Frequently Asked Questions

Do all cloud storage systems support all rclone commands

Yes they do. All the rclone commands (e.g. sync, copy, etc.) will work on all the remote storage systems.

Can I copy the config from one machine to another

Sure! Rclone stores all of its config in a single file. If you want to find this file, run rclone config file which will tell you where it is.

See the remote setup docs for more info.

How do I configure rclone on a remote / headless box with no browser?

This has now been documented in its own remote setup page.

Can rclone sync directly from drive to s3

Rclone can sync between two remote cloud storage systems just fine.

Note that it effectively downloads the file and uploads it again, so the node running rclone would need to have lots of bandwidth.

The syncs would be incremental (on a file by file basis).


rclone sync --interactive drive:Folder s3:bucket

Using rclone from multiple locations at the same time

You can use rclone from multiple places at the same time if you choose different subdirectory for the output, e.g.

Server A> rclone sync --interactive /tmp/whatever remote:ServerA
Server B> rclone sync --interactive /tmp/whatever remote:ServerB

If you sync to the same directory then you should use rclone copy otherwise the two instances of rclone may delete each other's files, e.g.

Server A> rclone copy /tmp/whatever remote:Backup
Server B> rclone copy /tmp/whatever remote:Backup

The file names you upload from Server A and Server B should be different in this case, otherwise some file systems (e.g. Drive) may make duplicates.

Why doesn't rclone support partial transfers / binary diffs like rsync?

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 (e.g. 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.

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.

It would be possible to make a sync system which stored binary diffs like rsync does, instead of whole objects, but that would break the 1:1 mapping of files on your hard disk to objects in the remote cloud storage system.

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.

Can rclone do bi-directional sync?

Yes, since rclone v1.58.0, bidirectional cloud sync is available.

Can I use rclone with an HTTP proxy?

Yes. rclone will follow the standard environment variables for proxies, similar to cURL and other programs.

In general the variables are called http_proxy (for services reached over http) and https_proxy (for services reached over https). Most public services will be using https, but you may wish to set both.

The content of the variable is protocol://server:port. The protocol value is the one used to talk to the proxy server, itself, and is commonly either http or socks5.

Slightly annoyingly, there is no standard for the name; some applications may use http_proxy but another one HTTP_PROXY. The Go libraries used by rclone will try both variations, but you may wish to set all possibilities. So, on Linux, you may end up with code similar to

export http_proxy=http://proxyserver:12345
export https_proxy=$http_proxy
export HTTP_PROXY=$http_proxy
export HTTPS_PROXY=$http_proxy

Note: If the proxy server requires a username and password, then use

export http_proxy=http://username:password@proxyserver:12345
export https_proxy=$http_proxy
export HTTP_PROXY=$http_proxy
export HTTPS_PROXY=$http_proxy

The NO_PROXY allows you to disable the proxy for specific hosts. Hosts must be comma separated, and can contain domains or parts. For instance "" also matches "".


export no_proxy=localhost,,
export NO_PROXY=$no_proxy

Note that the FTP backend does not support ftp_proxy yet.

Rclone gives x509: failed to load system roots and no roots provided error

This means that rclone can't find the SSL root certificates. Likely you are running rclone on a NAS with a cut-down Linux OS, or possibly on Solaris.

Rclone (via the Go runtime) tries to load the root certificates from these places on Linux.

"/etc/ssl/certs/ca-certificates.crt", // Debian/Ubuntu/Gentoo etc.
"/etc/pki/tls/certs/ca-bundle.crt",   // Fedora/RHEL
"/etc/ssl/ca-bundle.pem",             // OpenSUSE
"/etc/pki/tls/cacert.pem",            // OpenELEC

So doing something like this should fix the problem. It also sets the time which is important for SSL to work properly.

mkdir -p /etc/ssl/certs/
curl -o /etc/ssl/certs/ca-certificates.crt
ntpclient -s -h

The two environment variables SSL_CERT_FILE and SSL_CERT_DIR, mentioned in the x509 package, provide an additional way to provide the SSL root certificates.

Note that you may need to add the --insecure option to the curl command line if it doesn't work without.

curl --insecure -o /etc/ssl/certs/ca-certificates.crt

Rclone gives Failed to load config file: function not implemented error

Likely this means that you are running rclone on Linux version not supported by the go runtime, ie earlier than version 2.6.23.

See the system requirements section in the go install docs for full details.

All my uploaded docx/xlsx/pptx files appear as archive/zip

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

tcp lookup no such host

This happens when rclone cannot resolve a domain. Please check that your DNS setup is generally working, e.g.

# both should print a long list of possible IP addresses
dig          # resolve using your default DNS
dig @ # resolve with Google's DNS server

If you are using systemd-resolved (default on Arch Linux), ensure it is at version 233 or higher. Previous releases contain a bug which causes not all domains to be resolved properly.

The Go resolver decision can be influenced with the GODEBUG=netdns=... environment variable. This also allows to resolve certain issues with DNS resolution. On Windows or MacOS systems, try forcing use of the internal Go resolver by setting GODEBUG=netdns=go at runtime. On other systems (Linux, *BSD, etc) try forcing use of the system name resolver by setting GODEBUG=netdns=cgo (and recompile rclone from source with CGO enabled if necessary). See the name resolution section in the go docs.

Failed to start auth webserver on Windows

Error: config failed to refresh token: failed to start auth webserver: listen tcp bind: An attempt was made to access a socket in a way forbidden by its access permissions.
yyyy/mm/dd hh:mm:ss Fatal error: config failed to refresh token: failed to start auth webserver: listen tcp bind: An attempt was made to access a socket in a way forbidden by its access permissions.

This is sometimes caused by the Host Network Service causing issues with opening the port on the host.

A simple solution may be restarting the Host Network Service with eg. Powershell

Restart-Service hns

The total size reported in the stats for a sync is wrong and keeps changing

It is likely you have more than 10,000 files that need to be synced. By default, rclone only gets 10,000 files ahead in a sync so as not to use up too much memory. You can change this default with the --max-backlog flag.

Rclone is using too much memory or appears to have a memory leak

Rclone is written in Go which uses a garbage collector. The default settings for the garbage collector mean that it runs when the heap size has doubled.

However it is possible to tune the garbage collector to use less memory by setting GOGC to a lower value, say export GOGC=20. This will make the garbage collector work harder, reducing memory size at the expense of CPU usage.

The most common cause of rclone using lots of memory is a single directory with thousands or millions of files in. Rclone has to load this entirely into memory as rclone objects. Each rclone object takes 0.5k-1k of memory.

Rclone changes fullwidth Unicode punctuation marks in file names

For example: On a Windows system, you have a file with name Test:1.jpg, where is the Unicode fullwidth colon symbol. When using rclone to copy this to your Google Drive, you will notice that the file gets renamed to Test:1.jpg, where : is the regular (halfwidth) colon.

The reason for such renames is the way rclone handles different restricted filenames on different cloud storage systems. It tries to avoid ambiguous file names as much and allow moving files between many cloud storage systems transparently, by replacing invalid characters with similar looking Unicode characters when transferring to one storage system, and replacing back again when transferring to a different storage system where the original characters are supported. When the same Unicode characters are intentionally used in file names, this replacement strategy leads to unwanted renames. Read more here.