or destination.</p>
<p>Sync the source to the destination, changing the destination only. Doesn't transfer unchanged files, testing by size and modification time or MD5SUM. Destination is updated to match source, including deleting files if necessary.</p>
<p><strong>Important</strong>: Since this can cause data loss, test first with the <code>--dry-run</code> flag to see exactly what would be copied and deleted.</p>
<p>Note that files in the destination won't be deleted if there were any errors at any point.</p>
<p>It is always the contents of the directory that is synced, not the directory so when source:path is a directory, it's the contents of source:path that are copied, not the directory name and contents. See extended explanation in the <code>copy</code> command above if unsure.</p>
<p>If dest:path doesn't exist, it is created and the source:path contents go there.</p>
<p>Moves the contents of the source directory to the destination directory. Rclone will error if the source and destination overlap and the remote does not support a server side directory move operation.</p>
<p>If no filters are in use and if possible this will server side move <code>source:path</code> into <code>dest:path</code>. After this <code>source:path</code> will no longer longer exist.</p>
<p>Otherwise for each file in <code>source:path</code> selected by the filters (if any) this will move it into <code>dest:path</code>. If possible a server side move will be used, otherwise it will copy it (server side if possible) into <code>dest:path</code> then delete the original (if no errors on copy) in <code>source:path</code>.</p>
<p>Remove the path and all of its contents. Note that this does not obey include/exclude filters - everything will be removed. Use <code>delete</code> if you want to selectively delete files.</p>
<pre><code>rclone purge remote:path</code></pre>
<h2id="rclone-mkdir">rclone mkdir</h2>
<p>Make the path if it doesn't already exist.</p>
<h3id="synopsis-6">Synopsis</h3>
<p>Make the path if it doesn't already exist.</p>
<pre><code>rclone mkdir remote:path</code></pre>
<h2id="rclone-rmdir">rclone rmdir</h2>
<p>Remove the path if empty.</p>
<h3id="synopsis-7">Synopsis</h3>
<p>Remove the path. Note that you can't remove a path with objects in it, use purge for that.</p>
<pre><code>rclone rmdir remote:path</code></pre>
<h2id="rclone-check">rclone check</h2>
<p>Checks the files in the source and destination match.</p>
<p>Checks the files in the source and destination match. It compares sizes and MD5SUMs and prints a report of files which don't match. It doesn't alter the source or destination.</p>
<p>By default <code>dedup</code> interactively finds duplicate files and offers to delete all but one or rename them to be different. Only useful with Google Drive which can have duplicate file names.</p>
<p>The <code>dedupe</code> command will delete all but one of any identical (same md5sum) files it finds without confirmation. This means that for most duplicated files the <code>dedupe</code> command will not be interactive. You can use <code>--dry-run</code> to see what would happen without doing anything.</p>
<p>Copy files from source to dest, skipping already copied</p>
<h3id="synopsis-20">Synopsis</h3>
<p>If source:path is a file or directory then it copies it to a file or directory named dest:path.</p>
<p>This can be used to upload single files to other than their current name. If the source is a directory then it acts exactly like the copy command.</p>
<p>So</p>
<pre><code>rclone copyto src dst</code></pre>
<p>where src and dst are rclone paths, either remote:path or /path/to/local or C:.</p>
<p>This will:</p>
<pre><code>if src is file
copy it to dst, overwriting an existing file if it exists
if src is directory
copy it to dst, overwriting existing files if they exist
see copy command for full details</code></pre>
<p>This doesn't transfer unchanged files, testing by size and modification time or MD5SUM. It doesn't delete files from the destination.</p>
<p>This produces markdown docs for the rclone commands to the directory supplied. These are in a format suitable for hugo to render into the rclone.org website.</p>
<p>This can only write files seqentially, it can only seek when reading.</p>
<p>Rclone mount inherits rclone's directory handling. In rclone's world directories don't really exist. This means that empty directories will have a tendency to disappear once they fall out of the directory cache.</p>
<p>The bucket based FSes (eg swift, s3, google compute storage, b2) won't work from the root - you will need to specify a bucket, or a path within the bucket. So <code>swift:</code> won't work whereas <code>swift:bucket</code> will as will <code>swift:bucket/path</code>.</p>
<p>Only supported on Linux, FreeBSD and OS X at the moment.</p>
<h3id="rclone-mount-vs-rclone-synccopy">rclone mount vs rclone sync/copy</h3>
<p>File systems expect things to be 100% reliable, whereas cloud storage systems are a long way from 100% reliable. The rclone sync/copy commands cope with this with lots of retries. However rclone mount can't use retries in the same way without making local copies of the uploads. This might happen in the future, but for the moment rclone mount won't do that, so will be less reliable than the rclone command.</p>
<h3id="bugs">Bugs</h3>
<ul>
<li>All the remotes should work for read, but some may not for write
<p>Move file or directory from source to dest.</p>
<h3id="synopsis-25">Synopsis</h3>
<p>If source:path is a file or directory then it moves it to a file or directory named dest:path.</p>
<p>This can be used to rename files or upload single files to other than their existing name. If the source is a directory then it acts exacty like the move command.</p>
<p>So</p>
<pre><code>rclone moveto src dst</code></pre>
<p>where src and dst are rclone paths, either remote:path or /path/to/local or C:.</p>
<p>This will:</p>
<pre><code>if src is file
move it to dst, overwriting an existing file if it exists
if src is directory
move it to dst, overwriting existing files if they exist
see move command for full details</code></pre>
<p>This doesn't transfer unchanged files, testing by size and modification time or MD5SUM. src will be deleted on successful transfer.</p>
<p><strong>Important</strong>: Since this can cause data loss, test first with the --dry-run flag.</p>
<p>Remove any empty directoryies under the path.</p>
<h3id="synopsis-26">Synopsis</h3>
<p>This removes any empty directories (or directories that only contain empty directories) under the path that it finds, including the path if it has nothing in.</p>
<p>This is useful for tidying up remotes that rclone has left a lot of empty directories in.</p>
<h2id="copying-single-files">Copying single files</h2>
<p>rclone normally syncs or copies directories. However if the source remote points to a file, rclone will just copy that file. The destination remote must point to a directory - rclone will give the error <code>Failed to create file system for "remote:file": is a file not a directory</code> if it isn't.</p>
<p>For example, suppose you have a remote with a file in called <code>test.jpg</code>, then you could copy just that file like this</p>
<p>Where <code>/tmp/files</code> contains the single line</p>
<pre><code>test.jpg</code></pre>
<p>It is recommended to use <code>copy</code> when copying single files not <code>sync</code>. They have pretty much the same effect but <code>copy</code> will use a lot less memory.</p>
<h2id="quoting-and-the-shell">Quoting and the shell</h2>
<p>When you are typing commands to your computer you are using something called the command line shell. This interprets various characters in an OS specific way.</p>
<p>Here are some gotchas which may help users unfamiliar with the shell rules</p>
<h3id="linux-osx">Linux / OSX</h3>
<p>If your names have spaces or shell metacharacters (eg <code>*</code>, <code>?</code>, <code>$</code>, <code>'</code>, <code>"</code> etc) then you must quote them. Use single quotes <code>'</code> by default.</p>
<p>Drive, S3, Dropbox, Swift and Google Cloud Storage support server side copy.</p>
<p>This means if you want to copy one folder to another then rclone won't download all the files and re-upload them; it will instruct the server to copy them in place.</p>
<p>Rclone has a number of options to control its behaviour.</p>
<p>Options which use TIME use the go time parser. A duration string is a possibly signed sequence of decimal numbers, each with optional fraction and a unit suffix, such as "300ms", "-1.5h" or "2h45m". Valid time units are "ns", "us" (or "µs"), "ms", "s", "m", "h".</p>
<p>Options which use SIZE use kByte by default. However a suffix of <code>b</code> for bytes, <code>k</code> for kBytes, <code>M</code> for MBytes and <code>G</code> for GBytes may be used. These are the binary units, eg 1, 2**10, 2**20, 2**30 respectively.</p>
<p>Note that the units are Bytes/s not Bits/s. Typically connections are measured in Bits/s - to convert divide by 8. For example let's say you have a 10 Mbit/s connection and you wish rclone to use half of it - 5 Mbit/s. This is 5/8 = 0.625MByte/s so you would use a <code>--bwlimit 0.625M</code> parameter for rclone.</p>
<p>The number of checkers to run in parallel. Checkers do the equality checking of files during a sync. For some storage systems (eg s3, swift, dropbox) this can take a significant amount of time so they are run in parallel.</p>
<p>The default is to run 8 checkers in parallel.</p>
<h3id="c---checksum">-c, --checksum</h3>
<p>Normally rclone will look at modification time and size of files to see if they are equal. If you set this flag then rclone will check the file hash and size to determine if files are equal.</p>
<p>This is useful when the remote doesn't support setting modified time and a more accurate sync is desired than just checking the file size.</p>
<p>This is very useful when transferring between remotes which store the same hash type on the object, eg Drive and Swift. For details of which remotes support which hash type see the table in the <ahref="http://rclone.org/overview/">overview section</a>.</p>
<p>Eg <code>rclone --checksum sync s3:/bucket swift:/bucket</code> would run much quicker than without the <code>--checksum</code> flag.</p>
<p>When using this flag, rclone won't update mtimes of remote files if they are incorrect as it would normally.</p>
<p>Specify the location of the rclone config file. Normally this is in your home directory as a file called <code>.rclone.conf</code>. If you run <code>rclone -h</code> and look at the help for the <code>--config</code> option you will see where the default location is for you. Use this flag to override the config location, eg <code>rclone --config=".myconfig" .config</code>.</p>
<h3id="contimeouttime">--contimeout=TIME</h3>
<p>Set the connection timeout. This should be in go time format which looks like <code>5s</code> for 5 seconds, <code>10m</code> for 10 minutes, or <code>3h30m</code>.</p>
<p>The connection timeout is the amount of time rclone will wait for a connection to go through to a remote object storage system. It is <code>1m</code> by default.</p>
<p>Mode to run dedupe command in. One of <code>interactive</code>, <code>skip</code>, <code>first</code>, <code>newest</code>, <code>oldest</code>, <code>rename</code>. The default is <code>interactive</code>. See the dedupe command for more information as to what these options mean.</p>
<p>Do a trial run with no permanent changes. Use this to see what rclone would do without actually doing it. Useful when setting up the <code>sync</code> command which deletes files in the destination.</p>
<h3id="ignore-existing">--ignore-existing</h3>
<p>Using this option will make rclone unconditionally skip all files that exist on the destination, no matter the content of these files.</p>
<p>While this isn't a generally recommended option, it can be useful in cases where your files change due to encryption. However, it cannot correct partial transfers in case a transfer was interrupted.</p>
<p>Normally rclone will look at modification time and size of files to see if they are equal. If you set this flag then rclone will check only the modification time. If <code>--checksum</code> is set then it only checks the checksum.</p>
<p>It will also cause rclone to skip verifying the sizes are the same after transfer.</p>
<p>This can be useful for transferring files to and from onedrive which occasionally misreports the size of image files (see <ahref="https://github.com/ncw/rclone/issues/399">#399</a> for more info).</p>
<p>Using this option will cause rclone to unconditionally upload all files regardless of the state of files on the destination.</p>
<p>Normally rclone would skip any files that have the same modification time and are the same size (or have the same checksum if using <code>--checksum</code>).</p>
<p>Log all of rclone's output to FILE. This is not active by default. This can be useful for tracking down problems with syncs in combination with the <code>-v</code> flag. See the Logging section for more info.</p>
<p>This controls the number of low level retries rclone does.</p>
<p>A low level retry is used to retry a failing operation - typically one HTTP request. This might be uploading a chunk of a big file for example. You will see low level retries in the log with the <code>-v</code> flag.</p>
<p>This shouldn't need to be changed from the default in normal operations, however if you get a lot of low level retries you may wish to reduce the value so rclone moves on to a high level retry (see the <code>--retries</code> flag) quicker.</p>
<p>Disable low level retries with <code>--low-level-retries 1</code>.</p>
<p>This modifies the recursion depth for all the commands except purge.</p>
<p>So if you do <code>rclone --max-depth 1 ls remote:path</code> you will see only the files in the top level directory. Using <code>--max-depth 2</code> means you will see all the files in first two directory levels and so on.</p>
<p>For historical reasons the <code>lsd</code> command defaults to using a <code>--max-depth</code> of 1 - you can override this with the command line flag.</p>
<p>You can use this command to disable recursion (with <code>--max-depth 1</code>).</p>
<p>Note that if you use this with <code>sync</code> and <code>--delete-excluded</code> the files not recursed through are considered excluded and will be deleted on the destination. Test first with <code>--dry-run</code> if you are not sure what will happen.</p>
<p>When checking whether a file has been modified, this is the maximum allowed time difference that a file can have and still be considered equivalent.</p>
<p>The default is <code>1ns</code> unless this is overridden by a remote. For example OS X only stores modification times to the nearest second so if you are reading and writing to an OS X filing system this will be <code>1s</code> by default.</p>
<p>This command line flag allows you to override that computed default.</p>
<p>Don't set <code>Accept-Encoding: gzip</code>. This means that rclone won't ask the server for compressed files automatically. Useful if you've set the server to return files with <code>Content-Encoding: gzip</code> but you uploaded compressed files.</p>
<p>There is no need to set this in normal operation, and doing so will decrease the network transfer efficiency of rclone.</p>
<p>Normally rclone outputs stats and a completion message. If you set this flag it will make as little output as possible.</p>
<h3id="retries-int">--retries int</h3>
<p>Retry the entire sync if it fails this many times it fails (default 3).</p>
<p>Some remotes can be unreliable and a few retries helps pick up the files which didn't get transferred because of errors.</p>
<p>Disable retries with <code>--retries 1</code>.</p>
<h3id="size-only">--size-only</h3>
<p>Normally rclone will look at modification time and size of files to see if they are equal. If you set this flag then rclone will check only the size.</p>
<p>This can be useful transferring files from dropbox which have been modified by the desktop sync client which doesn't set checksums of modification times in the same way as rclone.</p>
<p>Commands which transfer data (<code>sync</code>, <code>copy</code>, <code>copyto</code>, <code>move</code>, <code>moveto</code>) will print data transfer stats at regular intervals to show their progress.</p>
<p>If you set the stats interval then all command can show stats. This can be useful when running other commands, <code>check</code> or <code>mount</code> for example.</p>
<p>Specifying the value <code>--delete-before</code> will delete all files present on the destination, but not on the source <em>before</em> starting the transfer of any new or updated files. This uses extra memory as it has to store the source listing before proceeding.</p>
<p>Specifying <code>--delete-during</code> (default value) will delete files while checking and uploading files. This is usually the fastest option. Currently this works the same as <code>--delete-after</code> but it may change in the future.</p>
<p>Specifying <code>--delete-after</code> will delay deletion of files until all new/updated files have been successfully transfered.</p>
<h3id="timeouttime">--timeout=TIME</h3>
<p>This sets the IO idle timeout. If a transfer has started but then becomes idle for this long it is considered broken and disconnected.</p>
<p>The default is <code>5m</code>. Set to 0 to disable.</p>
<h3id="transfersn">--transfers=N</h3>
<p>The number of file transfers to run in parallel. It can sometimes be useful to set this to a smaller number if the remote is giving a lot of timeouts or bigger if you have lots of bandwidth and a fast remote.</p>
<p>The default is to run 4 file transfers in parallel.</p>
<p>This forces rclone to skip any files which exist on the destination and have a modified time that is newer than the source file.</p>
<p>If an existing destination file has a modification time equal (within the computed modify window precision) to the source file's, it will be updated if the sizes are different.</p>
<p>On remotes which don't support mod time directly the time checked will be the uploaded time. This means that if uploading to one of these remoes, rclone will skip any files which exist on the destination and have an uploaded time that is newer than the modification time of the source file.</p>
<p>This can be useful when transferring to a remote which doesn't support mod times directly as it is more accurate than a <code>--size-only</code> check and faster than using <code>--checksum</code>.</p>
<p>Your configuration file contains information for logging in to your cloud services. This means that you should keep your <code>.rclone.conf</code> file in a secure location.</p>
<p>If you are in an environment where that isn't possible, you can add a password to your configuration. This means that you will have to enter the password every time you start rclone.</p>
<p>To add a password to your rclone configuration, execute <code>rclone config</code>.</p>
<pre><code>>rclone config
Current remotes:
e) Edit existing remote
n) New remote
d) Delete remote
s) Set configuration password
q) Quit config
e/n/d/s/q></code></pre>
<p>Go into <code>s</code>, Set configuration password:</p>
<pre><code>e/n/d/s/q> s
Your configuration is not encrypted.
If you add a password, you will protect your login information to cloud services.
<p>Your configuration is now encrypted, and every time you start rclone you will now be asked for the password. In the same menu you can change the password or completely remove encryption from your configuration.</p>
<p>There is no way to recover the configuration if you lose your password.</p>
<p>rclone uses <ahref="https://godoc.org/golang.org/x/crypto/nacl/secretbox">nacl secretbox</a> which in turn uses XSalsa20 and Poly1305 to encrypt and authenticate your configuration with secret-key cryptography. The password is SHA-256 hashed, which produces the key for secretbox. The hashed password is not stored.</p>
<p>While this provides very good security, we do not recommend storing your encrypted rclone configuration in public if it contains sensitive information, maybe except if you use a very strong password.</p>
<p>If it is safe in your environment, you can set the <code>RCLONE_CONFIG_PASS</code> environment variable to contain your password, in which case it will be used for decrypting the configuration.</p>
<p>You can set this for a session from a script. For unix like systems save this to a file called <code>set-rclone-password</code>:</p>
<pre><code>#!/bin/echo Source this file don't run it
read -s RCLONE_CONFIG_PASS
export RCLONE_CONFIG_PASS</code></pre>
<p>Then source the file when you want to use it. From the shell you would do <code>source set-rclone-password</code>. It will then ask you for the password and set it in the envonment variable.</p>
<p>If you are running rclone inside a script, you might want to disable password prompts. To do that, pass the parameter <code>--ask-password=false</code> to rclone. This will make rclone fail instead of asking for a password if <code>RCLONE_CONFIG_PASS</code> doesn't contain a valid password.</p>
<p>These options are useful when developing or debugging rclone. There are also some more remote specific options which aren't documented here which are used for testing. These start with remote name eg <code>--drive-test-option</code> - see the docs for the remote in question.</p>
<h3id="cpuprofilefile">--cpuprofile=FILE</h3>
<p>Write CPU profile to file. This can be analysed with <code>go tool pprof</code>.</p>
<p>Dump HTTP headers - will contain sensitive info such as <code>Authorization:</code> headers - use <code>--dump-headers</code> to dump without <code>Authorization:</code> headers. Can be very verbose. Useful for debugging only.</p>
<p>Dump HTTP headers with <code>Authorization:</code> lines removed. May still contain sensitive info. Can be very verbose. Useful for debugging only.</p>
<p>Use <code>--dump-auth</code> if you do want the <code>Authorization:</code> headers.</p>
<p><code>--no-check-certificate</code> controls whether a client verifies the server's certificate chain and host name. If <code>--no-check-certificate</code> is true, TLS accepts any certificate presented by the server and any host name in that certificate. In this mode, TLS is susceptible to man-in-the-middle attacks.</p>
<p>This option defaults to <code>false</code>.</p>
<p><strong>This should be used only for testing.</strong></p>
<p>The <code>--no-traverse</code> flag controls whether the destination file system is traversed when using the <code>copy</code> or <code>move</code> commands.</p>
<p>If you are only copying a small number of files and/or have a large number of files on the destination then <code>--no-traverse</code> will stop rclone listing the destination and save time.</p>
<p>However if you are copying a large number of files, escpecially if you are doing a copy where lots of the files haven't changed and won't need copying then you shouldn't use <code>--no-traverse</code>.</p>
<p>It can also be used to reduce the memory usage of rclone when copying - <code>rclone --no-traverse copy src dst</code> won't load either the source or destination listings into memory so will use the minimum amount of memory.</p>
<p>rclone has 3 levels of logging, <code>Error</code>, <code>Info</code> and <code>Debug</code>.</p>
<p>By default rclone logs <code>Error</code> and <code>Info</code> to standard error and <code>Debug</code> to standard output. This means you can redirect standard output and standard error to different places.</p>
<p>By default rclone will produce <code>Error</code> and <code>Info</code> level messages.</p>
<p>If you use the <code>-q</code> flag, rclone will only produce <code>Error</code> messages.</p>
<p>If you use the <code>-v</code> flag, rclone will produce <code>Error</code>, <code>Info</code> and <code>Debug</code> messages.</p>
<p>If you use the <code>--log-file=FILE</code> option, rclone will redirect <code>Error</code>, <code>Info</code> and <code>Debug</code> messages along with standard error to FILE.</p>
<p>If any errors occurred during the command, rclone with an exit code of <code>1</code>. This allows scripts to detect when rclone operations have failed.</p>
<p>During the startup phase rclone will exit immediately if an error is detected in the configuration. There will always be a log message immediately before exiting.</p>
<p>When rclone is running it will accumulate errors as it goes along, and only exit with an non-zero exit code if (after retries) there were no transfers with errors remaining. For every error counted there will be a high priority log message (visibile with <code>-q</code>) showing the message and which file caused the problem. A high priority message is also shown when starting a retry so the user can see that any previous error messages may not be valid after the retry. If rclone has done a retry it will log a high priority message if the retry was successful.</p>
<h1id="configuring-rclone-on-a-remote-headless-machine">Configuring rclone on a remote / headless machine</h1>
<p>Some of the configurations (those involving oauth2) require an Internet connected web browser.</p>
<p>If you are trying to set rclone up on a remote or headless box with no browser available on it (eg a NAS or a server in a datacenter) then you will need to use an alternative means of configuration. There are two ways of doing it, described below.</p>
<h2id="configuring-using-rclone-authorize">Configuring using rclone authorize</h2>
<p>On the headless box</p>
<pre><code>...
Remote config
Use auto config?
* Say Y if not sure
* Say N if you are working on a remote or headless machine
y) Yes
n) No
y/n> n
For this to work, you will need rclone available on a machine that has a web browser available.
<p>Now transfer it to the remote box (scp, cut paste, ftp, sftp etc) and place it in the correct place (use <code>rclone -h</code> on the remote box to find out where).</p>
<h1id="filtering-includes-and-excludes">Filtering, includes and excludes</h1>
<p>Rclone has a sophisticated set of include and exclude rules. Some of these are based on patterns and some on other things like file size.</p>
<p>The filters are applied for the <code>copy</code>, <code>sync</code>, <code>move</code>, <code>ls</code>, <code>lsl</code>, <code>md5sum</code>, <code>sha1sum</code>, <code>size</code>, <code>delete</code> and <code>check</code> operations. Note that <code>purge</code> does not obey the filters.</p>
<p>Each path as it passes through rclone is matched against the include and exclude rules like <code>--include</code>, <code>--exclude</code>, <code>--include-from</code>, <code>--exclude-from</code>, <code>--filter</code>, or <code>--filter-from</code>. The simplest way to try them out is using the <code>ls</code> command, or <code>--dry-run</code> together with <code>-v</code>.</p>
<p>If the pattern starts with a <code>/</code> then it only matches at the top level of the directory tree, <strong>relative to the root of the remote</strong> (not necessarily the root of the local drive). If it doesn't start with <code>/</code> then it is matched starting at the <strong>end of the path</strong>, but it will only match a complete path element:</p>
- doesn't match "directory/file.jpg"
- doesn't match "adir/file.jpg"</code></pre>
<p>A <code>?</code> matches any character except a slash <code>/</code>.</p>
<pre><code>l?ss - matches "less"
- matches "lass"
- doesn't match "floss"</code></pre>
<p>A <code>[</code> and <code>]</code> together make a a character class, such as <code>[a-z]</code> or <code>[aeiou]</code> or <code>[[:alpha:]]</code>. See the <ahref="https://golang.org/pkg/regexp/syntax/">go regexp docs</a> for more info on these.</p>
<pre><code>h[ae]llo - matches "hello"
- matches "hallo"
- doesn't match "hullo"</code></pre>
<p>A <code>{</code> and <code>}</code> define a choice between elements. It should contain a comma seperated list of patterns, any of which might match. These patterns can contain wildcards.</p>
<p>Note also that rclone filter globs can only be used in one of the filter command line flags, not in the specification of the remote, so <code>rclone copy "remote:dir*.jpg" /path/to/dir</code> won't work - what is required is <code>rclone --include "*.jpg" copy remote:dir /path/to/dir</code></p>
<h3id="directories">Directories</h3>
<p>Rclone keeps track of directories that could match any file patterns.</p>
<p>Directory matches are <strong>only</strong> used to optimise directory access patterns - you must still match the files that you want to match. Directory matches won't optimise anything on bucket based remotes (eg s3, swift, google compute storage, b2) which don't have a concept of directory.</p>
<p>Rclone maintains a combined list of include rules and exclude rules.</p>
<p>Each file is matched in order, starting from the top, against the rule in the list until it finds a match. The file is then included or excluded according to the rule type.</p>
<p>If the matcher fails to find a match after testing against all the entries in the list then the path is included.</p>
<p>A similar process is done on directory entries before recursing into them. This only works on remotes which have a concept of directory (Eg local, google drive, onedrive, amazon drive) and not on bucket based remotes (eg s3, swift, google compute storage, b2).</p>
<p>You can repeat the following options to add more than one rule of that type.</p>
<ul>
<li><code>--include</code></li>
<li><code>--include-from</code></li>
<li><code>--exclude</code></li>
<li><code>--exclude-from</code></li>
<li><code>--filter</code></li>
<li><code>--filter-from</code></li>
</ul>
<p>Note that all the options of the same type are processed together in the order above, regardless of what order they were placed on the command line.</p>
<p>So all <code>--include</code> options are processed first in the order they appeared on the command line, then all <code>--include-from</code> options etc.</p>
<p>To mix up the order includes and excludes, the <code>--filter</code> flag can be used.</p>
<p>Prepare a file like this <code>exclude-file.txt</code></p>
<pre><code># a sample exclude rule file
*.bak
file2.jpg</code></pre>
<p>Then use as <code>--exclude-from exclude-file.txt</code>. This will sync all files except those ending in <code>bak</code> and <code>file2.jpg</code>.</p>
<p>This is useful if you have a lot of rules.</p>
<h3id="include---include-files-matching-pattern"><code>--include</code> - Include files matching pattern</h3>
<p>Add a single include rule with <code>--include</code>.</p>
<p>Eg <code>--include *.{png,jpg}</code> to include all <code>png</code> and <code>jpg</code> files in the backup and no others.</p>
<p>This adds an implicit <code>--exclude *</code> at the very end of the filter list. This means you can mix <code>--include</code> and <code>--include-from</code> with the other filters (eg <code>--exclude</code>) but you must include all the files you want in the include statement. If this doesn't provide enough flexibility then you must use <code>--filter-from</code>.</p>
<h3id="include-from---read-include-patterns-from-file"><code>--include-from</code> - Read include patterns from file</h3>
<p>Prepare a file like this <code>include-file.txt</code></p>
<pre><code># a sample include rule file
*.jpg
*.png
file2.avi</code></pre>
<p>Then use as <code>--include-from include-file.txt</code>. This will sync all <code>jpg</code>, <code>png</code> files and <code>file2.avi</code>.</p>
<p>This is useful if you have a lot of rules.</p>
<p>This adds an implicit <code>--exclude *</code> at the very end of the filter list. This means you can mix <code>--include</code> and <code>--include-from</code> with the other filters (eg <code>--exclude</code>) but you must include all the files you want in the include statement. If this doesn't provide enough flexibility then you must use <code>--filter-from</code>.</p>
<h3id="filter---add-a-file-filtering-rule"><code>--filter</code> - Add a file-filtering rule</h3>
<p>This can be used to add a single include or exclude rule. Include rules start with <code>+</code> and exclude rules start with <code>-</code>. A special rule called <code>!</code> can be used to clear the existing rules.</p>
<p>Prepare a file like this <code>filter-file.txt</code></p>
<pre><code># a sample exclude rule file
- secret*.jpg
+ *.jpg
+ *.png
+ file2.avi
# exclude everything else
- *</code></pre>
<p>Then use as <code>--filter-from filter-file.txt</code>. The rules are processed in the order that they are defined.</p>
<p>This example will include all <code>jpg</code> and <code>png</code> files, exclude any files matching <code>secret*.jpg</code> and include <code>file2.avi</code>. Everything else will be excluded from the sync.</p>
<h3id="files-from---read-list-of-source-file-names"><code>--files-from</code> - Read list of source-file names</h3>
<p>This reads a list of file names from the file passed in and <strong>only</strong> these files are transferred. The filtering rules are ignored completely if you use this option.</p>
<p>Prepare a file like this <code>files-from.txt</code></p>
<pre><code># comment
file1.jpg
file2.jpg</code></pre>
<p>Then use as <code>--files-from files-from.txt</code>. This will only transfer <code>file1.jpg</code> and <code>file2.jpg</code> providing they exist.</p>
<p>For example, let's say you had a few files you want to back up regularly with these absolute paths:</p>
<pre><code>/home/user1/important
/home/user1/dir/file
/home/user2/stuff</code></pre>
<p>To copy these you'd find a common subdirectory - in this case <code>/home</code> and put the remaining files in <code>files-from.txt</code> with or without leading <code>/</code>, eg</p>
<pre><code>user1/important
user1/dir/file
user2/stuff</code></pre>
<p>You could then copy these to a remote like this</p>
<h3id="min-size---dont-transfer-any-file-smaller-than-this"><code>--min-size</code> - Don't transfer any file smaller than this</h3>
<p>This option controls the minimum size file which will be transferred. This defaults to <code>kBytes</code> but a suffix of <code>k</code>, <code>M</code>, or <code>G</code> can be used.</p>
<p>For example <code>--min-size 50k</code> means no files smaller than 50kByte will be transferred.</p>
<h3id="max-size---dont-transfer-any-file-larger-than-this"><code>--max-size</code> - Don't transfer any file larger than this</h3>
<p>This option controls the maximum size file which will be transferred. This defaults to <code>kBytes</code> but a suffix of <code>k</code>, <code>M</code>, or <code>G</code> can be used.</p>
<p>For example <code>--max-size 1G</code> means no files larger than 1GByte will be transferred.</p>
<h3id="max-age---dont-transfer-any-file-older-than-this"><code>--max-age</code> - Don't transfer any file older than this</h3>
<p>This option controls the maximum age of files to transfer. Give in seconds or with a suffix of:</p>
<ul>
<li><code>ms</code> - Milliseconds</li>
<li><code>s</code> - Seconds</li>
<li><code>m</code> - Minutes</li>
<li><code>h</code> - Hours</li>
<li><code>d</code> - Days</li>
<li><code>w</code> - Weeks</li>
<li><code>M</code> - Months</li>
<li><code>y</code> - Years</li>
</ul>
<p>For example <code>--max-age 2d</code> means no files older than 2 days will be transferred.</p>
<h3id="min-age---dont-transfer-any-file-younger-than-this"><code>--min-age</code> - Don't transfer any file younger than this</h3>
<p>This option controls the minimum age of files to transfer. Give in seconds or with a suffix (see <code>--max-age</code> for list of suffixes)</p>
<p>For example <code>--min-age 2d</code> means no files younger than 2 days will be transferred.</p>
<h3id="delete-excluded---delete-files-on-dest-excluded-from-sync"><code>--delete-excluded</code> - Delete files on dest excluded from sync</h3>
<p><strong>Important</strong> this flag is dangerous - use with <code>--dry-run</code> and <code>-v</code> first.</p>
<p>When doing <code>rclone sync</code> this will delete any files which are excluded from the sync on the destination.</p>
<p>If for example you did a sync from <code>A</code> to <code>B</code> without the <code>--min-size 50k</code> flag</p>
<pre><code>rclone sync A: B:</code></pre>
<p>Then you repeated it like this with the <code>--delete-excluded</code></p>
<p>The examples above may not work verbatim in your shell as they have shell metacharacters in them (eg <code>*</code>), and may require quoting.</p>
<p>Eg linux, OSX</p>
<ul>
<li><code>--include \*.jpg</code></li>
<li><code>--include '*.jpg'</code></li>
<li><code>--include='*.jpg'</code></li>
</ul>
<p>In Windows the expansion is done by the command not the shell so this should work fine</p>
<ul>
<li><code>--include *.jpg</code></li>
</ul>
<h1id="overview-of-cloud-storage-systems">Overview of cloud storage systems</h1>
<p>Each cloud storage system is slighly different. Rclone attempts to provide a unified interface to them, but some underlying differences show through.</p>
<h2id="features">Features</h2>
<p>Here is an overview of the major features of each cloud storage system.</p>
<p>The cloud storage system supports various hash types of the objects.<br/>
The hashes are used when transferring data as an integrity check and can be specifically used with the <code>--checksum</code> flag in syncs and in the <code>check</code> command.</p>
<p>To use the checksum checks between filesystems they must support a common hash type.</p>
<h3id="modtime">ModTime</h3>
<p>The cloud storage system supports setting modification times on objects. If it does then this enables a using the modification times as part of the sync. If not then only the size will be checked by default, though the MD5SUM can be checked with the <code>--checksum</code> flag.</p>
<p>All cloud storage systems support some kind of date on the object and these will be set when transferring from the cloud storage system.</p>
<h3id="case-insensitive">Case Insensitive</h3>
<p>If a cloud storage systems is case sensitive then it is possible to have two files which differ only in case, eg <code>file.txt</code> and <code>FILE.txt</code>. If a cloud storage system is case insensitive then that isn't possible.</p>
<p>This can cause problems when syncing between a case insensitive system and a case sensitive system. The symptom of this is that no matter how many times you run the sync it never completes fully.</p>
<p>The local filesystem may or may not be case sensitive depending on OS.</p>
<ul>
<li>Windows - usually case insensitive, though case is preserved</li>
<li>OSX - usually case insensitive, though it is possible to format case sensitive</li>
<li>Linux - usually case sensitive, but there are case insensitive file systems (eg FAT formatted USB keys)</li>
</ul>
<p>Most of the time this doesn't cause any problems as people tend to avoid files whose name differs only by case even on case sensitive systems.</p>
<h3id="duplicate-files">Duplicate files</h3>
<p>If a cloud storage system allows duplicate files then it can have two objects with the same name.</p>
<p>MIME types (also known as media types) classify types of documents using a simple text classification, eg <code>text/html</code> or <code>application/pdf</code>.</p>
<p>Some cloud storage systems support reading (<code>R</code>) the MIME type of objects and some support writing (<code>W</code>) the MIME type of objects.</p>
<p>The MIME type can be important if you are serving files directly to HTTP from the storage system.</p>
<p>If you are copying from a remote which supports reading (<code>R</code>) to a remote which supports writing (<code>W</code>) then rclone will preserve the MIME types. Otherwise they will be guessed from the extension, or the remote itself may assign the MIME type.</p>
<h2id="optional-features">Optional Features</h2>
<p>All the remotes support a basic set of features, but there are some optional features supported by some remotes used to make some operations more efficient.</p>
<p>This deletes a directory quicker than just deleting all the files in the directory.</p>
<p>† Note Swift and Hubic implement this in order to delete directory markers but they don't actually have a quicker way of deleting files other than deleting them individually.</p>
<h3id="copy">Copy</h3>
<p>Used when copying an object to and from the same remote. This known as a server side copy so you can copy a file without downloading it and uploading it again. It is used if you use <code>rclone copy</code> or <code>rclone move</code> if the remote doesn't support <code>Move</code> directly.</p>
<p>If the server doesn't support <code>Copy</code> directly then for copy operations the file is downloaded then re-uploaded.</p>
<h3id="move">Move</h3>
<p>Used when moving/renaming an object on the same remote. This is known as a server side move of a file. This is used in <code>rclone move</code> if the server doesn't support <code>DirMove</code>.</p>
<p>If the server isn't capable of <code>Move</code> then rclone simulates it with <code>Copy</code> then delete. If the server doesn't support <code>Copy</code> then rclone will download the file and re-upload it.</p>
<h3id="dirmove">DirMove</h3>
<p>This is used to implement <code>rclone move</code> to move a directory if possible. If it isn't then it will use <code>Move</code> on each file (which falls back to <code>Copy</code> then download and upload - see <code>Move</code> section).</p>
<h3id="cleanup">CleanUp</h3>
<p>This is used for emptying the trash for a remote by <code>rclone cleanup</code>.</p>
<p>If the server can't do <code>CleanUp</code> then <code>rclone cleanup</code> will return an error.</p>
<p>Paths are specified as <code>drive:path</code></p>
<p>Drive paths may be as deep as required, eg <code>drive:directory/subdirectory</code>.</p>
<p>The initial setup for drive involves getting a token from Google drive which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>Here is an example of how to make a remote called <code>remote</code>. First run:</p>
<pre><code> rclone config</code></pre>
<p>This will guide you through an interactive setup process:</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Google if you use auto config mode. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall, or use manual mode.</p>
<p>You can then use it like this,</p>
<p>List directories in top level of your drive</p>
<pre><code>rclone lsd remote:</code></pre>
<p>List all the files in your drive</p>
<pre><code>rclone ls remote:</code></pre>
<p>To copy a local directory to a drive directory called backup</p>
<p>Google drive stores modification times accurate to 1 ms.</p>
<h3id="revisions">Revisions</h3>
<p>Google drive stores revisions of files. When you upload a change to an existing file to google drive using rclone it will create a new revision of that file.</p>
<p>Revisions follow the standard google policy which at time of writing was</p>
<ul>
<li>They are deleted after 30 days or 100 revisions (whatever comes first).</li>
<li>They do not count towards a user storage quota.</li>
</ul>
<h3id="deleting-files">Deleting files</h3>
<p>By default rclone will delete files permanently when requested. If sending them to the trash is required instead then use the <code>--drive-use-trash</code> flag.</p>
<h3id="specific-options">Specific options</h3>
<p>Here are the command line options specific to this cloud storage system.</p>
<p>Only consider files owned by the authenticated user. Requires that --drive-full-list=true (default).</p>
<h4id="drive-formats">--drive-formats</h4>
<p>Google documents can only be exported from Google drive. When rclone downloads a Google doc it chooses a format to download depending upon this setting.</p>
<p>By default the formats are <code>docx,xlsx,pptx,svg</code> which are a sensible default for an editable document.</p>
<p>When choosing a format, rclone runs down the list provided in order and chooses the first file format the doc can be exported as from the list. If the file can't be exported to a format on the formats list, then rclone will choose a format from the default list.</p>
<p>If you prefer an archive copy then you might use <code>--drive-formats pdf</code>, or if you prefer openoffice/libreoffice formats you might use <code>--drive-formats ods,odt,odp</code>.</p>
<p>Note that rclone adds the extension to the google doc, so if it is calles <code>My Spreadsheet</code> on google docs, it will be exported as <code>My Spreadsheet.xlsx</code> or <code>My Spreadsheet.pdf</code> etc.</p>
<p>Here are the possible extensions with their corresponding mime types.</p>
<p>Drive has quite a lot of rate limiting. This causes rclone to be limited to transferring about 2 files per second only. Individual files may be transferred much faster at 100s of MBytes/s but lots of small files can take a long time.</p>
<h3id="making-your-own-client_id">Making your own client_id</h3>
<p>When you use rclone with Google drive in its default configuration you are using rclone's client_id. This is shared between all the rclone users. There is a global rate limit on the number of queries per second that each client_id can do set by Google. rclone already has a high quota and I will continue to make sure it is high enough by contacting Google.</p>
<p>However you might find you get better performance making your own client_id if you are a heavy user. Or you may not depending on exactly how Google have been raising rclone's rate limit.</p>
<p>Here is how to create your own Google Drive client ID for rclone:</p>
<olstyle="list-style-type: decimal">
<li><p>Log into the <ahref="https://console.developers.google.com/">Google API Console</a> with your Google account. It doesn't matter what Google account you use. (It need not be the same account as the Google Drive you want to access)</p></li>
<li><p>Select a project or create a new project.</p></li>
<li><p>Under Overview, Google APIs, Google Apps APIs, click "Drive API", then "Enable".</p></li>
<li><p>Click "Credentials" in the left-side panel (not "Go to credentials", which opens the wizard), then "Create credentials", then "OAuth client ID". It will prompt you to set the OAuth consent screen product name, if you haven't set one already.</p></li>
<li><p>Choose an application type of "other", and click "Create". (the default name is fine)</p></li>
<li><p>It will show you a client ID and client secret. Use these values in rclone config to add a new remote or edit an existing remote.</p></li>
</ol>
<p>(Thanks to <spanclass="citation">@balazer</span> on github for these instructions.)</p>
<p>Paths are specified as <code>remote:bucket</code> (or <code>remote:</code> for the <code>lsd</code> command.) You may put subdirectories in too, eg <code>remote:bucket/path/to/dir</code>.</p>
<p>Here is an example of making an s3 configuration. First run</p>
<pre><code>rclone config</code></pre>
<p>This will guide you through an interactive setup process.</p>
5 / Google Cloud Storage (this is not Google Drive)
\ "google cloud storage"
6 / Google Drive
\ "drive"
7 / Hubic
\ "hubic"
8 / Local Disk
\ "local"
9 / Microsoft OneDrive
\ "onedrive"
10 / Openstack Swift (Rackspace Cloud Files, Memset Memstore, OVH)
\ "swift"
11 / Yandex Disk
\ "yandex"
Storage> 2
Get AWS credentials from runtime (environment variables or EC2 meta data if no env vars). Only applies if access_key_id and secret_access_key is blank.
Choose a number from below, or type in your own value
1 / Enter AWS credentials in the next step
\ "false"
2 / Get AWS credentials from the environment (env vars or IAM)
\ "true"
env_auth> 1
AWS Access Key ID - leave blank for anonymous access or runtime credentials.
access_key_id> access_key
AWS Secret Access Key (password) - leave blank for anonymous access or runtime credentials.
<p>The modified time is stored as metadata on the object as <code>X-Amz-Meta-Mtime</code> as floating point since the epoch accurate to 1 ns.</p>
<h3id="multipart-uploads">Multipart uploads</h3>
<p>rclone supports multipart uploads with S3 which means that it can upload files bigger than 5GB. Note that files uploaded with multipart upload don't have an MD5SUM.</p>
<h3id="buckets-and-regions">Buckets and Regions</h3>
<p>With Amazon S3 you can list buckets (<code>rclone lsd</code>) using any region, but you can only access the content of a bucket from the region it was created in. If you attempt to access a bucket from the wrong region, you will get an error, <code>incorrect region, the bucket is not in 'XXX' region</code>.</p>
<p>There are two ways to supply <code>rclone</code> with a set of AWS credentials. In order of precedence:</p>
<ul>
<li>Directly in the rclone configuration file (as configured by <code>rclone config</code>)</li>
<li>set <code>access_key_id</code> and <code>secret_access_key</code></li>
<li>Runtime configuration:</li>
<li>set <code>env_auth</code> to <code>true</code> in the config file</li>
<li>Exporting the following environment variables before running <code>rclone</code>
<ul>
<li>Access Key ID: <code>AWS_ACCESS_KEY_ID</code> or <code>AWS_ACCESS_KEY</code></li>
<li>Secret Access Key: <code>AWS_SECRET_ACCESS_KEY</code> or <code>AWS_SECRET_KEY</code></li>
</ul></li>
<li>Running <code>rclone</code> on an EC2 instance with an IAM role</li>
</ul>
<p>If none of these option actually end up providing <code>rclone</code> with AWS credentials then S3 interaction will be non-authenticated (see below).</p>
Get AWS credentials from runtime (environment variables or EC2 meta data if no env vars). Only applies if access_key_id and secret_access_key is blank.
Choose a number from below, or type in your own value
* Enter AWS credentials in the next step
1) false
* Get AWS credentials from the environment (env vars or IAM)
2) true
env_auth> 1
AWS Access Key ID - leave blank for anonymous access or runtime credentials.
access_key_id>
AWS Secret Access Key (password) - leave blank for anonymous access or runtime credentials.
<p>You will be able to list and copy data but not upload it.</p>
<h3id="ceph">Ceph</h3>
<p>Ceph is an object storage system which presents an Amazon S3 interface.</p>
<p>To use rclone with ceph, you need to set the following parameters in the config.</p>
<pre><code>access_key_id = Whatever
secret_access_key = Whatever
endpoint = https://ceph.endpoint.goes.here/
region = other-v2-signature</code></pre>
<p>Note also that Ceph sometimes puts <code>/</code> in the passwords it gives users. If you read the secret access key using the command line tools you will get a JSON blob with the <code>/</code> escaped as <code>\/</code>. Make sure you only write <code>/</code> in the secret access key.</p>
<p>Eg the dump from Ceph looks something like this (irrelevant keys removed).</p>
<pre><code>{
"user_id": "xxx",
"display_name": "xxxx",
"keys": [
{
"user": "xxx",
"access_key": "xxxxxx",
"secret_key": "xxxxxx\/xxxx"
}
],
}</code></pre>
<p>Because this is a json dump, it is encoding the <code>/</code> as <code>\/</code>, so if you use the secret key as <code>xxxxxx/xxxx</code> it will work fine.</p>
<p>Minio doesn't support all the features of S3 yet. In particular it doesn't support MD5 checksums (ETags) or metadata. This means rclone can't check MD5SUMs or store the modified date. However you can work around this with the <code>--size-only</code> flag of rclone.</p>
<p>So once set up, for example to copy files into a bucket</p>
<p>Swift refers to <ahref="http://www.openstack.org/software/openstack-storage/">Openstack Object Storage</a>. Commercial implementations of that being:</p>
<p>Paths are specified as <code>remote:container</code> (or <code>remote:</code> for the <code>lsd</code> command.) You may put subdirectories in too, eg <code>remote:container/path/to/dir</code>.</p>
<p>Here is an example of making a swift configuration. First run</p>
<pre><code>rclone config</code></pre>
<p>This will guide you through an interactive setup process.</p>
echo "Please enter your OpenStack Password: "
read -sr OS_PASSWORD_INPUT
export OS_PASSWORD=$OS_PASSWORD_INPUT
export OS_REGION_NAME="SBG1"
if [ -z "$OS_REGION_NAME" ]; then unset OS_REGION_NAME; fi</code></pre>
<p>The config file needs to look something like this where <code>$OS_USERNAME</code> represents the value of the <code>OS_USERNAME</code> variable - <code>123abc567xy</code> in the example above.</p>
<pre><code>[remote]
type = swift
user = $OS_USERNAME
key = $OS_PASSWORD
auth = $OS_AUTH_URL
tenant = $OS_TENANT_NAME</code></pre>
<p>Note that you may (or may not) need to set <code>region</code> too - try without first.</p>
<p>The Swift API doesn't return a correct MD5SUM for segmented files (Dynamic or Static Large Objects) so rclone won't check or use the MD5SUM for these.</p>
<h4id="rclone-gives-failed-to-create-file-system-for-remote-bad-request">Rclone gives Failed to create file system for "remote:": Bad Request</h4>
<p>Due to an oddity of the underlying swift library, it gives a "Bad Request" error rather than a more sensible error when the authentication fails for Swift.</p>
<p>So this most likely means your username / password is wrong. You can investigate further with the <code>--dump-bodies</code> flag.</p>
<p>Paths are specified as <code>remote:path</code></p>
<p>Dropbox paths may be as deep as required, eg <code>remote:directory/subdirectory</code>.</p>
<p>The initial setup for dropbox involves getting a token from Dropbox which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>Here is an example of how to make a remote called <code>remote</code>. First run:</p>
<pre><code> rclone config</code></pre>
<p>This will guide you through an interactive setup process:</p>
<p>Note that Dropbox is case insensitive so you can't have a file called "Hello.doc" and one called "hello.doc".</p>
<p>There are some file names such as <code>thumbs.db</code> which Dropbox can't store. There is a full list of them in the <ahref="https://www.dropbox.com/en/help/145">"Ignored Files" section of this document</a>. Rclone will issue an error message <code>File name disallowed - not uploading</code> if it attempt to upload one of those file names, but the sync won't fail.</p>
<p>If you have more than 10,000 files in a directory then <code>rclone purge dropbox:dir</code> will return the error <code>Failed to purge: There are too many files involved in this operation</code>. As a work-around do an <code>rclone delete dropbix:dir</code> followed by an <code>rclone rmdir dropbox:dir</code>.</p>
<p>Paths are specified as <code>remote:bucket</code> (or <code>remote:</code> for the <code>lsd</code> command.) You may put subdirectories in too, eg <code>remote:bucket/path/to/dir</code>.</p>
<p>The initial setup for google cloud storage involves getting a token from Google Cloud Storage which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>Here is an example of how to make a remote called <code>remote</code>. First run:</p>
<pre><code> rclone config</code></pre>
<p>This will guide you through an interactive setup process:</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Google if you use auto config mode. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall, or use manual mode.</p>
<p>This remote is called <code>remote</code> and can now be used like this</p>
<p>You can set up rclone with Google Cloud Storage in an unattended mode, i.e. not tied to a specific end-user Google account. This is useful when you want to synchronise files onto machines that don't have actively logged-in users, for example build machines.</p>
<p>To get credentials for Google Cloud Platform <ahref="https://cloud.google.com/iam/docs/service-accounts">IAM Service Accounts</a>, please head to the <ahref="https://console.cloud.google.com/permissions/serviceaccounts">Service Account</a> section of the Google Developer Console. Service Accounts behave just like normal <code>User</code> permissions in <ahref="https://cloud.google.com/storage/docs/access-control">Google Cloud Storage ACLs</a>, so you can limit their access (e.g. make them read only). After creating an account, a JSON file containing the Service Account's credentials will be downloaded onto your machines. These credentials are what rclone will use for authentication.</p>
<p>To use a Service Account instead of OAuth2 token flow, enter the path to your Service Account credentials at the <code>service_account_file</code> prompt and rclone won't use the browser based authentication flow.</p>
<p>Google google cloud storage stores md5sums natively and rclone stores modification times as metadata on the object, under the "mtime" key in RFC3339 format accurate to 1ns.</p>
<p>The initial setup for Amazon Drive involves getting a token from Amazon which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>See the <ahref="http://rclone.org/remote_setup/">remote setup docs</a> for how to set it up on a machine with no Internet browser available.</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Amazon. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall.</p>
<p>Once configured you can then use <code>rclone</code> like this,</p>
<p>Any files you delete with rclone will end up in the trash. Amazon don't provide an API to permanently delete files, nor to empty the trash, so you will have to do that with one of Amazon's apps or via the Amazon Drive website.</p>
<h3id="using-with-non-.com-amazon-accounts">Using with non <code>.com</code> Amazon accounts</h3>
<p>Let's say you usually use <code>amazon.co.uk</code>. When you authenticate with rclone it will take you to an <code>amazon.com</code> page to log in. Your <code>amazon.co.uk</code> email and password should work here just fine.</p>
<p>Files this size or more will be downloaded via their <code>tempLink</code>. This is to work around a problem with Amazon Drive which blocks downloads of files bigger than about 10GB. The default for this is 9GB which shouldn't need to be changed.</p>
<p>To download files above this threshold, rclone requests a <code>tempLink</code> which downloads the file through a temporary URL directly from the underlying S3 storage.</p>
<p>Sometimes Amazon Drive gives an error when a file has been fully uploaded but the file appears anyway after a little while. This happens sometimes for files over 1GB in size and nearly every time for files bigger than 10GB. This parameter controls the time rclone waits for the file to appear.</p>
<p>The default value for this parameter is 3 minutes per GB, so by default it will wait 3 minutes for every GB uploaded to see if the file appears.</p>
<p>You can disable this feature by setting it to 0. This may cause conflict errors as rclone retries the failed upload but the file will most likely appear correctly eventually.</p>
<p>These values were determined empirically by observing lots of uploads of big files for a range of file sizes.</p>
<p>Upload with the <code>-v</code> flag to see more info about what rclone is doing in this situation.</p>
<p>Note that Amazon Drive is case insensitive so you can't have a file called "Hello.doc" and one called "hello.doc".</p>
<p>Amazon Drive has rate limiting so you may notice errors in the sync (429 errors). rclone will automatically retry the sync up to 3 times by default (see <code>--retries</code> flag) which should hopefully work around this problem.</p>
<p>Amazon Drive has an internal limit of file sizes that can be uploaded to the service. This limit is not officially published, but all files larger than this will fail.</p>
<p>Unfortunatly there is no way for rclone to see that this failure is because of file size, so it will retry the operation, as any other failure. To avoid this problem, use <code>--max-size 50000M</code> option to limit the maximum size of uploaded files. Note that <code>--max-size</code> does not split files into segments, it only ignores files over this size.</p>
<h2id="microsoft-one-drive">Microsoft One Drive</h2>
<p>Paths are specified as <code>remote:path</code></p>
<p>Paths may be as deep as required, eg <code>remote:directory/subdirectory</code>.</p>
<p>The initial setup for One Drive involves getting a token from Microsoft which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>Here is an example of how to make a remote called <code>remote</code>. First run:</p>
<pre><code> rclone config</code></pre>
<p>This will guide you through an interactive setup process:</p>
<p>See the <ahref="http://rclone.org/remote_setup/">remote setup docs</a> for how to set it up on a machine with no Internet browser available.</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Microsoft. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall.</p>
<p>Once configured you can then use <code>rclone</code> like this,</p>
<p>List directories in top level of your One Drive</p>
<pre><code>rclone lsd remote:</code></pre>
<p>List all the files in your One Drive</p>
<pre><code>rclone ls remote:</code></pre>
<p>To copy a local directory to an One Drive directory called backup</p>
<h3id="modified-time-and-hashes">Modified time and hashes</h3>
<p>One Drive allows modification times to be set on objects accurate to 1 second. These will be used to detect whether objects need syncing or not.</p>
<p>One drive supports SHA1 type hashes, so you can use <code>--checksum</code> flag.</p>
<h3id="deleting-files-2">Deleting files</h3>
<p>Any files you delete with rclone will end up in the trash. Microsoft doesn't provide an API to permanently delete files, nor to empty the trash, so you will have to do that with one of Microsoft's apps or via the One Drive website.</p>
<p>Note that One Drive is case insensitive so you can't have a file called "Hello.doc" and one called "hello.doc".</p>
<p>Rclone only supports your default One Drive, and doesn't work with One Drive for business. Both these issues may be fixed at some point depending on user demand!</p>
<p>There are quite a few characters that can't be in One Drive file names. These can't occur on Windows platforms, but on non-Windows platforms they are common. Rclone will map these names to and from an identical looking unicode equivalent. For example if a file has a <code>?</code> in it will be mapped to <code>？</code> instead.</p>
<p>Paths are specified as <code>remote:path</code></p>
<p>Paths are specified as <code>remote:container</code> (or <code>remote:</code> for the <code>lsd</code> command.) You may put subdirectories in too, eg <code>remote:container/path/to/dir</code>.</p>
<p>The initial setup for Hubic involves getting a token from Hubic which you need to do in your browser. <code>rclone config</code> walks you through it.</p>
<p>Here is an example of how to make a remote called <code>remote</code>. First run:</p>
<pre><code> rclone config</code></pre>
<p>This will guide you through an interactive setup process:</p>
<p>See the <ahref="http://rclone.org/remote_setup/">remote setup docs</a> for how to set it up on a machine with no Internet browser available.</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Hubic. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall.</p>
<p>Once configured you can then use <code>rclone</code> like this,</p>
<p>List containers in the top level of your Hubic</p>
<pre><code>rclone lsd remote:</code></pre>
<p>List all the files in your Hubic</p>
<pre><code>rclone ls remote:</code></pre>
<p>To copy a local directory to an Hubic directory called backup</p>
<p>If you want the directory to be visible in the official <em>Hubic browser</em>, you need to copy your files to the <code>default</code> directory</p>
<p>This uses the normal OpenStack Swift mechanism to refresh the Swift API credentials and ignores the expires field returned by the Hubic API.</p>
<p>The Swift API doesn't return a correct MD5SUM for segmented files (Dynamic or Static Large Objects) so rclone won't check or use the MD5SUM for these.</p>
<p>B2 is <ahref="https://www.backblaze.com/b2/">Backblaze's cloud storage system</a>.</p>
<p>Paths are specified as <code>remote:bucket</code> (or <code>remote:</code> for the <code>lsd</code> command.) You may put subdirectories in too, eg <code>remote:bucket/path/to/dir</code>.</p>
<p>Here is an example of making a b2 configuration. First run</p>
<pre><code>rclone config</code></pre>
<p>This will guide you through an interactive setup process. You will need your account number (a short hex number) and key (a long hex number) which you can get from the b2 control panel.</p>
<p>The modified time is stored as metadata on the object as <code>X-Bz-Info-src_last_modified_millis</code> as milliseconds since 1970-01-01 in the Backblaze standard. Other tools should be able to use this as a modified time.</p>
<p>Modified times are used in syncing and are fully supported except in the case of updating a modification time on an existing object. In this case the object will be uploaded again as B2 doesn't have an API method to set the modification time independent of doing an upload.</p>
<p>Large files which are uploaded in chunks will store their SHA1 on the object as <code>X-Bz-Info-large_file_sha1</code> as recommended by Backblaze.</p>
<p>Backblaze recommends that you do lots of transfers simultaneously for maximum speed. In tests from my SSD equiped laptop the optimum setting is about <code>--transfers 32</code> though higher numbers may be used for a slight speed improvement. The optimum number for you may vary depending on your hardware, how big the files are, how much you want to load your computer, etc. The default of <code>--transfers 4</code> is definitely too low for Backblaze B2 though.</p>
<p>Note that uploading big files (bigger than 200 MB by default) will use a 96 MB RAM buffer by default. There can be at most <code>--transfers</code> of these in use at any moment, so this sets the upper limit on the memory used.</p>
<p>When rclone uploads a new version of a file it creates a <ahref="https://www.backblaze.com/b2/docs/file_versions.html">new version of it</a>. Likewise when you delete a file, the old version will still be available.</p>
<p>Old versions of files are visible using the <code>--b2-versions</code> flag.</p>
<p>If you wish to remove all the old versions then you can use the <code>rclone cleanup remote:bucket</code> command which will delete all the old versions of files, leaving the current ones intact. You can also supply a path and only old versions under that path will be deleted, eg <code>rclone cleanup remote:bucket/path/to/stuff</code>.</p>
<p>When you <code>purge</code> a bucket, the current and the old versions will be deleted then the bucket will be deleted.</p>
<p>However <code>delete</code> will cause the current versions of the files to become hidden old versions.</p>
<p>Here is a session showing the listing and and retreival of an old version followed by a <code>cleanup</code> of the old versions.</p>
<p>Show current version and all the versions with <code>--b2-versions</code> flag.</p>
<p>It is useful to know how many requests are sent to the server in different scenarios.</p>
<p>All copy commands send the following 4 requests:</p>
<pre><code>/b2api/v1/b2_authorize_account
/b2api/v1/b2_create_bucket
/b2api/v1/b2_list_buckets
/b2api/v1/b2_list_file_names</code></pre>
<p>The <code>b2_list_file_names</code> request will be sent once for every 1k files in the remote path, providing the checksum and modification time of the listed files. As of version 1.33 issue <ahref="https://github.com/ncw/rclone/issues/818">#818</a> causes extra requests to be sent when using B2 with Crypt. When a copy operation does not require any files to be uploaded, no more requests will be sent.</p>
<p>Uploading files that do not require chunking, will send 2 requests per file upload:</p>
<pre><code>/b2api/v1/b2_get_upload_url
/b2api/v1/b2_upload_file/</code></pre>
<p>Uploading files requiring chunking, will send 2 requests (one each to start and finish the upload) and another 2 requests for each chunk:</p>
<pre><code>/b2api/v1/b2_start_large_file
/b2api/v1/b2_get_upload_part_url
/b2api/v1/b2_upload_part/
/b2api/v1/b2_finish_large_file</code></pre>
<h3id="b2-with-crypt">B2 with crypt</h3>
<p>When using B2 with <code>crypt</code> files are encrypted into a temporary location and streamed from there. This is required to calculate the encrypted file's checksum before beginning the upload. On Windows the %TMPDIR% environment variable is used as the temporary location. If the file requires chunking, both the chunking and encryption will take place in memory.</p>
<p>When uploading large files chunk the file into this size. Note that these chunks are buffered in memory and there might a maximum of <code>--transfers</code> chunks in progress at once. 100,000,000 Bytes is the minimim size (default 96M).</p>
<p>Cutoff for switching to chunked upload (default 190.735 MiB == 200 MB). Files above this size will be uploaded in chunks of <code>--b2-chunk-size</code>.</p>
<p>This value should be set no larger than 4.657GiB (== 5GB) as this is the largest file size that can be uploaded.</p>
<p>These will be set in the <code>X-Bz-Test-Mode</code> header which is documented in the <ahref="https://www.backblaze.com/b2/docs/integration_checklist.html">b2 integrations checklist</a>.</p>
<h4id="b2-versions">--b2-versions</h4>
<p>When set rclone will show and act on older versions of files. For example</p>
<p>Listing without <code>--b2-versions</code></p>
<pre><code>$ rclone -q ls b2:cleanup-test
9 one.txt</code></pre>
<p>And with</p>
<pre><code>$ rclone -q --b2-versions ls b2:cleanup-test
9 one.txt
8 one-v2016-07-04-141032-000.txt
16 one-v2016-07-04-141003-000.txt
15 one-v2016-07-02-155621-000.txt</code></pre>
<p>Showing that the current version is unchanged but older versions can be seen. These have the UTC date that they were uploaded to the server to the nearest millisecond appended to them.</p>
<p>Note that when using <code>--b2-versions</code> no file write operations are permitted, so you can't upload files or delete them.</p>
<p>See the <ahref="http://rclone.org/remote_setup/">remote setup docs</a> for how to set it up on a machine with no Internet browser available.</p>
<p>Note that rclone runs a webserver on your local machine to collect the token as returned from Yandex Disk. This only runs from the moment it opens your browser to the moment you get back the verification code. This is on <code>http://127.0.0.1:53682/</code> and this it may require you to unblock it temporarily if you are running a host firewall.</p>
<p>Once configured you can then use <code>rclone</code> like this,</p>
<p>Modified times are supported and are stored accurate to 1 ns in custom metadata called <code>rclone_modified</code> in RFC3339 with nanoseconds format.</p>
<h3id="md5-checksums">MD5 checksums</h3>
<p>MD5 checksums are natively supported by Yandex Disk.</p>
<p>The <code>crypt</code> remote encrypts and decrypts another remote.</p>
<p>To use it first set up the underlying remote following the config instructions for that remote. You can also use a local pathname instead of a remote which will encrypt and decrypt from that directory which might be useful for encrypting onto a USB stick for example.</p>
<p>First check your chosen remote is working - we'll call it <code>remote:path</code> in these docs. Note that anything inside <code>remote:path</code> will be encrypted and anything outside won't. This means that if you are using a bucket based remote (eg S3, B2, swift) then you should probably put the bucket in the remote <code>s3:bucket</code>. If you just use <code>s3:</code> then rclone will make encrypted bucket names too (if using file name encryption) which may or may not be what you want.</p>
<p>Now configure <code>crypt</code> using <code>rclone config</code>. We will call this one <code>secret</code> to differentiate it from the <code>remote</code>.</p>
<pre><code>No remotes found - make a new one
n) New remote
s) Set configuration password
q) Quit config
n/s/q> n
name> secret
Type of storage to configure.
Choose a number from below, or type in your own value
1 / Amazon Drive
\ "amazon cloud drive"
2 / Amazon S3 (also Dreamhost, Ceph, Minio)
\ "s3"
3 / Backblaze B2
\ "b2"
4 / Dropbox
\ "dropbox"
5 / Encrypt/Decrypt a remote
\ "crypt"
6 / Google Cloud Storage (this is not Google Drive)
\ "google cloud storage"
7 / Google Drive
\ "drive"
8 / Hubic
\ "hubic"
9 / Local Disk
\ "local"
10 / Microsoft OneDrive
\ "onedrive"
11 / Openstack Swift (Rackspace Cloud Files, Memset Memstore, OVH)
<p><strong>Important</strong> The password is stored in the config file is lightly obscured so it isn't immediately obvious what it is. It is in no way secure unless you use config file encryption.</p>
<p>A long passphrase is recommended, or you can use a random one. Note that if you reconfigure rclone with the same passwords/passphrases elsewhere it will be compatible - all the secrets used are derived from those two passwords/passphrases.</p>
<p>Note that rclone does not encrypt * file length - this can be calcuated within 16 bytes * modification time - used for syncing</p>
<h2id="specifying-the-remote">Specifying the remote</h2>
<p>In normal use, make sure the remote has a <code>:</code> in. If you specify the remote without a <code>:</code> then rclone will use a local directory of that name. So if you use a remote of <code>/path/to/secret/files</code> then rclone will encrypt stuff to that directory. If you use a remote of <code>name</code> then rclone will put files in a directory called <code>name</code> in the current directory.</p>
<p>If you specify the remote as <code>remote:path/to/dir</code> then rclone will store encrypted files in <code>path/to/dir</code> on the remote. If you are using file name encryption, then when you save files to <code>secret:subdir/subfile</code> this will store them in the unencrypted path <code>path/to/dir</code> but the <code>subdir/subpath</code> bit will be encrypted.</p>
<p>Note that unless you want encrypted bucket names (which are difficult to manage because you won't know what directory they represent in web interfaces etc), you should probably specify a bucket, eg <code>remote:secretbucket</code> when using bucket based remotes such as S3, Swift, Hubic, B2, GCS.</p>
<p>Note that this retains the directory structure which means you can do this</p>
<pre><code>$ rclone -q ls secret:subdir
8 file2.txt
9 file3.txt
10 subsubdir/file4.txt</code></pre>
<p>If don't use file name encryption then the remote will look like this - note the <code>.bin</code> extensions added to prevent the cloud provider attempting to interpret the data.</p>
<pre><code>$ rclone -q ls remote:path
54 file0.txt.bin
57 subdir/file3.txt.bin
56 subdir/file2.txt.bin
58 subdir/subsubdir/file4.txt.bin
55 file1.txt.bin</code></pre>
<h3id="file-name-encryption-modes">File name encryption modes</h3>
<p>Here are some of the features of the file name encryption modes</p>
<p>Off * doesn't hide file names or directory structure * allows for longer file names (~246 characters) * can use sub paths and copy single files</p>
<p>Standard * file names encrypted * file names can't be as long (~156 characters) * can use sub paths and copy single files * directory structure visibile * identical files names will have identical uploaded names * can use shortcuts to shorten the directory recursion</p>
<p>Cloud storage systems have various limits on file name length and total path length which you are more likely to hit using "Standard" file name encryption. If you keep your file names to below 156 characters in length then you should be OK on all providers.</p>
<p>There may be an even more secure file name encryption mode in the future which will address the long file name problem.</p>
<p>The initial nonce is generated from the operating systems crypto strong random number genrator. The nonce is incremented for each chunk read making sure each nonce is unique for each block written. The chance of a nonce being re-used is miniscule. If you wrote an exabyte of data (10¹⁸ bytes) you would have a probability of approximately 2×10⁻³² of re-using a nonce.</p>
<h4id="chunk">Chunk</h4>
<p>Each chunk will contain 64kB of data, except for the last one which may have less data. The data chunk is in standard NACL secretbox format. Secretbox uses XSalsa20 and Poly1305 to encrypt and authenticate messages.</p>
<p>Each chunk contains:</p>
<ul>
<li>16 Bytes of Poly1305 authenticator</li>
<li>1 - 65536 bytes XSalsa20 encrypted data</li>
</ul>
<p>64k chunk size was chosen as the best performing chunk size (the authenticator takes too much time below this and the performance drops off due to cache effects above this). Note that these chunks are buffered in memory so they can't be too big.</p>
<p>This uses a 32 byte (256 bit key) key derived from the user password.</p>
<h4id="examples">Examples</h4>
<p>1 byte file will encrypt to</p>
<ul>
<li>32 bytes header</li>
<li>17 bytes data chunk</li>
</ul>
<p>49 bytes total</p>
<p>1MB (1048576 bytes) file will encrypt to</p>
<ul>
<li>32 bytes header</li>
<li>16 chunks of 65568 bytes</li>
</ul>
<p>1049120 bytes total (a 0.05% overhead). This is the overhead for big files.</p>
<h3id="name-encryption">Name encryption</h3>
<p>File names are encrypted segment by segment - the path is broken up into <code>/</code> separated strings and these are encrypted individually.</p>
<p>File segments are padded using using PKCS#7 to a multiple of 16 bytes before encryption.</p>
<p>They are then encrypted with EME using AES with 256 bit key. EME (ECB-Mix-ECB) is a wide-block encryption mode presented in the 2003 paper "A Parallelizable Enciphering Mode" by Halevi and Rogaway.</p>
<p>This makes for determinstic encryption which is what we want - the same filename must encrypt to the same thing otherwise we can't find it on the cloud storage system.</p>
<p>This means that</p>
<ul>
<li>filenames with the same name will encrypt the same</li>
<li>filenames which start the same won't have a common prefix</li>
</ul>
<p>This uses a 32 byte key (256 bits) and a 16 byte (128 bits) IV both of which are derived from the user password.</p>
<p>After encryption they are written out using a modified version of standard <code>base32</code> encoding as described in RFC4648. The standard encoding is modified in two ways:</p>
<ul>
<li>it becomes lower case (no-one likes upper case filenames!)</li>
<li>we strip the padding character <code>=</code></li>
</ul>
<p><code>base32</code> is used rather than the more efficient <code>base64</code> so rclone can be used on case insensitive remotes (eg Windows, Amazon Drive).</p>
<h3id="key-derivation">Key derivation</h3>
<p>Rclone uses <code>scrypt</code> with parameters <code>N=16384, r=8, p=1</code> with a an optional user supplied salt (password2) to derive the 32+32+16 = 80 bytes of key material required. If the user doesn't supply a salt then rclone uses an internal one.</p>
<p><code>scrypt</code> makes it impractical to mount a dictionary attack on rclone encrypted data. For full protection agains this you should always use a salt.</p>
<p>Will sync <code>/home/source</code> to <code>/tmp/destination</code></p>
<p>These can be configured into the config file for consistencies sake, but it is probably easier not to.</p>
<h3id="modified-time-7">Modified time</h3>
<p>Rclone reads and writes the modified time using an accuracy determined by the OS. Typically this is 1ns on Linux, 10 ns on Windows and 1 Second on OS X.</p>
<p>Filenames are expected to be encoded in UTF-8 on disk. This is the normal case for Windows and OS X.</p>
<p>There is a bit more uncertainty in the Linux world, but new distributions will have UTF-8 encoded files names. If you are using an old Linux filesystem with non UTF-8 file names (eg latin1) then you can use the <code>convmv</code> tool to convert the filesystem to UTF-8. This tool is available in most distributions' package managers.</p>
<p>If an invalid (non-UTF8) filename is read, the invalid caracters will be replaced with the unicode replacement character, '<27>'. <code>rclone</code> will emit a debug message in this case (use <code>-v</code> to see), eg</p>
<pre><code>Local file system at .: Replacing invalid UTF-8 characters in "gro\xdf"</code></pre>
<h3id="long-paths-on-windows">Long paths on Windows</h3>
<p>Rclone handles long paths automatically, by converting all paths to long <ahref="https://msdn.microsoft.com/en-us/library/windows/desktop/aa365247(v=vs.85).aspx#maxpath">UNC paths</a> which allows paths up to 32,767 characters.</p>
<p>This is why you will see that your paths, for instance <code>c:\files</code> is converted to the UNC path <code>\\?\c:\files</code> in the output, and <code>\\server\share</code> is converted to <code>\\?\UNC\server\share</code>.</p>
<p>However, in rare cases this may cause problems with buggy file system drivers like <ahref="https://github.com/ncw/rclone/issues/261">EncFS</a>. To disable UNC conversion globally, add this to your <code>.rclone.conf</code> file:</p>
<pre><code>[local]
nounc = true</code></pre>
<p>If you want to selectively disable UNC, you can add it to a separate entry like this:</p>
<p>This will use UNC paths on <code>c:\src</code> but not on <code>z:\dst</code>. Of course this will cause problems if the absolute path length of a file exceeds 258 characters on z, so only use this option if you have to.</p>
<p>This tells rclone to stay in the filesystem specified by the root and not to recurse into different file systems.</p>
<p>For example if you have a directory heirachy like this</p>
<pre><code>root
├── disk1 - disk1 mounted on the root
│ └── file3 - stored on disk1
├── disk2 - disk2 mounted on the root
│ └── file4 - stored on disk12
├── file1 - stored on the root disk
└── file2 - stored on the root disk</code></pre>
<p>Using <code>rclone --one-file-system copy root remote:</code> will only copy <code>file1</code> and <code>file2</code>. Eg</p>
<pre><code>$ rclone -q --one-file-system ls root
0 file1
0 file2</code></pre>
<pre><code>$ rclone -q ls root
0 disk1/file3
0 disk2/file4
0 file1
0 file2</code></pre>
<p><strong>NB</strong> Rclone (like most unix tools such as <code>du</code>, <code>rsync</code> and <code>tar</code>) treats a bind mount to the same device as being on the same filesystem.</p>
<p><strong>NB</strong> This flag is only available on Unix based systems. On systems where it isn't supported (eg Windows) it will not appear as an valid flag.</p>
<li>Oauth support redone - fix many bugs and improve usability
<ul>
<li>Use "golang.org/x/oauth2" as oauth libary of choice</li>
<li>Improve oauth usability for smoother initial signup</li>
<li>drive, googlecloudstorage: optionally use auto config for the oauth token</li>
</ul></li>
<li>Implement --dump-headers and --dump-bodies debug flags</li>
<li>Show multiple matched commands if abbreviation too short</li>
<li>Implement server side move where possible</li>
<li>local</li>
<li>Always use UNC paths internally on Windows - fixes a lot of bugs</li>
<li>dropbox</li>
<li>force use of our custom transport which makes timeouts work</li>
<li>Thanks to Klaus Post for lots of help with this release</li>
</ul></li>
<li>v1.19 - 2015-08-28
<ul>
<li>New features</li>
<li>Server side copies for s3/swift/drive/dropbox/gcs</li>
<li>Move command - uses server side copies if it can</li>
<li>Implement --retries flag - tries 3 times by default</li>
<li>Build for plan9/amd64 and solaris/amd64 too</li>
<li>Fixes</li>
<li>Make a current version download with a fixed URL for scripting</li>
<li>Ignore rmdir in limited fs rather than throwing error</li>
<li>dropbox</li>
<li>Increase chunk size to improve upload speeds massively</li>
<li>Issue an error message when trying to upload bad file name</li>
</ul></li>
<li>v1.18 - 2015-08-17
<ul>
<li>drive</li>
<li>Add <code>--drive-use-trash</code> flag so rclone trashes instead of deletes</li>
<li>Add "Forbidden to download" message for files with no downloadURL</li>
<li>dropbox</li>
<li>Remove datastore
<ul>
<li>This was deprecated and it caused a lot of problems</li>
<li>Modification times and MD5SUMs no longer stored</li>
</ul></li>
<li>Fix uploading files > 2GB</li>
<li>s3</li>
<li>use official AWS SDK from github.com/aws/aws-sdk-go</li>
<li><strong>NB</strong> will most likely require you to delete and recreate remote</li>
<li>enable multipart upload which enables files > 5GB</li>
<li>tested with Ceph / RadosGW / S3 emulation</li>
<li>many thanks to Sam Liston and Brian Haymore at the <ahref="https://www.chpc.utah.edu/">Utah Center for High Performance Computing</a> for a Ceph test account</li>
<li>misc</li>
<li>Show errors when reading the config file</li>
<li>Do not print stats in quiet mode - thanks Leonid Shalupov</li>
<li>Add FAQ</li>
<li>Fix created directories not obeying umask</li>
<li>dropbox: fix case insensitivity issues - thanks Leonid Shalupov</li>
</ul></li>
<li>v1.16 - 2015-06-09
<ul>
<li>Fix uploading big files which was causing timeouts or panics</li>
<li>Don't check md5sum after download with --size-only</li>
</ul></li>
<li>v1.15 - 2015-06-06
<ul>
<li>Add --checksum flag to only discard transfers by MD5SUM - thanks Alex Couper</li>
<li>Implement --size-only flag to sync on size not checksum & modtime</li>
<li>Expand docs and remove duplicated information</li>
<li>Document rclone's limitations with directories</li>
<li>dropbox: update docs about case insensitivity</li>
</ul></li>
<li>v1.14 - 2015-05-21
<ul>
<li>local: fix encoding of non utf-8 file names - fixes a duplicate file problem</li>
<li>drive: docs about rate limiting</li>
<li>google cloud storage: Fix compile after API change in "google.golang.org/api/storage/v1"</li>
</ul></li>
<li>v1.13 - 2015-05-10
<ul>
<li>Revise documentation (especially sync)</li>
<li>Implement --timeout and --conntimeout</li>
<li>s3: ignore etags from multipart uploads which aren't md5sums</li>
</ul></li>
<li>v1.12 - 2015-03-15
<ul>
<li>drive: Use chunked upload for files above a certain size</li>
<li>drive: add --drive-chunk-size and --drive-upload-cutoff parameters</li>
<li>drive: switch to insert from update when a failed copy deletes the upload</li>
<li>core: Log duplicate files if they are detected</li>
</ul></li>
<li>v1.11 - 2015-03-04
<ul>
<li>swift: add region parameter</li>
<li>drive: fix crash on failed to update remote mtime</li>
<li>In remote paths, change native directory separators to /</li>
<li>Add synchronization to ls/lsl/lsd output to stop corruptions</li>
<li>Ensure all stats/log messages to go stderr</li>
<li>Add --log-file flag to log everything (including panics) to file</li>
<li>Make it possible to disable stats printing with --stats=0</li>
<li>Implement --bwlimit to limit data transfer bandwidth</li>
</ul></li>
<li>v1.10 - 2015-02-12
<ul>
<li>s3: list an unlimited number of items</li>
<li>Fix getting stuck in the configurator</li>
</ul></li>
<li>v1.09 - 2015-02-07
<ul>
<li>windows: Stop drive letters (eg C:) getting mixed up with remotes (eg drive:)</li>
<li>local: Fix directory separators on Windows</li>
<li>drive: fix rate limit exceeded errors</li>
</ul></li>
<li>v1.08 - 2015-02-04
<ul>
<li>drive: fix subdirectory listing to not list entire drive</li>
<li>drive: Fix SetModTime</li>
<li>dropbox: adapt code to recent library changes</li>
</ul></li>
<li>v1.07 - 2014-12-23
<ul>
<li>google cloud storage: fix memory leak</li>
</ul></li>
<li>v1.06 - 2014-12-12
<ul>
<li>Fix "Couldn't find home directory" on OSX</li>
<li>swift: Add tenant parameter</li>
<li>Use new location of Google API packages</li>
</ul></li>
<li>v1.05 - 2014-08-09
<ul>
<li>Improved tests and consequently lots of minor fixes</li>
<li>core: Fix race detected by go race detector</li>
<li>core: Fixes after running errcheck</li>
<li>drive: reset root directory on Rmdir and Purge</li>
<li>fs: Document that Purger returns error on empty directory, test and fix</li>
<li>google cloud storage: fix ListDir on subdirectory</li>
<li>google cloud storage: re-read metadata in SetModTime</li>
<li>s3: make reading metadata more reliable to work around eventual consistency problems</li>
<li>s3: strip trailing / from ListDir()</li>
<li>swift: return directories without / in ListDir</li>
</ul></li>
<li>v1.04 - 2014-07-21
<ul>
<li>google cloud storage: Fix crash on Update</li>
</ul></li>
<li>v1.03 - 2014-07-20
<ul>
<li>swift, s3, dropbox: fix updated files being marked as corrupted</li>
<li>Make compile with go 1.1 again</li>
</ul></li>
<li>v1.02 - 2014-07-19
<ul>
<li>Implement Dropbox remote</li>
<li>Implement Google Cloud Storage remote</li>
<li>Verify Md5sums and Sizes after copies</li>
<li>Remove times from "ls" command - lists sizes only</li>
<li>Add add "lsl" - lists times and sizes</li>
<li>Add "md5sum" command</li>
</ul></li>
<li>v1.01 - 2014-07-04
<ul>
<li>drive: fix transfer of big files using up lots of memory</li>
</ul></li>
<li>v1.00 - 2014-07-03
<ul>
<li>drive: fix whole second dates</li>
</ul></li>
<li>v0.99 - 2014-06-26
<ul>
<li>Fix --dry-run not working</li>
<li>Make compatible with go 1.1</li>
</ul></li>
<li>v0.98 - 2014-05-30
<ul>
<li>s3: Treat missing Content-Length as 0 for some ceph installations</li>
<li>rclonetest: add file with a space in</li>
</ul></li>
<li>v0.97 - 2014-05-05
<ul>
<li>Implement copying of single files</li>
<li>s3 & swift: support paths inside containers/buckets</li>
</ul></li>
<li>v0.96 - 2014-04-24
<ul>
<li>drive: Fix multiple files of same name being created</li>
<li>drive: Use o.Update and fs.Put to optimise transfers</li>
<li>Add version number, -V and --version</li>
</ul></li>
<li>v0.95 - 2014-03-28
<ul>
<li>rclone.org: website, docs and graphics</li>
<li>drive: fix path parsing</li>
</ul></li>
<li>v0.94 - 2014-03-27
<ul>
<li>Change remote format one last time</li>
<li>GNU style flags</li>
</ul></li>
<li>v0.93 - 2014-03-16
<ul>
<li>drive: store token in config file</li>
<li>cross compile other versions</li>
<li>set strict permissions on config file</li>
</ul></li>
<li>v0.92 - 2014-03-15
<ul>
<li>Config fixes and --config option</li>
</ul></li>
<li>v0.91 - 2014-03-15
<ul>
<li>Make config file</li>
</ul></li>
<li>v0.90 - 2013-06-27
<ul>
<li>Project named rclone</li>
</ul></li>
<li>v0.00 - 2012-11-18
<ul>
<li>Project started</li>
</ul></li>
</ul>
<h2id="bugs-and-limitations">Bugs and Limitations</h2>
<h3id="empty-directories-are-left-behind-not-created">Empty directories are left behind / not created</h3>
<p>With remotes that have a concept of directory, eg Local and Drive, empty directories may be left behind, or not created when one was expected.</p>
<p>This is because rclone doesn't have a concept of a directory - it only works on objects. Most of the object storage systems can't actually store a directory so there is nowhere for rclone to store anything about directories.</p>
<p>You can work round this to some extent with the<code>purge</code> command which will delete everything under the path, <strong>inluding</strong> empty directories.</p>
<p>This may be fixed at some point in <ahref="https://github.com/ncw/rclone/issues/100">Issue #100</a></p>
<p>For the same reason as the above, rclone doesn't have a concept of a directory - it only works on objects, therefore it can't preserve the timestamps of directories.</p>
<h3id="do-all-cloud-storage-systems-support-all-rclone-commands">Do all cloud storage systems support all rclone commands</h3>
<p>Yes they do. All the rclone commands (eg <code>sync</code>, <code>copy</code> etc) will work on all the remote storage systems.</p>
<h3id="can-i-copy-the-config-from-one-machine-to-another">Can I copy the config from one machine to another</h3>
<p>Sure! Rclone stores all of its config in a single file. If you want to find this file, the simplest way is to run <code>rclone -h</code> and look at the help for the <code>--config</code> flag which will tell you where it is.</p>
<p>See the <ahref="http://rclone.org/remote_setup/">remote setup docs</a> for more info.</p>
<h3id="how-do-i-configure-rclone-on-a-remote-headless-box-with-no-browser">How do I configure rclone on a remote / headless box with no browser?</h3>
<p>This has now been documented in its own <ahref="http://rclone.org/remote_setup/">remote setup page</a>.</p>
<h3id="can-rclone-sync-directly-from-drive-to-s3">Can rclone sync directly from drive to s3</h3>
<p>Rclone can sync between two remote cloud storage systems just fine.</p>
<p>Note that it effectively downloads the file and uploads it again, so the node running rclone would need to have lots of bandwidth.</p>
<p>The syncs would be incremental (on a file by file basis).</p>
Server B> rclone copy /tmp/whatever remote:Backup</code></pre>
<p>The file names you upload from Server A and Server B should be different in this case, otherwise some file systems (eg Drive) may make duplicates.</p>
<h3id="why-doesnt-rclone-support-partial-transfers-binary-diffs-like-rsync">Why doesn't rclone support partial transfers / binary diffs like rsync?</h3>
<p>Rclone stores each file you transfer as a native object on the remote cloud storage system. This means that you can see the files you upload as expected using alternative access methods (eg using the Google Drive web interface). There is a 1:1 mapping between files on your hard disk and objects created in the cloud storage system.</p>
<p>Cloud storage systems (at least none I've come across yet) don't support partially uploading an object. You can't take an existing object, and change some bytes in the middle of it.</p>
<p>It would be possible to make a sync system which stored binary diffs instead of whole objects like rclone does, but that would break the 1:1 mapping of files on your hard disk to objects in the remote cloud storage system.</p>
<p>All the cloud storage systems support partial downloads of content, so it would be possible to make partial downloads work. However to make this work efficiently this would require storing a significant amount of metadata, which breaks the desired 1:1 mapping of files to objects.</p>
<h3id="can-rclone-do-bi-directional-sync">Can rclone do bi-directional sync?</h3>
<p>No, not at present. rclone only does uni-directional sync from A -> B. It may do in the future though since it has all the primitives - it just requires writing the algorithm to do it.</p>
<h3id="can-i-use-rclone-with-an-http-proxy">Can I use rclone with an HTTP proxy?</h3>
<p>Yes. rclone will use the environment variables <code>HTTP_PROXY</code>, <code>HTTPS_PROXY</code> and <code>NO_PROXY</code>, similar to cURL and other programs.</p>
<p><code>HTTPS_PROXY</code> takes precedence over <code>HTTP_PROXY</code> for https requests.</p>
<p>The environment values may be either a complete URL or a "host[:port]", in which case the "http" scheme is assumed.</p>
<p>The <code>NO_PROXY</code> allows you to disable the proxy for specific hosts. Hosts must be comma separated, and can contain domains or parts. For instance "foo.com" also matches "bar.foo.com".</p>
<h3id="rclone-gives-x509-failed-to-load-system-roots-and-no-roots-provided-error">Rclone gives x509: failed to load system roots and no roots provided error</h3>
<p>This means that <code>rclone</code> can't file the SSL root certificates. Likely you are running <code>rclone</code> on a NAS with a cut-down Linux OS, or possibly on Solaris.</p>
<h3id="rclone-gives-failed-to-load-config-file-function-not-implemented-error">Rclone gives Failed to load config file: function not implemented error</h3>
<p>Likely this means that you are running rclone on Linux version not supported by the go runtime, ie earlier than version 2.6.23.</p>
<p>See the <ahref="https://golang.org/doc/install">system requirements section in the go install docs</a> for full details.</p>
<h3id="all-my-uploaded-docxxlsxpptx-files-appear-as-archivezip">All my uploaded docx/xlsx/pptx files appear as archive/zip</h3>
<p>This is caused by uploading these files from a Windows computer which hasn't got the Microsoft Office suite installed. The easiest way to fix is to install the Word viewer and the Microsoft Office Compatibility Pack for Word, Excel, and PowerPoint 2007 and later versions' file formats</p>
